Maternal Age

INFORMATION TO USERS

This manuscript h a s b een reproduced from the microfilm m aster. UMI films the text directly from th e original or copy submitted. Thus, som e th e sis and dissertation copies are in typewriter face, while others may be from an y type of com puter printer. The quality of this reproduction is dependent upon the quality o f the copy submitted. Broken or indistinct print, colored or poor quality illustrations

and photographs, print bleedthrough, substandard margins, and im proper alignm ent can adversely affect reproduction. In th e unlikely event th at the author did not send UMI a complete m anuscript and there are missing pages, th ese will be noted.

Also, if unauthorized

copyright material had to be removed, a note will indicate the deletion. O versize materials (e.g., m aps,

drawings, charts) are

reproduced

by

sectioning the original, beginning at th e upper left-hand com er and continuing from left to right in equal sections with small overlaps. Photographs included in the original manuscript have been reproduced xerographically in this copy.

H igher quality 6” x 9” black an d white

photographic prints are available for a n y photographs or illustrations appearing in this copy for an additional charge. C ontact UMI directly to order.

P roQ uest Information and Learning 300 North Z eeb Road, Ann Arbor, Ml 48106-1346 USA 800-521-0600

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.


THE UNIVERSITY OF CHICAGO

IMPACT OF ADVANCED MATERNAL AGE ON THE RISK O F ADVERSE BIRTH OUTCOMES IN THE UNITED STATES

A DISSERTATION SUBMITTED TO THE FACULTY OF THE IRVING B. HARRIS GRADUATE SCHOOL OF PUBLIC POLICY STUDIES IN CANDIDACY FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

BY BABAK KHOSHNOOD

CHICAGO, ILLINOIS JUNE 2001


UMI Number. 3010887

__ ___

__ @

UMI

UMI Microform 3010887 Copyright 2001 by Bell & Howell Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code.

Bell & Howell Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346


Copyright © 2001 by Babak Khoshnood All rights reserved.


DEDICATION: To my parents, Manuchehr Khoshnood and Parvaneh Khallaghi


TABLE OF CONTENTS LIST OF FIG U R E S...........................................................................................................vi LIST OF TABLES ........................................................................................................... vii ACKNOW LEDGEM ENTS............................................................................................. xi A BSTRA CT...................................................................................................................... xii Chapter 1. INTRODUCTION ................................................................................................. 1 Reference List........................................................................................................14 2. RECENT TRENDS IN MATERNAL AGE-SPECIFIC FERTILITY AND THE ROLE OF SOCIOLOGICAL, ECONOMIC AND PSYCHOLOGICAL FACTORS IN THESE TRENDS ..................................15 2.1 Overall trends in fertility in the United States, 1917-1997......... 17 2.2 Recent trends towards delayed childbearing in the United States, 1970-1997............................................................... 25 2.3 Sociological, economic and psychological factors in timing o f childbirth..........................................................................30 2.4 Summary o f reasons for/correlates o f delayed childbearing ......57 Reference List....................................................................................................... 59 3. EFFECTS OF SOCIOECONOMIC FACTORS ASSOCIATED WITH DELAYED CHILDBEARING ON THE RISK OF ADVERSE BIRTH OUTCOMES: REVIEW OF LITERATURE................................................... 63 3.1 3.2 3.3 Reference

Effects o f ethnicity on adverse birth outcom es............................. 64 Effects o f education and income .....................................................71 Effects o f occupation........................................................................76 List....................................................................................................... 81

4. EFFECTS OF ADVANCED MATERNAL AGE ON REPRODUCTIVE OUTCOMES: REVIEW OF LITERATURE......................................... .'........84 4.1 Effects o f age on fecundability (ability to conceive) ................... 85 4.2 Effects o f age on probability o f m iscarriage..................................89 4.3 Effects o f age on pregnancy, labor and delivery...........................90 4.4 Effects o f age on adverse birth outcom es...................................... 95 4.5 Summary o f age effects ......................................................... 101 Reference List..................................................................................................... 104

5. ANALYSES OF THE EFFECTS OF ADVANCED MATERNAL AGE ON BIRTH OUTCOM ES.................................................................................107 iv


5.1

questions o f interest for the empirical study in the dissertation.....................................................................................107 5.2 Conceptual framework for the empirical analyses......................113 5.3 Data Sources ...................................................................................115 5.4 Empirical specifications ................................................................121 5.5 Model specifications....................................................................... 135 Reference List..................................................................................................... 138 Main

6. RESULTS...........................................................................................................141 6.1 Effects o f age on low birth weight and preterm delivery — First births ......................................................................................143 6.2 Effects o f age on low birth weight and preterm delivery — Births o f second or higher order ................................................. 158 6.3 Effects o f age on Down syndrome: Effect modification by socioeconomic factors...................................................................174 6.4 Effects o f age on cause-specific infant m ortality........................ 178

7. CONCLUSIONS ...............................................................................................181 Reference List..................................................................................................... 199 Appendices A. TABLES AND FIGURES IN CHAPTER 2 ................................................ 202 B. TABLES AND FIGURES IN CHAPTER 6 ................................................ 216

v


LIST OF FIGURES A. FIGURES IN CHAPTER 2 Figure 2.1 - Crude birth rate (CBR) and total fertility rate (TFR), United States, 1917-1997 ...................................................................................................................... 203 Figure 2.2 - Age-specific fertility rates, United States, 1960-1997 ......................... 206 Figure 2.3 - Shifts in age-specific fertility rates, 1960, 1975, 1985, 1997 (Series 1-4), United S ta te s........................................................................................... 207 Figure 2.4 - Trends in age-specific fertility rates, first births, United States, 1970-1997 ...................................................................................................................... 210 Figure 2.5 - Shifts in age-specific fertility rates, first births, United States, 1975, 1980, 1985, 1997 (Series 1-4) ...........................................................................211 Figure 2.6 - Trends in age-specific fertility rates, parity > 2, United States, 1970-1997 ...................................................................................................................... 214 Figure 2.7 - Shifts in age-specific fertility, parity > 2, United States, 1975, 1980,1985,1997 (Series 1 -4 )....................................................................................... 215 B. FIGURES IN CHAPTER 6 Figure 6.1 - Risk for Down syndrome at birth by maternal age and race / ethnicity, United State, 1989 - 1991 ...........................................................................250 Figure 6.2 - Risk for Down syndrome at birth by maternal age and education non-Hispanic Whites - United State, 1989 —1991 .....................................................254 Figure 6.3 - Risk for Down syndrome at birth by maternal age and education African Americans - United State, 1989 - 1991 ........................................................ 255

vi


LIST OF TABLES A. TABLES IN CHAPTER 2 Table 2.1 - Total and Age-specific fertility rates by age o f mother, United States, 1960,1970-1997 ................................................................................... 204 Table 2.2 - Number o f first births by age o f the mother, United States, 1 9 7 0 -1 9 9 7 ..................................................................................................................... 208 Table 2.3 - First birth rates by age o f the mother, United States, 1 9 7 0 - 1997..................................................................................................................... 209 Table 2.4 - Percent distribution o f first births by age o f the mother, United States, 1970-1997............................................................................................... 212 Table 2.5 - Birth rates by age o f the mother, parity > 2, United States, 1 9 7 0 - 1977..................................................................................................................... 213 B. TABLES IN CHAPTER 6 Table 6 .1(a) - Total number and proportion o f live births by maternal age among four racial/ethnic groups - First Births - United States,1989 —1991 ............217 Table 6.1(b) - Total number and proportion o f live births by maternal age among four racial/ethnic groups - Parity > 1 - United States, 1989—1991 ...............218 Table 6.2(a) - Odds ratios for the association o f delayed childbearing (maternal age > 35 years) with low (< 2.5 kg) birth weight among four racial / ethnic groups - First Births - United States, 1989 —1991 ......................................... 219 Table 6.2(b) - Odds ratios for the association o f delayed childbearing (maternal age > 35 years) with low (< 2.5 kg) birth weight among four racial / ethnic groups - Parity > 1 - United States, 1989 - 1991 ............................................220 Table 6.2(c) - Risk differences for the association o f delayed childbearing (maternal age >35 years) with low (< 2.5 kg) birth weight among four racial / ethnic groups - First Births - United States, 1989 - 1991 ......................................... 221 Table 6.2(d) - Risk differences for the association o f delayed childbearing (maternal age >35 years) with low (< 2.5 kg) birth weight among four racial / ethnic groups - Parity > 1 - United States, 1989 —1991 ................................222 Table 6.2(e) - Odds ratios for the association o f delayed childbearing with low birth weight (< 2.5 kg) among four racial / ethnic groups First Births - United States, 1989 - 1991 .................................................................... 223 vii


Table 6.2(f) - Odds ratios for the association o f delayed childbearing with low birth weight (< 2.5 kg) among four racial / ethnic groups —Parity > 1 — United States, 1989-1991 .......................................................................................... 224 Table 6.3(a) - Attributable fractions o f low birth weight among the exposed (AFe) and attributable fractions in the population (AFp) associated with delayed childbearing (> 35 years) among four racial/ethnic groups First Births - United States, 1989 —1991 ....................................................................225 Table 6.3(b) - Attributable fractions o f low birth weight among the exposed (AFe) and attributable fractions in the population (AFp) associated with delayed childbearing (> 35 years) among four racial/ethnic groups Parity > 1 - United States, 1989 —1991 ....................................................................... 226 Table 6.3(c) - Attributable fractions o f low birth weight in the population (AFp) associated with delayed childbearing (>35 years) among four racial/ethnic groups - All Births - United States, 1989 - 1991 ...................................................... 227 Table 6.4(a) - Logistic regression analysis o f the association between delayed childbearing (> 35 years) and risk o f low (< 2.5 kg) birth weight among four racial/ethnic groups - First births - United States, 1989 - 1991 ................................ 228 Table 6.4(b) - Logistic regression analysis o f the association between delayed childbearing (> 35 years) and risk o f low (< 2.5 kg) birth weight among four racial/ethnic groups - Parity > 1 - United States, 1989 - 1991 .................................229 Table 6.4(c) - Logistic regression analysis o f the association between delayed childbearing and risk o f low (< 2.5 kg) birth weight among four racial/ethnic groups - First births - United States, 1989 —1991 ..................................................... 230 Table 6.4(d) - Logistic regression analysis o f the association between delayed childbearing and risk o f low (< 2.5 kg) birth weight among four racial/ethnic groups - Parity > 1 - United States, 1989 —1991 ...................................................... 231 Table 6.4(e) - Logistic regression analysis o f the association between delayed childbearing and risk o f low (< 2.5 kg) birth weight among four racial/ethnic groups - First births - United States, 1989 - 1991 [prenatal care not controlled] ... 232 Table 6.4(f) - Logistic regression analysis o f the association between delayed childbearing and risk o f low (< 2.5 kg) birth weight among four racial/ethnic groups - Parity > 1 - United States, 1989 - 1991 [prenatal care not controlled]

233

Table 6.5(a) - Odds ratios for the association o f delayed childbearing (maternal age > 35 years) with preterm (< 37 weeks) delivery among four racial / ethnic groups - First births - United States, 1989 - 1991 .....................234 viii


Table 6.5(b) - Odds ratios for the association o f delayed childbearing (maternal age > 35 years) with preterm (< 37 weeks) delivery among four racial / ethnic groups - Parity > 1 - United States, 1989 —1991 ...............................235 Table 6.5(c) - Risk differences for the association o f delayed childbearing (maternal age > 3 5 years) with preterm (< 37 weeks) delivery among four racial / ethnic groups - First Births - United States, 1989 —1991 ............................ 236 Table 6.5(d) - Risk differences for the association o f delayed childbearing (maternal age > 3 5 years) with preterm ( 1 - United States, 1989 —1991 ...............................237 Table 6.5(e) - Odds ratios for the association o f delayed childbearing with preterm ( 35 years) among four racial/ethnic groups First births - United States, 1989 —1991 ....................................................................240 Table 6.6(b) - Attributable fractions o f preterm delivery among the exposed (AFe) and attributable fractions in the population (AFp) associated with delayed childbearing (> 35 years) among four racial/ethnic groups Parity > 1 - United States, 1989 —1991 ...................................................................... 241 Table 6.6(c) - Attributable fractions o f preterm delivery in the population (AFp) associated with delayed childbearing (>35 years) among four racial/ethnic groups - All Births - United States, 1989 - 1991 ................................242 Table 6.7(a) - Logistic regression analysis of the association between delayed childbearing (>35 years) and risk o f preterm (< 37 weeks) delivery among four racial/ethnic groups - First births - United States, 1989 —1991 .......................243 Table 6.7(b) - Logistic regression analysis of the association between delayed childbearing (>35 years) and risk o f preterm (< 37 weeks) delivery among four racial/ethnic groups - Parity > 1 - United States, 1989 —1991 .........................244 Table 6.7(c) - Logistic regression analysis of the association between delayed childbearing and risk o f preterm (< 3 7 weeks) delivery among four racial/ethnic groups - First births - United States, 1989 —1991 ...............................245 ix


Table 6.7(d) - Logistic regression analysis o f the association between delayed childbearing and risk o f preterm (< 3 7 weeks) delivery among four racial/ethnic groups - Parity > 1 - United States, 1989 —1991 ..................................246 Table 6.7(e) - Logistic regression analysis o f the association between delayed childbearing and risk of preterm (< 3 7 weeks) delivery among four racial/ethnic groups - First births - United States, 1989 - 1991 [prenatal care not controlled] ... 247 Table 6.7(f) - Logistic regression analysis o f the association between delayed childbearing and risk o f preterm (< 37 weeks) delivery among four racial/ethnic groups - Parity > 1 - United States,1989 —1991 [prenatal care not controlled] .......248 Table 6.8 - Risk o f Down syndrome at birth by age o f the mother for three racial/ethnic groups - United States, 1989 - 1991 ..................................................... 249 Table 6.9 - Impact o f delayed childbearing on the risk for Down syndrome for three different racial/ethnic groups in the United States, 1989 —1991 ...............251 Table 6.10 - Odds o f amniocentesis use by age and race of the mother — United States, 1989-1991 ............................................................................................252 Table 6.11 - Risk o f Down syndrome at birth by age and education o f the mother for African Americans and non-Hispanic whites —United States, 1989 —1991 .... 253 Table 6.12 - Odds o f amniocentesis use by maternal age and education for African Americans and non-Hispanic whites - United States, 1989 —1991 ........... 256 Table 6.13(a) - Competing risks analysis o f the effect o f delayed childbearing (maternal age > 3 5 years) on cause-specific infant mortality - United States, 1990 ..................................................................................................................................257 Table 6.13(b) - Competing risks analysis o f the effect o f delayed childbearing (maternal age > 3 5 years) on cause-specific infant mortality —United States, 1990 ................................................................................................................................. 258

x


ACKNOWLEDGEMENTS: I am grateful to my advisors David Meltzer, Willard Maiming, Robert Michael, Edward Lawlor and Kwang-sun Lee for many helpful comments and suggestions. I have also benefited from discussions with Alberto Palloni, Tomas Philipson, Stephen Wall, and Robert Willis. All errors are o f course my own.

xi


ABSTRACT Delayed childbearing has increased in the United States over the past three decades. In contrast to teenage childbearing, the potential effects o f delayed childbearing on the risk for adverse birth and infant outcomes have not received much attention in public policy debates. There are plausible reasons to suggest that biological changes that are associated with advanced maternal age might increase the risks for several adverse reproductive outcomes. However, there are also important socioeconomic and behavioral factors that are associated with both delayed childbearing and the risks for adverse outcomes that may confound o r modify the biological relationship between advanced maternal age and adverse outcomes. The aim o f this dissertation is to assess the individual- and population-level impact o f delayed childbearing on risks for low (< 2.5 kg) birth weight, preterm ( 1] is likely to be lower than the socioeconomic profile o f first births to women at older ages. Finally, as discussed in the previous section, given the methodological issues relevant to the interpretation o f psychological literature, it is difficult to ascertain whether the net effect o f psychological correlates o f delayed childbearing, is to increase or decrease the levels o f anxiety, depression, stress or other potential contributors to increased risk o f adverse birth outcomes such as low birth weight or preterm delivery.

Reference List (1) Heuser RL. Fertility Tables for Birth Cohorts by Color: United States 1917-1973. DHEW Publication No. (HRA) 76-1152. 1976. Rockville, MD, National Center for Health Statistics. Ref Type: Report (2) Ryder NB. Components o f temporal variations in American fertility. In: Hioms RW, editor. Demographic patterns in developed societies. London: Taylor and Francis, 1980: 15-54. (3) Lee RD, Casterline JB. Introduction. Fertility in the United States: New patterns, new theories. New York: Population Council, 1996: 1-15. (4) Morgan SP. Characteristic Features o f Modem American Fertility. In: Casterline JB, Lee RD, Foote KA, editors. Fertility in the United States: New patterns, new theories. New York: Population Council, 1996: 19-63. (5) Ventura SJ. Births: Final Data for 1997.47(18). 4-29-1999. Rockville, MD, National Center for Health Statistics. Ref Type: Report (6) Ventura SJ. Trends and variations in first births to older women, United States, 1970-86. Vital Health Stat 1989; 21:1-27. (7) Hotz VJ, Klerman JA, Willis RJ. The economics o f fertility in developed countries. In: M.R.Rosenzweig, O.Stark, editors. Handbook o f Population and Family Economics. Elsevier Science B.V., 1997: 275-347. (8) Chen R, Morgan SP. Recent trends in the timing of first births in the United States. Demography 1991; 28(4):513-533.


60 (9) Rindfiiss RR, Bumpass L, StJohn C. Education and fertility: Implications for the roles women occupy. American Sociological Review 1980; 45:431-447. (10) Rindfuss RR, Morgan SP, Swicegood G. First Births in American: Changes in the Timing o f Parenthood. Berkeley: University of California Press, 1988. (11) Rindfuss RR, Morgan SP, Ofiutt K. Education and the changing age pattern of American fertility: 1963-1989. Demography 1996; 33(3):277-290. (12) Bloom D, Trussell J. What are the determinants o f delayed childbearing and permanent childlessness in the United States. Demography 1984; 21(4):591-6l 1. (13) Becker GS. An economic analysis of fertility. 209-231. 1960. Princeton, NJ, NBER. Universities-National Bureau of Economic Research Conference Series 11. Ref Type: Serial (Book,Monograph) (14) Becker GS. A Treatise on the Family. Enlarged ed. Cambridge, MA: Harvard University Press, 1991. (15) Michael RT. Education and the Derived Demand for Children. In: Schultz T, editor. Economic o f the Family: Marriage, Children and Human Capital. Chicago: The University o f Chicago Press, 1974: 120-156. (16) Michael RT, Willis R. Contraception and Fertility: Household Production under Uncertainty. In: Terleckyj NE, editor. Household Production and Consumption. New York: National Bureau o f Economic Research, 1976: 27-93. (17) Willis RJ. Economic Theory of Fertility Behavior. In: Schultz T, editor. Economic o f the Family: Marriage, Children and Human Capital. Chicago: The University of Chicago Press, 1974: 25-75. (18) Coale A, McNeil D. The distribution by age of the frequency o f first marriage in a femal cohort. Journal o f the American Statistical Association 1972; 67:743-749. (19) Trussell J, Bloom D. Estimating the covariates o f age at marriage and first birth. Population Studies 1983; 37:403-416. (20) Trussell J. Illustrative analysis: Age at first birth in Sri Lanka and Thailand. 13. 1981. London, World Fertility Survey. Scientific Reports. Ref Type: Report (21) Casterline JB, Trussell J. Age at first birth. 15. 1980. London, World Fertility Survey. Comparative studies. Ref Type: Report (22) Bloom D. What's happening to the age at first birth in the United States? A study of recent cohorts. Demography 1982; 19:351-370. (23) Golding C. The meaning of college in the lives o f American women: The past one-hundred years. 4099. 1992. Cambridege, MA, National Bureau o f Economic Research. Working


61 Paper. Ref Type: Report (24) Jacobsen JP. The economics o f gender. Cambridge, MA: Blackwell, 1994. (25) Becker GS. A theory o f the allocation oftime. Economic Journal 1965; 75:493-517. (26) Heckman JJ, Willis R. Estimation of a Stochastic Model o f reproduction. In: Terleckyj N, editor. Household Production and Consumption. New York: National Bureau of Economic Research, 1976: 99-138. (27) Newman JL. Economic analysis o f the spacing o f births. American Economic Review 1983; 72(2):33-37. (28) Newman JL, McCulloh C. A hazard rate approach to the timing of births. Econometrica 1984; 52:939-961. (29) Newman JL. A stochastic dynamic model of fertility. Research in Population Economics 1988; 6:41-68. (30) Happel SK, Hill J.K., Low SJ. An economic analysis of the timing of childbirth. Population Studies 1984; 38:299-311. (31) Moffitt R. Profiles of fertility, labour supply and wages of married women: a complete life cycle model. Review o f Economic Studies 1984; 51:263-278. (32) Cigno A., Ermisch J. A microeconomic analysis of the timing of births. European Economic Review 1989; 33:737-760. (33) Wolpin K. An estimable dynamic stochastic model o f fertility and child mortality. Journal of Political Economy 1984; 92:852-874. (34) Hotz VJ, Miller R. Conditional choice probabilities and the estimation o f dynamic models. Review of Economic Studies 1993; 60:497-530. (35) Alden A. Delayed childbearing: Issues and implications. Dissertation Abstracts International 42(7-B), Jan. 1982. (36) Baber KM. College women's career and motherhood expectations: New options, old dilemmas. Sex Roles 1989; 19:203. (37) Baber KM. Delayed childbearing: The psychosocial aspects o f the decision-making process. Dissertation Abstracts International 44(I0-A), Apr. 1984. (38) Dion KK. Delayed parenthood and women's expectations about the transition to parenthood. International Journal o f Behavioral Development 1995; 18:333. (39) Coady SS. Delayed childbearing: Correlates of maternal satisfaction at one year postpartum. Dissertation Abstracts International 43(10-B), Apr. 1983.

Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.

62 (40) Meltzer RH. An investigation of the experience o f combining the roles o f mother and worker among older first time mothers. [Dissertation] New York University, 1986. (41) Nord CW. Delayed childbearing in the United States: An exploration of the implications for women’s and children’s lives. Dissertation Abstracts International 49(9-A), Mar. 1989. (42) Bowen SM. Intimacy and generativity in on-time and delayed childbearing women. Dissertation Abstracts International 50(4-A), Oct. 1990. (43) Marchant L. The impact o f delayed parenthood on stress, anxiety and depression among women. [Dissertation] Texas A&M University, 1991. (44) Deitch KV. The consequences of delayed childbearing: A social time model. Dissertation Abstracts International 53(5-A), Nov. 1993. (45) Mueller MC. Was it worth the wait? Delayed parenthood as experienced by parents and their children. [Dissertation] University of Toledo, 1993. (46) Gander M. Discrepancies between childbearing expectations and the perception o f the actual experience of the mature primipara. [Dissertation] University of Manitoba (Canada), 1992. (47) Liebmann-Smith J. Social consequences o f delayed childbearing and infertility. [Dissertation] City University o f New York, 1995. (48) Issod JL. A comparison o f "on-time" and "delayed" parenthood. American Mental Health Counselors Association Journal 1988; 9:97.


63

CHAPTER 3 - EFFECTS OF SOCIOECONOMIC FACTORS ASSOCIATED WITH DELAYED CHILDBEARING ON THE RISK OF ADVERSE BIRTH OUTCOMES: REVIEW OF LITERATURE This chapter is a review o f the literature on the effects o f socioeconomic factors associated with delayed childbearing on the risk for adverse outcomes. As discussed in Chapter 2, these factors include ethnicity, education, income and occupation. These factors, in turn, independently affect the risk o f several adverse birth outcomes as discussed below. Overall, women o f higher socioeconomic status, who are more likely to delay child bearing, are at lower risk for adverse birth outcomes. In addition, it appears that independent o f socioeconomic factors such as maternal education, income and marital status, there are substantial ethnic variations in the risk of adverse birth outcomes. The chapter discusses the literature on the effects o f ethnicity, education, income and occupation on the risk of adverse birth outcomes. It is organized as follows: Section 3.1 reviews the literature on the effects o f ethnicity on birth outcomes, which also includes a discussion o f the interactions between maternal age and race (Section 3.1.1) and o f ethnic differences in the use o f prenatal diagnosis (Section 3.1.2). Section 3.2 discusses the literature on the effects o f education and income and Section 3.3 reviews the literature on the effects o f occupation on the risk for adverse birth outcomes.


64

3.1 Effects o f ethnicity on adverse birth outcomes An extensive literature has documented racial or ethnic differences in essentially all o f the measured indices o f adverse pregnancy outcomes such as fetal death, low and very low birth weight, and neonatal mortality I~13. African Americans have a two to three fold higher risk o f fetal death, preterm delivery, low birth weight and infant death as compared to Whites. Although African Americans have higher proportion o f births with maternal risk factors (or markers), such as young age, lower level o f education and unmarried status, these differences do not fully account for their higher rates o f adverse outcomes 1’9. In addition to the overall racial differences in the risk o f adverse birth outcomes between African Americans and Whites, there appear to be racial differences in the effects o f other risk factors, i.e., interactions between race and other risk factors such as maternal age, education and marital status that are known to affect adverse birth outcomes. Kleinman and Kessel1 studied racial differences in low birth weight to identify risk factors responsible for differences in low birth weight between African Americans and whites in the United States. Using national data in 1983, they found the risk o f very low birth weight (< 1.5 kg) to be three fold higher, and the risk o f moderately low birth weight (1.5-2.5 kg) to be 2.3 fold higher in African Americans as compared with Whites. They investigated the effects o f four risk factors, maternal age, parity, marital status and education on rates o f very low birth


65 weight and moderately low birth weight. Using these four factors to define low risk and high risk groups, they found racial differences in birth weight to be greater among low risk than among high risk mothers, especially for very low birth weight; the risk ratios in very low birth weight comparing low-risk and high-risk African Americans to Whites were 3.4 and 1.7 respectively. Other studies have also shown the interaction noted above between race and other sociodemographic risk factors for adverse birth outcomes; i.e., racial disparities in birth outcomes have been shown to be greater in “low-risk” as compared with “high-risk” women. Using the 1983 and 1984 national data to study racial differences in infant mortality, Kleinman and colleagues 2 used maternal age, parity, education, marital status and place o f birth (native vs. foreign bom) to define three broad maternal risk groups. Lower risk births occurred to mothers who were married, more educated, in their twenty to early thirties, multiparous (birth orders 2-4), and foreign bom. In particular they showed that for both races, foreign-born women had about a 20 % lower risk o f infant mortality as compared with native-born women suggesting a favorable selection (“healthy immigrant effect”) in both races for foreign-born women. Kleinman and colleagues also showed that between the low and high levels o f maternal risk, infant mortality increased by a factor of approximately two for African Americans and three for Whites in the United States. Similarly, using birth data in Chicago in the years 1982 and 1983, and 1980 median family income o f the mother’s


66

census tract, Collins and David 10 found that among high risk mothers in the poorest areas the proportion o f low birth weight infants in African Americans and Whites were less divergent than in higher income areas. The reason(s) for this greater racial disparity in “low-risk” as compared with “high-risk” women are not clear. The definitions used in the literature to define low and high-risk categories clearly encompass heterogeneous groups o f women in both races. Therefore, it is certainly possible that racial disparities in income, education, place o f residence, or other socioeconomic and behavioral factors known to affect the risk o f adverse birth outcomes might be greater in “low-risk” as compared with “high-risk” women. If so, the greater racial disparities among “low-risk” women might reflect, at least in part, the greater divergence in resources and psychosocial stresses between African Americans and Whites who are defined as “low risk” as compared to their “high risk” counterparts. Most o f the previous studies have focused on the racial disparities in pregnancy outcomes between African Americans and Whites. However, there is a growing body o f evidence that indicates there are differences in pregnancy outcomes among several other racial or ethnic groups in the United States. Becerra and colleagues 12 used the 1983 and 1984 Linked Births and Infant Death datasets to compare infant mortality risks among single-delivery infants o f Hispanic descent with those among singleton infants o f non-Hispanic whites. Neonatal mortality risk was higher among Puerto Rican islanders [Relative Risk= 2.3] and continental Puerto


67 Ricans [Relative Risk = 1.5] and lower among Cuban- Americans [Relative Risk = 1.0] and Mexican Americans [Relative Risk = 1.0]. The post-neonatal (28 to 364 days) mortality risk was highest among continental Puerto Ricans [Relative Risk = 1.2] and lowest among Cuban-Americans [Relative Risk = 0.6]. Overall, these results and those o f Collins and Shay 14 underscore the heterogeneity o f birth outcomes in the Hispanic population in the Untied States. Furthermore, traditional measures o f socioeconomic status such as education and income do not seem to explain ethnic differences within Hispanic groups or between Hispanic subgroups and non-Hispanic whites. In particular, the favorable outcomes o f births to Mexican American women with lower socioeconomic status has been termed an ‘epidemiological paradox’ and remains unexplained 14. Overall, the vast literature that is concerned with the racial/ethnic differences in pregnancy outcomes has succeeded far better in documenting the existence and extent o f inequalities in outcomes than in explaining why these inequalities persist. In particular, the higher neonatal mortality in African American infants does not seem to be principally due to less access to high quality neonatal intensive care for African American infants who are preterm or very low birth weight as compared with their white counterparts. In fact, very low birth weight African American infants have a survival advantage compared with whites 15. In order to understand the issue o f racial/ethnic differences in pregnancy outcomes in general, one needs to consider the meaning and utility o f the designations


68

race and ethnicity in studies o f human health and disease l6. Stephen Gould contends that from a strictly biological point o f view:

the continued racial

classification o f Homo sapiens represents an outmoded approach to the general problem o f differentiation within a species ” 11. Others hold that the concept o f biological race may have some limited usefulness in clinical medicine, epidemiology and public health 18. They maintain that the concepts o f race and ethnicity, although imprecise, are useful in descriptive epidemiological studies that may yield etiologic hypotheses to be explored by more detailed studies. At best, however, the use o f racial/ethnic groups may be an interim solution to the problem o f describing variability with reference to disease and health patterns

3.1.2

1X

.

Interaction between maternal age and race —The “weathering hypothesis” Writing mostly in reference to teenage childbearing and its contribution, or

lack thereof to racial disparities in birth outcomes, Geronimus has forwarded the “weathering hypothesis” to explain interaction between maternal age and race 8’19. The empirical basis for this hypothesis was first explored by Geronimus in a population-based study to explore whether the excessive neonatal mortality rates among infants with teenage mothers were due to young maternal age per se or due to socioeconomic

circumstances

of teenage

mothers

I9. Maternal

age-specific

comparison o f neonatal mortality rates between African Americans and whites in a population-based sample o f approximately 306 thousand first births to African


69 American and white mothers for three states in the years 1976-1979 showed that at every maternal age, higher neonatal mortality rates were observed for African Americans compared to whites. Furthermore, African Americans above age 23 experienced higher neonatal mortality rates than most African American or white teenagers. Hence the racial disparity in neonatal mortality was greater at older maternal ages than at younger ages, suggesting that the higher neonatal mortality among African Americans was not due to the greater frequency o f teenage childbearing among African Americans. In addition, she observed that the effect o f advancing maternal age on the risk for neonatal mortality was greater for African Americans than it was for whites. She suggested that this might be due to greater • s; namely that “weathering” with age for African Americans as compared with Whites the health o f African American women might begin to deteriorate earlier and at a greater rate as a consequence o f cumulative socioeconomic disadvantage.

3.1.2

Ethnic differences in use o f prenatal diagnosis Previous studies have documented ethnic differences in use o f prenatal

diagnosis in the United States and abroad 2021. In a study o f Georgia women, Sokal and colleagues 20 showed racial and geographic variations in the rates o f prenatal chromosomal diagnosis. This study examined the use o f prenatal chromosomal diagnosis in 1978 by Georgia residents who would be 40 years and older at delivery, i.e., women at high risk o f Down syndrome, other autosomal trisomies and


70 chromosomal anomalies. It should be noted that this study examined the use o f prenatal diagnosis in 1978 when it was still relatively new technology even though it had been recommended as a routine procedure inaJAALi editorial in 1974. The study by Sokal and colleagues showed that overall 15 % o f Georgia women 40 years and older made use o f prenatal chromosomal diagnosis. However, there were substantial racial and geographic differences in the ratio o f women who used prenatal diagnosis. Prenatal chromosomal diagnosis use ratios ranged from about 60 % among whites in two large urban counties (Fulton and DeKalb, the most populous counties in the Atlanta area) to 0.5 % among blacks outside Augusta and Atlanta health districts. Black women in Atlanta and Augusta health districts had a use ratio o f 19.8% (p 4.5 kg). Infants o f older mothers without any clinically manifest sign o f diabetes or glucose intolerance also appear to be an increased risk o f macrosomia. Infants with macrosomia are at higher risk for traumatic deliveries including brachial plexus injuries and shoulder dystocia (dislocations). Advancing maternal age is also associated with increased risk o f hypertensive disorders including preeclampsia, other forms o f pregnancy-induced hypertension, and chronic hypertension 2-23-25. In a recent study, Bianco et al reported about a two fold increase in the risk o f preeclampsia for both first births (nulliparas) and births o f second or higher order (multiparas) births (nulliparas: odds ratio, 1.8, 95 % C.I., 1.32.6; multiparas, odds ratio, 1.9, 95 % C.I., 1.2-2.9) among older gravidas. Hypertensive disorders can in turn increase the risk o f intrauterine growth retardation, preterm delivery and perinatal mortality. In addition, it has been suggested 2,5 that uteroplacental underperfiision in either a chronic or acute form and with or without

permission of the copyright owner. Further reproduction prohibited without permission.

m aternal

hypertension might serve as a common pathway for fetal compromise

either in the form o f intrauterine growth retardation o f the fetus or fetal hypoxia. Naeye 5 studied age-related vascular lesions in the uterus that might affect uteroplacental perfusion. He found that older women had sclerotic lesions in the intramyometrial arteries in the uterus that can limit blood flow. These lesions were not the atherosclreotic lesions that develop with age in elastic arteries, but rather fibrotic lesions that are confined to the media o f muscular arteries. Collagen and lon g itu dinally oriented smooth muscle had replaced the normally circularly oriented

smooth muscle in the arteries’ media. These findings were reported based on examination o f normal uteri from nonpregnant women 17 to 19 years of age without a history o f hypertension who had come to autopsy as the result o f accidental death. He found that the percentage o f intramyometrial arteries with sclerotic lesions increased from 11% at 17-19 years o f age to 37% at 20 to 29 years, 61% at 30 to 39 years and 83% after age 39 (p 40 years o f age were respectively, 1.5 (95 % C.I., 1.2 -1.8), 1.7 (95 % C.I., 1.4 - 2.2) and 2.3 (95 % C.I., 1.6 - 3.4). There were similar trends for preterm delivery and for very low birth weight; however the effect o f age for very low birth weight infants did not reach significance at 0.05 level. No statistically significant maternal age effect was found for any o f the outcomes in infants bom to African American mothers. However, only 127 births to African American women aged 35 years or more were studied. It has been suggested that there may be differential effects o f advanced maternal age in African Americans as compared with whites. Geronimus32 has hypothesized that the effects o f aging may be greater under circumstances o f social


97 disadvantage and in particular for African Americans in the United States (the “weathering hypothesis”). However, as noted above, findings o f the few previous studies on this subject have not been consistentI2’15’32.

4.4.2

Birth defects The in utero incidence and birth prevalence o f chromosomal anomalies and

most notably, autosomal trisomies are known to increase with advancing maternal age j3. Down syndrome (trisomy 21) is one o f the most common congenital anomalies and is the foremost known genetic cause o f mental retardation 33-34. It is the most intensively studied human chromosome abnormality and yet little is known about its cause. The established risk factors for Down syndrome include, most importantly, advancing maternal age, and to a quantitatively lesser degree, family history o f trisomy or relevant chromosomal rearrangements34. In studies that for the most part predate the advent o f prenatal diagnosis, birth prevalence o f Down syndrome is reported to be lowest for women 20 to 24 years o f age (1/1400 births), and to increase exponentially from early thirties. The birth prevalence o f Down syndrome was about 1/350 (four fold increase) for women 35 years old and 1/25 (about 50 fold increase) for women 45 years o ld 35. The estimated rates o f other less common chromosomal abnormalities, most notably, the Edwards syndrome (Trisomy 18), and Klinefelter syndrome (XXY genotype) also appear to increase exponentially with age 35. The rates o f Patau


syndrome (Trisomy 13) also increase w ith age; however the age effect is less marked for Trisomy 13 presumably due to the association o f a large fraction o f cases with translocations that are not affected by age. H ook 35 has estimated the rates o f all c linically significant

cytogenetic abnormalities to increase from about

2

per

1000

at

the youngest maternal ages to about 2.6 per 1000 at age 30, 5.6 per 1000 at age 35, 15.8 per 1000 at age 40 and 53.7 per 1000 at age 45. Trisomy 21 causes about 95 % o f Down syndrome cases, the remaining proportion being caused by translocations 36. Trisomy 21 typically results from nondisjunction during meiosis. Nondisjunction may occur during Meiosis I when the chromosome pairs foil to separate or during Meiosis II when chromatids fail to separate. Molecular techniques have made it possible to detect the stage and parental origin o f the extra chromosome 21 in the case o f Down syndrome by identification of DNA polymorphisms. The results o f the study by Yoon and colleagues

36

suggests

that advanced maternal age increases the risk o f both Meiosis I and Meiosis II errors. Since errors in Meiosis I can occur as early as the mother’s fetal life, the association between maternal age and Meiosis I errors does not pinpoint the timing o f errors. On the other hand, the association between advanced age and Meiosis II suggests that there is at least one maternal age-related mechanism that acts around the time o f conception.


4.4.3

Perinatal mortality The literature on the effects o f advancing maternal age on perinatal mortality

suggests that in the modem obstetric and neonatal care era there is only a modest agerelated effect on overall perinatal mortality3-23’25-26. As noted above, several pregnancy complications that increase the risk o f perinatal mortality, including hypertensive disorder, diabetes, placenta previa and abruptio placenta, have been shown to increase with maternal age 23-25’26. jn addition, while the findings from previous studies are inconsistent, advancing maternal age has been shown to be associated with increased risk o f low birth weight and preterm delivery in populationbased studies with adequate power to show such effects. Low birth weight and preterm delivery are in turn the principal predictors o f neonatal mortality37'39. However, it appears that modem obstetric and neonatal care has diminished or completely alleviated the age-related increases in the overall risk o f perinatal mortality 1 23-25’26. In a study o f women 40 years and older, Bianco and colleagues 25 concluded that although maternal morbidity was increased in the older gravidas, the overall neonatal outcome did not appear to be affected. Prysak et al

studied the outcomes

o f women and infants bom to women 35 years and older and concluded that nulliparous women 35 years and older had higher rates o f antepartum, intrapartum and newborn complications compared with their 25-29 year old counterparts but not an increased perinatal mortality rate. They also stated that despite increased risk o f


100 complications, maternal and perinatal outcomes were good. Neither o f these studies distinguished between difference causes o f infant death nor did they assess the effects o f age on causes o f postneonatal mortality. In an investigation o f N ew York City births in the years 1976 to 1978, Kiely, Paneth and Susser 8 assessed the effects o f maternal age on perinatal mortality in the only investigation that has divided perinatal mortality into four components which have distinct epidemiological and medical features: late fetal deaths that occurred before labor (late antepartum fetal deaths), fetal deaths during labor (intrapartum fetal deaths), neonatal deaths, and perinatal deaths due to congenital anomalies. In analyses that controlled for prior fetal loss, public vs. private service, race, marital status and maternal education, they found that increasing maternal age increased the risk of antepartum fetal deaths but not intrapartum fetal deaths. Stein 1 has suggested that improved obstetric care might have diminished or entirely alleviated the age-related increases in the risk o f adverse intrapartum mortality. Kiely and colleagues also found that older maternal age was associated with perinatal deaths due to congenital anomalies and in the case o f first births, neonatal deaths 8. However, none o f the previous studies have done a comprehensive causespecific analysis of the effects o f age on neonatal, as well as, postneonatal mortality in a recent population-based cohort in the United States.


101 4.5

Summary o f age effects There is empirical evidence, as well as plausible biological links3'4"7'36'40,

between advanced maternal age and several adverse reproductive outcomes. Nevertheless, the extensive literature on the effects o f advanced maternal age on pregnancy outcomes is by no means consistent and can be difficult to interpret particularly with regard to low birth weight, preterm delivery and infant mortality. While many studies have shown higher risks for several adverse outcomes among older mothers, the conclusions of other studies have been quite reassuring and even enthusiastic about the quality of pregnancy and childbirth among older women. Several methodological problems have affected many o f the studies on the effects o f delayed childbearing on the risks for adverse outcomes. In particular, most o f the previous studies o f the relationship between advanced maternal age and risks for adverse outcomes have been hospital-based studies. Problems of inference from these hospital-based studies o f the effects o f delayed childbearing on reproductive outcomes include selection bias with inadequate control for confounders 2’13"14'25, lack o f sufficient power to detect reasonable differences in outcomes particularly if the effects are to be adjusted for the effects o f potential confounders

12-14-23*25

and

questions regarding generalizability o f the findings to other groups in the population 12.13

There have been few population-based studies o f the effects o f delayed childbearing on reproductive outcomes9"12. There have also been inconsistencies with


102

regard to whether there are differential effects o f advanced maternal age in African Americans as compared with whites 12'15’32. Furthermore, the impact o f delayed childbearing on the range o f adverse outcomes considered here has not been examined using United States data from a recent national cohort. In particular, previous studies have not examined the differential effect o f maternal age on the risk for Down syndrome in different ethnic and education groups. N or have previous studies examined the differential effects o f maternal age on a comprehensive set o f cause-specific infant mortality rates in a nationally representative birth cohort in the United States. A major problem with essentially all of the previous hospital-based studies has been lack of sufficient power to detect reasonable differences in adverse birth outcomes. Assuming a 1:1 ratio o f comparison groups and an alpha (Type I error rate) o f 0.05, for a study to have an 80 % power of detecting a 50 % increase in the risk for very low birth weight (from 1 % to 1.5 %), a sample size o f approximately 8,145 births is required for mothers in each o f the comparison groups; same assumptions imply that detecting a 50 % increase in low birth weight rates (from 5 % to 7.5 %) requires 1,550 births in each group. By comparison, the study by Barkan et a l 14 which noted no increase in low birth weight and preterm delivery rates due to delayed childbearing included 313 births to mothers > 30 years o f age and 33 births to mothers > 3 5 years o f age. The study by Bianco et a l 25 also suggesting no increase in low birth weight and very low birth weight rates in mothers > 40 years o f age was


based on 607 nulliparous and 797 multiparous women > 40 years o f age. Based on the sample sizes in the study by Bianco et al, the power to detect a 50 % difference in very low weight rates due to delayed childbearing was approximately 18 % for both nulliparous and multiparous mothers. Bianco et al note that their study is the largest published series o f patients over the age o f 40 reported to date. The data sets and the set o f analyses in the present study will allow comparisons o f the risks associated with delayed childbearing on several major birth outcomes among different ethnic groups using recent national data. To date, the present study represents the largest study o f the effects of maternal age on adverse birth outcomes and one with sufficient power to answer the questions posed in the empirical analyses. Combining measures o f relative risk associated with delayed childbearing with prevalence o f delayed childbearing at the population level, the study provides estimates o f the total impact o f delayed childbearing on the risks for adverse outcomes in the United States. In addition, the study addresses interaction effects between maternal age and other attributes o f mothers who bear children at older ages. Specifically, effect modification by maternal ethnicity, education and parity are considered, which have not been adequately addressed in the previous literature. Assessment o f these interaction effects helps define groups that may be particularly vulnerable or protected from the adverse effects o f delayed childbearing. In addition, these interactions can also generate hypotheses about the underlying mechanisms that produce the adverse effects o f advancing maternal age in different


104 groups in the population and suggest medical interventions and policy options to ameliorate them.

Reference List (1) Stein Z. Aging and Reproduction: I. Fecundity, Fertility and Gestation. In: Kline J, Stein Z, Susser M, editors. Conception to Birth: Epidemiology o f Prenatal Development. New York: Oxford University Press, 1989: 259-281. (2) O'Reilly-Green C, Cohen WR. Pregnancy in women aged 40 and older. [Review] [103 refs]. Obstetrics & Gynecology Clinics o f North America 1993; 20(2):313-331. (3) Kline J, Stein Z, Susser M Conception to Birth: Epidemiology of Prenatal Development. New York: Oxford University Press, 1989. (4) Cano F, Simon C, Remohi J, Pellicer A. Effect o f aging on the female reproductive system: evidence for a role o f uterine senescence in the decline in female fecundity. Fertility & Sterility 1995; 64(3):584-589. (5) Naeye RL. Maternal age, obstetric complications, and the outcome o f pregnancy. Obstetrics & Gynecology 1983; 61:210-216. (6) Pellicer A, Simon C, Remohi J. Effects o f aging on the female reproductive system. [Review] [37 refs]. Human Reproduction 1995; 10 SuppI 2:77-83. (7) Abdalla HI, Burton G, Kirkland A, Johnson MR, Leonard T, Brooks AA et al. Age, pregnancy and miscarriage: uterine versus ovarian factors. Human Reproduction 1993; 8(9): 1512-1517. (8) Kiely JL, Paneth N, Susser M. An assessment o f the effects of maternal age and parity in different components of perinatal mortality. Am J Epidemiol 1986 Mar 123:444-454. (9) Lee KS, Ferguson RM, Corpuz M. Gartner LM. Maternal age and incidence o f low birth weight at term: a population study. American Journal o f Obstetrics & Gynecology 1988; 158(l):84-89. (10) Cnattingius S. Does age potentiate the smoking-related risk of fetal growth retardation? Early Human Development 1989; 20:203-211. (11) Cnattingius S, Forman MR, Berendes HW, Isotalo L. Delayed childbearing and risk of adverse perinatal outcome. A population-based study [see comments]. JAMA 1992 Aug 19 268:886-890. (12) Aldous MB, Edmonson MB. Maternal age at first childbirth and risk of low birth weight and preterm delivery in Washington State. JAMA 1993; 270(21):2574-2577.


105 (13) Berkowitz GS, Skovron ML, Lapinski RH, Berkowitz RL. Delayed childbearing and the outcome o f pregnancy [see comments]. N Engl J Med 1990 Mar 8 322:659-664. (14) Barkan SE, Bracken MB. Delayed childbearing: no evidence for increased risk of low birth weight and preterm delivery. Am J Epidemiol 1987 Jan 125:101-109. (15) Grimes DA, Gross GK. Pregnancy outcomes in black women aged 35 and older. Obstet Gynecol 1981 Nov 58:614-620. (16) Federation des Centres d’Etude et de Conservation du Sperme Humaine DSaMJM. Femal fecundity as a function o f age. New England Journal o f Medicine 1982; 307:404-406. (17) Bongaarts J. Infertility After Age 30: A False Alarm. Family Planning Perspectives 1982; 14(2):75-78. (18) Menken J, Trussell J, Larsen U. Age and Infertility. Science 1986; 233:1389-1393. (19) Vessey MP, Wright NH, McPherson K, Wiggins P. Fertility after stopping different methods o f contraception. British Medical Journal 1978; 2:265. (20) McFalls JA J. The risks o f reproductive impairments in the later years o f delayed childbearing. Annu Rev Sociol 1990; 16:491-519. (21) Wilcox AJ, Weinberg CR, O'Connor JF, Baird DD, Schlatterer JP, Canfield RE et al. Incidence of early loss o f pregnancy. New England Journal o f Medicine 1988; 319:189-194. (22) Bonnie Miller Rubin, Ronald Kotulak. NEVER TOO OLD A REALITY FOR DADS-AND MOMS. Chicago Tribune 1997 Apr 25. (23) Prysak M, Lorenz RP, Kisly A. Pregnancy outcome in nulliparous women 35 years and older. Obstetrics & Gynecology 1995; 85(l):65-70. (24) Dollberg S, Seidman DS, Armon Y, Stevenson DK, Gale R. Adverse perinatal outcome in the older primipara. Journal o f Perinatology 1996; 16(2 Pt l):93-97. (25) Bianco A, Stone J, Lynch L, Lapinski R, Berkowitz G, Berkowitz RL. Pregnancy outcome at age 40 and older. Obstetrics & Gynecology 1996; 87(6):917-922. (26) Berendes H, Forman M. Delayed childbearing: Trends and Consequences. In: Kiely M, editor. Reproductive and Perinatal Epidemiology. Boca Raton, FL: CRC Press, Inc., 1991: 27-42. (27) Newcomb WW, Rodriguez M, Johnson JWC. Reproduction in the older gravida. Journal of Reproductive Medicine 1991;839-845. (28) Placek P, Taffel SM, Moien M. 1986 C-section risk; VBAC inch upward. Am J Publ Hlth 1988; 78:562. (29) Placek P, Keppel KG, Taffel SM, Liss T. Electronic fetal monitoring in relation to cesarean section delivery, for live births and still births in the U.S., 1980. Public Health Reports 1984; 99:173.


106 (30) Witter FR, Repke JT, Niebyl JR. The effect of maternal age on primary cesarean section rate. Int J Gynecol Obstet 1988; 27:51. (3 1) Taffel SM. Cesarean delivery in the United States, 1990. Vital & Health Statistics 1994; Series 21, Data on N(51):l-24. (32) Geronimus AT. The weathering hypothesis and the health of African-American women and infants: evidence and speculations. [Review]. Ethnicity & Disease 1992; 2(3):207-221. (33) Leek I. Structural Birth Defects. In: Pless IB, editor. The Epidemiology of Childhood Disorders. New York: Oxford University Press, 1994: 66-117. (34) Bell J. The epidemiology of Down's syndrome. [Review] [47 refs]. Medical Journal of Australia 1991; 155(2):115-117. (35) Hook EB. Rates o f chromosome abnormalities at different maternal ages. Obstet Gynecol 1981;58:282-285. (36) Yoon P, Freeman S, Sherman S, Taft L, Gu Y, Pettay D et al. Advanced maternal age and the risk o f Down syndrome characterized by the meiotic stage o f chromosomal error: a population-based study. American Journal of Human Genetics 1996; 58(3):628-633. (37) Lee KS, Paneth N, Gartner LM, Pearlman M. The very low-birth-weight rate: Principal predictor of neonatal mortality in industrialized populations. Journal of Pediatrics 1980; 97(5):759-764. (38) Lee KS, Khoshnood B, Hsieh H, Kim BI, Schreiber MD, Mittendorf R. Which birthweight groups contributed most to the overall reduction in the neonatal mortality rate in the United States from I960 to 1986? Paediatric & Perinatal Epidemiology 1995; 9(4):420-430. (39) Lee KS, Paneth N, Gartner LM, Pearlman MA, Gruss L. Neonatal mortality: an analysis of the recent improvement in the United States. Am J Public Health 1980 Jan 70:15-21. (40) Abruzzo MA, Hassold TJ. Etiology o f nondisjunction in humans. [Review] [52 refs]. Environmental & Molecular Mutagenesis 1995; 25 SuppI 26:38-47.


107

CHAPTER 5 - ANALYSES OF THE EFFECTS OF ADVANCED MATERNAL AGE ON BIRTH OUTCOMES 5.1

Main questions o f interest for the empirical study in the dissertation This chapter describes the empirical analyses that address the effects o f advanced

maternal age on birth outcomes in the dissertation. The analyses assess the individuallevel effects and population-level impact o f advanced maternal age on the following set o f adverse outcomes: i)

Low birth weight (< 2.5 kg) and preterm (< 37 weeks completed gestation) delivery

ii)

Structural birth defects: Down syndrome and neural tube defects (anencephaly and spina bifida)

iii)

Infant mortality: overall infant mortality, as well as, cause-specific infant mortality due to

11

causes o f death: maternal complications, complications o f

placenta, hypoxia and asphyxia, prematurity, congenital anomalies, respiratory distress syndrome (RDS), pneumonia, infections, sudden infant death syndrome (SIDS), unintentional (accidents) and intentional injuries (homicides). Both the overall (main) effects o f advanced maternal age and specific interaction effects between age and socioeconomic factors and birth order are estimated. In the case o f main effects, the aim is to answer the set o f questions that are concerned with:


108 1. What is the net (adjusted) individual-level effect and population-level impact o f delayed childbearing on the risk o f adverse birth and infant outcom es after taking into account other socioeconomic and behavioralfactors that are associated with both delayed childbearing and adverse outcomes? The aim o f the first set of questions is to arrive at estimates o f the overall risks associated with delayed childbearing when main effects provide an adequate description o f the effects o f advanced age on adverse birth outcomes; i.e., when interaction effects are not dominant or essential for characterizing the relationship between advanced age and the risk o f adverse outcomes. The risks associated with advanced age are estimated in terms o f both the relative risks (odds ratios) and the risk differences since both the relative (e.g., odds ratio) and the absolute (e.g., risk difference) measures of risk might be important for women in their assessments o f the risks associated with delayed childbearing. Combining measures o f relative risk associated with delayed childbearing with prevalence o f delayed childbearing at the population level, the dissertation also provides estimates (population attributable risks) o f the total impact o f delayed childbearing on the risks for adverse outcomes at the population level. The second set o f questions address interaction effects between maternal age and other attributes of mothers who bear children at older ages, i.e., 2. How is the effect o f delayed childbearing on the risk fo r adverse outcomes m odified by socioeconomic circumstances o f the mothers?


109 Effect modification, by maternal education, ethnicity, and parity are considered. There are two main a priori reasons for focusing on this specific set o f interactions. The first reason is that previous literature suggests that these set o f factors are correlated with birth timing maternal age [Chapter 2] and might also interact in important ways with the effect o f age on the risk o f adverse outcomes [Chapters 3 and 4]. The second reason is that the assessment o f the interaction effects between delayed childbearing and these attributes helps define groups that may be particularly vulnerable or protected from the adverse effects o f delayed childbearing. Finally, interaction effects might also suggest mechanisms by which age might interact with other attributes such as education and ethnicity in affecting the risk o f adverse outcomes and perhaps also ways o f reducing the risks associated with delayed childbearing. Previous literature suggests that more highly educated mothers may be protected against some o f the adverse birth outcomes [Chapter 3]. However, previous studies have not addressed the specific case o f congenital anomalies. In particular, socioeconomic differentials in the use o f prenatal diagnosis and selective termination o f affected fetuses have not received adequate attention in the literature. Therefore, one o f the main questions the dissertation aims to answer concerns the impact o f socioeconomic, specifically, education and ethnicity, differentials in the use o f prenatal diagnostic services. It is hypothesized that mothers who are less educated, or those o f particular ethnic groups might have less access and/or utilization rates of


110 prenatal diagnostic services and selective termination and hence they might be at greater risk for the structural birth defects that are associated with delayed child bearing. Finally, birth order might also modify the effect o f advanced age. For example, the effects of advanced age on low birth weight and preterm delivery might be more pronounced for first births. This might be due to the specific effects o f advanced age on hypertensive disorders during pregnancy that are most common in first births. Specifically, the following hypotheses are tested in the empirical estimations: Hypothesis 1 —The effect ofparity: The increased risks o f low birth w eight and preterm delivery associated with advanced maternal age are greater fo r fir s t births as compared with births o f second or higher order. Advanced maternal age increases the risk o f preeclampsia and other hypertensive disorders that might affect the length o f gestation and intrauterine growth. The risk o f hypertensive disorders during pregnancy, and in particular the risk o f preeclampsia, is substantially greater for first births as compared with births o f second or higher order. Therefore, the increased o f risk o f low birth weight and preterm delivery is likely to be higher for first births.


I ll Hypothesis 2: Effect o f socioeconomic factors: The increased risks o f adverse birth outcomes with advanced maternal age are greater after adjustment is made fo r the higher socioeconomic status o f women who delay childbearing. Women who delay childbearing tend to have a higher socioeconomic status (Chapter 2); in particular they are likely to have higher levels o f education and income. Higher socioeconomic status is associated with lower risks o f adverse birth outcomes (Chapter 3). Therefore, adjustment for socioeconomic factors is likely to increase estimates o f the risks associated with advanced maternal age.

Hypothesis 3: Effect o f race: The increased risks o f adverse birth outcomes associated with advanced maternal age m ight be greater fo r African Americans due to greater rate o f accumulation o f stresses associated with aging; the “weathering hypothesis ”. Geronimus has argued that aging might be more accelerated for African Americans and in particular African Americans women due to a “weathering” effect (Chapter 3). Also because o f higher base line risks in African Americans and Puerto Ricans, the risk differences associated with advanced age would be higher even with equivalent measures o f relative risk such as odds ratio.

Hypothesis 4: The effects o f maternal age on the risk fo r Down syndrome m ight vary by ethnicity and education. Specifically, the risk m ight be greater fo r African


112 Americans and Mexican Americans due to lesser access to or use o f prenatal diagnosis services and selective termination o f affectedfetuses. The age-related risk o f Down syndrome might also be greater fo r women with lower levels o f education within a given ethnic group due to lesser use o f prenatal diagnosis. Previous studies have shown differences by ethnicity and education in the use of medical care and specifically in the use o f prenatal diagnosis (Chapter 3), which in turn could result in differences by ethnicity and education in the effects o f age on the risk o f Down syndrome and other birth defects.

Hypothesis 5: Differential effects o f advanced age on cause-specific infant mortality: Advanced maternal age increases the risk associated with biologically related causes o f infant mortality: infant deaths due to congenital anomalies and abnormalities o f the placenta (placenta previa and abruptio placentae). On the other hand advanced maternal age might be associated with lower risks o f infant mortality due to the set o f causes that are related to childcare practices such as infant deaths due to unintentional and intentional injuries. This association might be due to the effect o f birth tim ing itself or to the effect o f unmeasured socioeconomic facto rs associated with delayed childbearing or both.


5.2

Conceptual framework for the empirical analyses A main problem in conducting and interpreting the results o f any analysis that

aim s

to assess the causal relationship between maternal age and the risk for adverse

birth and infant outcomes is the endogeneity o f maternal age. As discussed in Chapters 3 and 4, there are important socioeconomic and behavioral factors that are associated with both delayed childbearing and the risks for adverse outcomes. It is also possible that independent o f any effects o f aging on the reproductive function, lower ‘baseline' levels o f fecundity may be associated with delayed childbearing as well as various forms o f reproductive dysfunction. I f so, mothers with lower levels o f fecundity are more likely to (involuntarily) bear children at older ages and might also have higher risks for adverse outcome without there being any causal relationship between their age per se and risks for adverse outcomes. Considering the stochastic nature o f reproductive outcomes, the risk for an adverse outcome in relation to maternal age may be modeled as: Pt = P ( t,X ( t\ M ( X ,t) ) , where P, is the probability o f the occurrence for the adverse outcome i, t is the mother’s age, ATis a vector o f household or maternal characteristics excluding the mother’s age that by definition reduce the risk for the adverse outcome i and M is the use o f medical inputs (e.g., prenatal care, amniocentesis, ultrasound).


114 Fu thennore x£ X = (X x,...,X n) ,M = (M {,...,Mm) , their dP;

5P

n dP; dX-

^

ct

■‘Z âXi dt

L=

L+ y

L

m SR

L+ y

L_

L

j t x8M j' dt

• ’

dM - dM„ dM- & L - __ =L+ y __ ____ L or / Z‘l gT- ' dt

where:

5Pf

— —< 0 (by assunytion)

dXf

5P

— < 0 (underlie plausible assurptionthat the net effect o f medical utterventions is to reduce

dMj

risks for adverse outcomes).

On the other hand, in genreal, the signs of the other terms are ambiguous - i.e., dX, n dX, _ . >0 or < 0 and dt dt dM , dM , ^-> 0 o r '-< 0. dt dt Sociological and economic studies o f birth timing decisions (Chapter 2) suggest that the socioeconomic status o f women who delay their first births tends to be higher that that o f women who have first births at younger ages. Since higher socioeconomic status is in general associated with lower risk o f adverse outcomes (Chapter 3), the first dX: component, i.e., — - > 0 probably dominates for most adverse outcomes. Also women dt o f higher socioeconomic status might have greater demand for “quality” (as in Becker’s quality-quantity interaction, see Chapter 2) or for reducing uncertainty and hence may behave in ways that reduce the risk o f adverse birth outcomes. Hence it is more likely fo r^ O . dt


115 The magnitude o f

dM : —, as well as the age invariant use o f medical inputs may dt

be correlated or interact with at least some o f the X {'s. Finally, it is reasonable to hypothesize th a t

dM,. . J „ —> 0 . Mothers who have delayed childbearing (with advice from dt

health care providers) may seek to mitigate the effects of delayed childbearing by using additional medical inputs or they may use additional inputs because they may be more concerned about their future prospects for childbearing or have greater demand for reducing quality uncertainty. Note that the above statements regarding the higher socioeconomic status o f women who delay childbearing only apply for first births. Since women o f lower socioeconomic status tend to have higher fertility rates and hence greater number o f births o f higher parities at younger as well as older ages [Chapter 2], the socioeconomic characteristics o f births to older women at higher parities [i.e., birth order > 1 ] is likely to be lower than the socioeconomic profile o f women who have first births at older ages.

5.3

Data sources The Linked Infant Birth and Death data for the United States for the years 1989 to 1991 1 are the data source for the study. The linked infant birth and death data

for the United States are from the National Center for Health Statistics at the Centers for Disease Prevention and Control that began linking birth and infant death data for public use in 1983. The data set links essentially all live births and infant (up to one


year o f age) deaths in the United States. For each record the data includes infant’s birth weight, gestational age, survival status and cause o f death for infonts who died. The data also includes sociodemographic (e.g., age, education, marital status, race/ethnicity, state of residence) and reproductive history (e.g., parity, interval since last birth, prenatal care) for the mother and the sociodemographic characteristics o f the father. There was a major revision and expansion o f the data sets beginning in 1989 *As a result, several important items including maternal tobacco and alcohol use during pregnancy, presence o f several medical and obstetrical high risk conditions (e.g., preeclampsia, gestational diabetes) and congenital anomalies (e.g., Down syndrome, Neural Tube Defects) were added to the data sets.

5.3.1

List o f variables and measurement issues

5.3.1.1 Adverse birth and infant outcomes (dependent variables) The adverse birth outcomes in the study include low birth weight (< 2.5 kg) further subdivided into moderately low (1.5-2.5 kg) and very low (< 1.5 kg) birth weight categories. Preterm (< 37 weeks) delivery is similarly subdivided into moderately preterm (32-37 weeks) and very preterm (< 32 weeks) gestations. These outcomes will be coded as binary variables. The cutoff points are somewhat arbitrary and do result in some loss o f information as the underlying variables, birth weight and gestational age, are by nature continuous variables. However, low birth weight,


117 preterm delivery and their corresponding subcategories are established as benchmarks in the epidemiological literature on adverse birth outcomes and reflect the fact that for both birth weight and gestational age, the risk for mortality and morbidity in the newborns decreases sharply as birth weight and gestational age increase beyond the extremely low values o f birth weight and gestational age and becomes relatively flat thereafter 2'4. Birth weight can be measured unambiguously and with fairly high degree o f accuracy and is presumably reported as such on the majority o f birth certificates The observed relationship between birth weight and mortality in the data sets used in the study also supports accuracy o f birth weight data. Gestational age measurement is more ambiguous and subject to a higher degree o f measurement and reporting error. For neither birth weight o f gestational age, however, there is any evidence or a priori reason to think that measurement or reporting error or the frequency o f missing values might be related to mother’s age. The overall birth prevalence o f Down syndrome and congenital anomalies is likely to be underreported in birth certificate data 5. However, previous reports have not found any differential misclassification (i.e., under/over reporting) o f Down syndrome by maternal age 5’6. Furthermore, results o f the analyses [Chapter 6 ] show that the relationship between maternal age and the risk o f Down syndrome is consistent with previous literature with a flat relationship between maternal age and


118 risk o f Down, syndrome up to ages 30 to 35 and an exponential rise in the risk for Down syndrome as mother’s age advances beyond 35 years o f age. For the cause-specific analysis o f infant mortality, the 61 causes o f infant deaths reported by the National Center for Health Statistics are grouped into 11 causes o f death following a modified version o f the grouping suggested by Dollfixs et a l 7: congenital anomalies (ICD 740-759), sudden infant death syndrome (SIDS ICD 798), respiratory distress syndrome (RDS —ICD 769), prematurity (ICD 765), pneumonia (ICD 480-487), maternal complications (ICD 761), hypoxia and asphyxia (ICD 768), complications o f placenta (ICD 762) ), infections (ICD 771), unintentional injuries (ICD E800-E949) and intentional injuries (ICD E960-E969). Deaths due to intentional injuries include infant deaths due to “child battering or other maltreatment” and “other homicide”. Unintentional deaths include accidental deaths due to suffocation or choking, motor vehicle accidents, drowning, falls, fires, exposures to excessive heat or cold, being struck by falling objects and other unintentional injuries. The National Center for Health Statistics data on causes o f infant death is the main source o f reporting on cause-specific infant mortality data in the United States. Furthermore, the accuracy o f death certificate data for some o f the causes o f death and in particular injury-related deaths has been validated 8' 10. On the other hand, deaths due to intentional injuries (homicides) are probably underreported in the birth certificate data 11,12.


119 Data on mortality excludes unlinked deaths. Unlinked deaths accounted for less than 2.5 % o f all deaths. The distribution o f unlinked deaths is not entirely random - for example, African Americans have slightly higher rates o f unlinked deaths as do very low birth weight babies. Therefore, the mortality estimates for these groups may slightly underestimate true rates. There is no evidence, however, o f linkage varying by maternal age and in any case the small percentage o f unlinked deaths would not be expected to produce effects that would be substantially important in the study.

5.3.1.2 Predictor [right hand side (rhs) variables] In addition to maternal age, the following set o f predictor variables will be used in the analyses for estimating the adjusted effects o f maternal age on the risks for adverse birth and infant outcomes: 1 ) socioeconomic characteristics o f the parents: marital status, maternal education and parental education; 2 ) maternal behaviors during pregnancy: cigarette smoking and alcohol intake during pregnancy and adequacy o f prenatal care; and 3) reproductive history, birth order (parity), and for mothers with parity > 1 , previous preterm delivery and interval since last pregnancy [computed from interbirth interval reported in the data set minus the gestational age o f the index pregnancy]. Data on socioeconomic characteristics o f the parents do not include income or occupation. However, previous reports support the validity o f maternal education as a


120

proxy for socioeconomic status in studies o f birth and infant outcome 13. For maternal behaviors during pregnancy, cigarette smoking is reported as the (average) number o f cigarettes per day and alcohol intake as the number o f drinks per week. Both o f these measures are based on self-reports o f mothers and in the case o f alcohol intake a substantial number o f records have missing values. Prenatal care is reported as the Kessner Index 14 that is based on the time o f initiation o f prenatal care, number o f prenatal visits and gestational age o f the newborn, as well as, the individual items o f the Kessner Index. The variable maternal age is presumably measured and reported fairly accurately. However, as noted above, some o f the predictor variables may be subject to measurement and/or reporting error. However, there have not been any previous reports documenting biased measurement error in any o f the predictor variables according to maternal age. There is also no a priori reason to assume that measurement errors might not be neutral for younger vs. older mothers. Nevertheless, to the extent that the true variables (i.e., when measured without error) are correlated with maternal age, the measurement error may be transmitted to the estimates o f the age effect even if measurement error itself is uncorrelated with age. It is difficult, however, to predict and gauge the exact extent o f any biases in the estimates o f age effect that may be due to measurement error in rhs variables especially in the case of several rhs variables.


The frequency o f missing values is also not known to vary by maternal age and preliminary analyses did not reveal any substantial differences in the frequency o f missing values by maternal age. In the empirical analyses, births with m ining predictor variables tire retained by including dummy variables indicating an “unknown” category for each predictor variable.

5.4

Empirical specifications The following sections present a general discussion o f the specification issues

with regard to estimating the age effects (advanced maternal age) on the risks for adverse birth and infant outcomes. This first section focuses on the main statistical issue of concern; i.e., the possible effects o f unobservable/unobserved (omitted) variables in the empirical models for estimating the age effects on adverse outcomes. Use o f proxy variables and endogenous right hand side (rhs) control variables and the consequences in terms o f the possible biases in the estimates o f age effect are discussed. In the second section, the specific strategy for estimating effect modification by medical inputs —i.e., how the adverse effects o f age might be modified or compensated by the use of medical inputs is discussed. The main medical inputs in the study are use o f prenatal care, amniocentesis and ultrasound.


122 5.4.1

Use o f proxy variables This section presents a discussion o f the consequences o f estimating age

effects when one is not able to include an unobservable variable that might account for possible socioeconomic, behavioral or biological differences in baseline [i.e., unrelated to the effect o f aging per se] risk for adverse outcomes in younger vs. older mothers - and the choice o f using proxies for the unobservable variable (e.g., by using maternal education, prenatal care, other observable variables in the data sets) or the misspecification o f the omitted variable equation. The following discussion o f the use o f proxy variables uses the ordinary least squares (OLS) specification. As described in more detail in the sections that follow, the actual estimations in the dissertation are done using logistic regression and Cox proportional hazard models that more appropriately estimate the risk o f binary adverse outcomes or differentials in cause-specific hazards o f infant mortality. However, the substantive results below should still hold as first order approximation. The following derivation follows the work by Wickens

15

that deals with the

issue of using proxies and derives the formulas for the bias equations for the omitted variable vs. the proxy case. Suppose the true model to be estimated is: (1)

y — Tfi + Z y + U ,


123 where T represents the matrix o f observations on k fixed variables (e.g., vector o f age effects), but Z, another fixed variable, is not observable but is in the true modeL, (y *

0

). Consider two estimates o f the age effects that are possible given available

data: (i) that obtained by OLS from the misspecified equation: (2)

y = T/3 + v

denoted by t>2 and (ii) that obtained by OLS from the equation (3)

y = T /3+ P S + e

denoted by b3 , which includes P, an observable proxy variable for Z that is related to Z through the model: (4)

P = Z 0 + s.

T herefore: y = T/3 + Z y + u = Tfi 4 - {Z.0 + e ) S + e and hence: (5) y = 0 8 a n d e = u — Se.

The large sample bias o f \>i from the omitted variable model is: (6 ) p lim(£>, - J3) = M ^y, where — Iim„_„ T 'T I n i s assumed to be finite and non - singular and M j? - limn_ a3 T'Z / n is assumed to be finite. Hence, b2 is asymptotically unbiased if either or yare zero. The asymptotic bias o f b 3 from the model that includes the proxy variable can be obtained from: -i

(7)

plim

b3

P

M -n

d ,

5

M pr

1

\ M t' ] M pp

where M ab = lim„

l M Pe\


a ’b / n.

124 The inverse o f the 2 x 2 partitioned matrix can be written as: Afjy .A'fpp

jp AdTpp

1

r A /^ ( / + M tpF2 M pt A f £ ) —F, M pj

M ^-M tpF2 where F2

F2 = ( M pp - MpjM^-M-ppY1 Also: M Te = 0 and M Pe = -8 cr2. Afpp(I hfppF2MPT■^Tt') —F2 Af pj' Afj~p

Hence: plim V A

—hdpp M tp F2 So*

8 ct^-M pj- M pp F2 hd yt hd yp = & r |F 2 •s

_ —

S a 2tF2

-/

.

b /tj-p A'f'j'p

M pp — M.pp Mpp Afpp )v - i

-/

( 8)

But: (9) Mpp —MppMppM.pp — 0 and E (y )< 0 . The OLS bias in the estimate o f age effect that results from including an observable endogenous rhs variable, such as the observable educational level o f women, can be derived by the following expression that is in terms o f plim’s since the endogenous proxy is a random variable rather than a fixed one: plimi V A .

1 8

r

~TT P'T

-1

P'P

1

(T'D (P 'P ) ~ (T 'P X P 'T ) Therefore: plim(Z^

T 'P ' P'P

_ p ,T

'T V P'e -T 'P j 0 ‘ T ,T I p ,e 1

_-[-T'PP'e] [(T'T)(P'P)-(T'P)2 J

^ — .• — P ) — plimi

Note that { T ' T ) { P ' P ) - i T ' P f > 0 and T'P > 0. Hence if: (i) plim (P'e) > 0 , b2 is biased downward from /?, whereas if: (ii) plim(P'e) < 0, b3 is biased upward from /?. Note that in contrast to the case o f using a proxy variable to control for an unobservable variable discussed above, in the general case o f using an endogenous rhs variable as a control variable, the bias in the age estimate (bz) cannot be said to be unambiguously less when the control variable is included as compared with the omitted variable estimate (bz). Furthermore, one cannot unambiguously sign the bias based on the empirical relationship between bz and

63.

For example,

63

>

62

would be

consistent with either the case o f both bz and 6 3 underestimating the true effect with bz


127 being less biased than

bz

if (1) is true, or the case o f 6 3 overestimating the true

effect and bz underestimating the true effect if (ii) is true. Nevertheless, the above derivation o f the expected bias in the age estimate in the presence o f an endogenous rhs control variable does yield several positive results that can be used for bounding the estimates o f age effects. First, regardless o f whether (i) or (ii) is true, if both b z and

63

are positive,

then it can be said unambiguously that the true estimate is also positive and that the risk for adverse outcomes does increase with age. Furthermore, if evidence from previous literature suggests that (i) is more likely, both bz and

63

are underestimates o f

the true effect and hence the less biased o f the two would be the larger estimate and may be taken as the minimum effect o f age. On the other hand, if (ii) is deemed more likely, then bz underestimates whereas bz and

63

63

overestimates the true age effect and hence

can be taken as the lower and upper bounds, respectively, for the true age

effect. Finally, note that the above statements hold as long as age itself is uncorrelated with the error term once the proxy or the endogenous rhs control variable is included in the regression for estimating the age effect. If age remains correlated with the error term then the bias in the estimates depends on the sign of:


128 p Iim {(P 'P )(rre) - ( T rP)(P'e)}. Hence if: (i) plim (r'e) < 0 and plim (P'e) > 0 then 6 3 is unambiguously biased downward, whereas if: (ii) plim (r'e) > 0 and plim (P'e) < 0 then b3 is biased upward. On the other hand, if: (Hi) plim (P'e) and plim (r'e) have the same sign, then the direction o f the bias is ambiguous.

As discussed in Chapters 2 and 3, previous literature suggests that to the extent that P may not fully capture the favorable socioeconomic and behavioral profile o f mothers who delay childbearing, plim(Pe) may be positive and hence both bz and 6 3 would underestimate the true age effect. In addition, as noted in Chapter 2, previous studies suggest that delayed childbearing by young women is for the most part a matter o f choice and not due to any reproductive dysfunction which makes the possibility o f a positive correlation between age and the error term less likely. Nevertheless, unaccounted biologically- or behaviorally mediated adverse selection for older mothers might result in plim(7e) to be positive which, all else equal, would tend to produce overestimates o f the age effect with the final direction o f the bias depending on the direction o f the correlation between T and P and the error term. In the analysis o f second and higher order births, two proxies of reproductive function, interpregnancy interval and previous preterm birth, are used as rhs variables that can serve as controls for adverse selection. Nevertheless, given the observational nature o f the analysis, and unavailability o f appropriate instruments for an instrumental variable estimation strategy, the issue o f adverse selection cannot be fully resolved


129 with the available data here. Therefore, the estimates o f the age effect will be presented and discussed with appropriate elaboration o f the caveats and limitations with regard to the magnitude and direction o f the bias as noted above. In addition, the biological mechanisms discussed in Chapter 4 can substantiate a basis for the empirical estimates o f the age effect. Given the uncertainty with regard to bias in the estimates o f age effect as discussed above, one may wish to take the lesser o f b2 (unadjusted) and b3 (adjusted) estimates as the more conservative estimate o f the age effect. In that case, the question o f the sampling distribution o f the resulting p arises given that b2 and b3 are not independent. To find this distribution, one option would be to obtain the distribution by bootstrap methods, which would be difficult given the sample sizes in this study. Another option would be to find the cumulative distribution function (cdf) o f the distribution by numerical integration. In practice, the b2 and b3 estimates obtained empirically turned out to be fairly close and almost always in the same direction with b3 (adjusted) estimates o f the age effect equal to or greater than the b2 (unadjusted) estimates. Therefore, while the problem o f finding the distribution o f f5 was a theoretical possibility, in practice it was not a substantial question in terms o f the obtained empirical estimates o f the age effect.


130

5.4.3 Use o f medical inputs The medical inputs in the study are use o f prenatal care and amniocentesis. Estimating the impact o f prenatal care on low birth weight and other birth outcomes is subject to both adverse and favorable selection biases that relate to the mothers’ and health care providers’ choice o f timing and frequency o f prenatal care in relation to the underlying risk profile o f mothers. An extensive biomedical as well as social science literature is concerned with this subject. Researchers who have studied this question carefully

16

have often concluded that most, if not all, o f the beneficial

effects ascribed to prenatal care are produced as a result o f favorable selection —i.e„ mothers who are at lower risk for adverse outcomes tend to follow what is considered an adequate schedule o f prenatal care visits. Given, the main purpose o f the dissertation, i.e., estimating the effect o f advanced maternal age on adverse birth and infant outcomes, and limitations o f the available data, the present analysis will not attempt to resolve the selection issues as they relate to estimation of the effect o f prenatal care. Instead, use o f prenatal care is accounted for in empirical models (as a right hand side variable in logistic regression and Cox proportional hazard models) for estimating age effects in order to arrive at adjusted estimates o f advanced maternal age that are not due to either adverse or favorable selection by maternal age as might be partially proxied by prenatal care, and/or by differential use and hence possible benefit from prenatal care services. In this sense, prenatal care will be used as one component o f the observables (i.e.,


131 control rhs variables). Again, as discussed in the previous section, the potential role o f unobservables leads to ambiguity about the biases that might be inherent in the crude (omitted variable) and/or observables-adjusted estimates o f age effects in models that include prenatal care. In contrast to prenatal care, amniocentesis does have a directly established mechanism in reducing the adverse birth outcomes associated with structural birth defects —i.e., Down syndrome and neural tube defects. Prenatal screening methods, and conclusive diagnostic techniques, are available for the early pregnancy detection o f birth defects. Current noninvasive screening strategies include the use o f maternal serum biochemical methods 17, and to an increasing extent by early pregnancy ultrasound evaluation for anatomic markers associated with aneuploidy I8*1919. Definitive prenatal diagnosis, on the other hand, requires invasive fetal testing most commonly by amniocentesis, occasionally by chorionic villus sampling, and in rare instances fetal blood sampling. Traditionally, invasive fetal testing has been offered to empirically “high risk” groups including (a) women o f “advanced maternal age” defined somewhat arbitrarily as age 35 or more at time o f delivery, and (b) selected women with a relevant family history. Women choosing to undertake prenatal diagnostic procedures are known to have a high utilization o f abortion when an affected fetus has been cytogenetically identified 20. In the dissertation, data on age-specific use o f amniocentesis are used to estimate how the effect o f advanced maternal age on the risk for birth defects is


132 modified (or compensated) by use o f prenatal diagnostic technologies. The main questions o f interest in this part o f the project are: 1 ) to what extent the effects o f advanced maternal age are compensated by the use o f amniocentesis in different socioeconomic groups [i.e., as defined by level o f education and/or race/ethnicity]? And 2) whether differential use o f prenatal diagnosis services increase socioeconomic differences in birth outcomes at older ages? The empirical strategy to answer these questions is different from what has been discussed so far. This is due to the fact that the available data sources only include information on live births and do not include data on fetal deaths due to spontaneous and/or elective terminations o f affected or normal fetuses. This is particularly important for estimating the effect o f prenatal diagnosis services (e.g., amniocentesis) where the impact on birth outcomes is by diagnosis followed by selective termination o f affected fetuses. The ideal data set would therefore include data on the outcomes of all pregnancies and not just those that end in live births, as is the case in the available data sets. Therefore, in light o f the data limitations, in order to estimate the effect o f amniocentesis on the risk for birth defects, an indirect and more tentative empirical strategy that leads to suggestive rather than definitive inferences is used. The twostep strategy and a discussion o f its attendant limitations are described below.


133 First, empirical models are estimated to test whether the effect(s) o f advanced maternal age do indeed vary by socioeconomic characteristics o f the women, e.g., maternal education and/or race/ethnicity. More formally, suppose y g = Ti^ li + X i/?2 i + U xfisi + £ as in (1) above except that i and j index the ith birth defect (e.g., Down syndrome) and j t h socioeconomic group. Next, the effect o f membership in group i on the age-specific rates o f amniocentesis are estimated: mv = T f a I(i + X ia 2i, -t-UiOTjjf +rj , as above except that mj is the frequency o f the jth medical input (e.g., amniocentesis) among live births —i.e., the birth prevalence of the jth medical input as distinct from its frequency of use among all pregnancies and anj is the vector o f age effects for group i and medical input j, X,- is the vector of observable socioeconomic and behavioral characteristics o f group i (o f course other than what is indexed by i) that may be correlates o f advanced age in group i, as well as, medical input j, oi2 ij is the vector o f effects associated with X in group i an d , U,- is the vector o f variables in group i that are unobservable / unmeasured / omitted from the data sets used in the study that may be correlates o f age, use o f medical input and/or the observable variables (X; or what is indexed by i), and a 3,-j is the vector o f the effects associated with the unobservables.


134 Since we can only estimate models that include the observables, the same arguments as discussed in the previous section imply that, in general, the age effects in group i for birth defect j, as well as, for medical input j, may be biased and that the sign and magnitude o f any such biases would be ambiguous. In addition, since the relevant effect o f medical inputs here presumably operates through the unobserved selective terminations, a direct estimate o f the effect o f medical inputs is not possible. Furthermore, it is only possible to perform approximate and indirect consistency checks between the differential usage rate o f medical inputs vis-a-vis differences in the effects o f advanced maternal age on birth defects across socioeconomic groups. This latter limitation is due to both the unobserved terminations and hence unobserved frequency o f amniocentesis among terminations / spontaneous abortions, as well as, due to the fact that amniocentesis (and ultrasound studies) are done for indications other than prenatal diagnosis o f the congenital anomalies in the study. Therefore, the empirical estimates can only be shown to be qualitatively and (at least approximately quantitatively) consistent with a pattern where differential use o f medical inputs results in increases in the socioeconomic disparities in birth outcomes at older ages. However alternative explanations cannot be ruled out with certainty given data limitations and potentials for bias / confounding due to unobservables. Hence the conclusions must remain tentative and suggestive rather than definitive. The birth defect in the study, Down syndrome, is chosen to represent a major birth defect in terms o f its population impact and also to have a well-known


135 biological mechanism and epidemiological features (Chapter 4). In addition, it can diagnosed prenatally (e.g., by amniocentesis) and is hence subject to selective termination. Therefore, the known biological and epidemiological attributes o f Down syndrome can be used to complement the discussion of empirical results and to assess the possible role o f unobservables or any related biases.

5.5

Model specifications The analyses in the dissertation includes the following specifications to assess the

effect o f advanced maternal age on the risks for adverse outcomes while taking into account other socioeconomic and behavioral factors that might be associated with both delayed childbearing and adverse outcomes: 1 ) logistic regression models for low birth weight, preterm delivery and birth prevalence of Down Syndrome and 2) competing risks hazard models o f cause-specific infant mortality.

5.5.1

Logistic regression models

Logistic regression models 21 are used to estimate the net effect o f advanced maternal age on the risk o f low birth weight, preterm delivery and birth prevalence o f Down syndrome. The set o f control (rhs) variables includes maternal education, martial status, race/ethnicity, prenatal care, interpregnancy interval, birth order and maternal smoking. Both main effect models and models that include interaction terms with particular variables o f interest, parity, maternal education and race/ethnicity are


136 estimated. Birth data for the years 1989-1991 from the National Center for Health Statistics are used for this set o f analyses.

5.5.2

Competing risks hazard models o f cause-specific infant mortality Effect o f delayed childbearing on overall infant mortality, as well as, cause-

specific infant mortality due to

11

causes o f death: maternal complications,

complications o f placenta, hypoxia and asphyxia, prematurity, congenital anomalies, respiratory distress syndrome (RDS), pneumonia, infections, sudden infant death syndrome (SIDS), unintentional (accidents) and intentional injuries (homicides) is considered. Linked birth and infant death data for the year 1990 are used for this set o f analyses. Studies o f cause-specific mortality are subject to the competing risk problem 22, which is related to the fact that individuals are exposed to several risks o f death which compete with one another for the life o f a person. It is necessary to resolve the competing risks problem in order to:

0

determine how the probability o f death for a

given cause will change when the probability o f death for another cause changes and ii) assess the impact o f covariates on cause-specific hazards o f death. Suppose that a population is subjected to K causes o f death, ci,C2 ,—,cic with a random vector, T = (Ti, T 2 ,...,Tk) representing the times o f death from ci,C2 ,...,ck respectively. The competing risks problem arises when one wishes to estimate a “net”


137 or “pure” hazard rate o f death from a specific cause o f death in the absence o f the other causes o f death When only cause-specific times at death are known, the competing risk problem is usually solved by assuming cause-specific hazard rates to be mutually independent. This is the method used by demographers to calculate multiple cause o f death (multiple decrement / single decrem ent) life tables. Tsiatis

23

showed that

unless one accepts the unverifiable and the often untenable assumption that causespecific hazard rates are mutually independent, a given set of crude survival probabilities does not identify the corresponding net probabilities - i.e., one cannot obtain a unique solution for the cause specific net hazard rates. This is true when there are no regressors or covariates. Yashin et a l 24 and Heckman and Honore

25

show how access to regressors overturns the non-identification theorem o f Cox and Tsiatis for both proportional hazard and accelerated failure time models. In its simplest form hazard models that include covariates can be estimated by assuming cause-specific hazard rates to be conditionally independent given the set o f regressors . While this assumption remains subject to Tsiatis’ criticism, it is a weaker and an arguably more defensible assumption when strong predictors o f mortality such as birth weight are available as is the case here. For the set o f analyses here, the 61 causes o f infant deaths reported by the National Center for Health Statistics are first grouped using a modified version o f the grouping suggested by Dollfiis et a l 1. Cox proportional hazard models o f cause-


138 specific hazard, o f mortality are then estimated assuming conditional independence o f the times to death for different causes o f death given the set o f covariates in the model. Effect o f delayed childbearing on the hazard for all cause mortality, as well as, cause-specific mortality due to

11

causes o f death: maternal

complications, complications o f placenta, hypoxia and asphyxia, prematurity, congenital anom alies, respiratory distress syndrome (R D S ), pneumonia, infections, sudden infant death syndrome (S I D S ), unintentional (accidents) and intentional injuries (homicides) is estimated. Three sets o f models are estimated for each cause o f death and for all causes combined. Model I estimates the crude risk o f infant mortality associated with advanced maternal age for each cause o f death and all-cause mortality (crude age effect model). Model II estimates the effect o f advanced maternal age adjusted for the potentially confounding effects o f maternal education, marital status, race/ethnicity, parity, interpregnancy interval, prenatal care, tobacco and alcohol use during pregnancy (adjusted age effect). Model III estimates the adjusted effect o f advanced maternal age on infant mortality conditional on the birth outcome; the model includes birth weight as well as the set o f sociodemographic and reproductive history characteristics.

Reference List (1) National Center for Health Statistics. Linked Birth/Infant Death Data Set: 1991 Birth Cohort. Public Use Data File Documentation. 1995. Hyattesville, MD, Centers for Disease Prevention


139 and Control, National Center for Health Statistics. Ref Type: Data File (2) Lee KS, Paneth N, Gartner LM, Pear 1man M. The v e r y low-birth-weight rate: Principal predictor of neonatal mortality in industrialized populations. Journal o f Pediatrics 1980; 97(5):759-764. (3) Wilcox AJ, Russell IT. Birthweight and perinatal m ortality: III. Towards a new method of analysis. International Journal ofEpidemiology 1986; 15(2):188-196. (4) Lee KS, Khoshnood B, Hsieh H, Kim BI, Schreiber Bv4D, Mittendorf R. Which birthweight groups contributed most to the overall reduction in tttie neonatal mortality rate in the United States from I960 to 1986? Paediatric & Perinatal Epidemiology 1995; 9(4):420-430. (5) Centers for Disease Control and Prevention. Down syndrom e prevalence at birth—United States, 1983-1990. MMWR 1994; 43:617-622. (6) Bishop J, Huether C, Toris C, Lorey F, Deddens J. EZpidemioIogic study of Down syndrome in a racially diverse California population, 1989-1991. A-merican Journal ofEpidemiology 1997; 145(2):134-147. (7) Dollfus C, Patetta M, Siegel E, Cross AW. Infont moortality: a practical approach to the analysis of the leading causes of death and risk factors. Pediaatrics 1990 Aug 86:176-183. (8) Fife D. Injuries and deaths among elderly persons. /American Journal ofEpidemiology 1987; 126:936-941. (9) Olsen SJ, Durkin MS. Validity o f hospital discharges data regarding intentionality of fatal pediatric injuries. Epidemiology 1996; 7:644-647. (10) Dijkhuis H, Zwerling C, Parrish G, Bennett T, Kemjper HCG. Medical examiner data in injury surveillance: a comparison with death certificates. A*merican Journal ofEpidemiology 1994; 139:637-643. (11) Jason J, Carpenter MM, Tyler CW. Underrecording o f infont homicide in the United States. American Journal o f Public Health 1983; 73:195-19**7. (12) Ewigman B, Kivlahan C, Land G. The Missouri Chnld Fatality Study*. Underreporting of maltreatment fatalities among children younger tham five years of age, 1983 through 1986. Pediatrics 1993; 91:330-337. (13) Parker JD, Schoendorf KC, Kiely JL. Associations boetween measures of socioeconomic status and low birth weight, small for gestational age, and premature delivery in the United States. Ann Epidemiol 1994; 4:271-278. (14) Kotelchuck M. An evaluation of the Kessner adequaacy o f prenatal care index and a proposed adequacy of prenatal care utilization index. A m erican Journal of Public Health 1994; 84:14141420. (15) Wickens MR. A note on the use o f proxy variables. CEconometrica 1972; 40:759-761.


140 (16) Fink A, Yano EM, Goya D. Prenatal programs: What the literature reveals. Obstet Gynecol 1992; 80:867-872. (17) Haddow JE, Palomaki GE, Knight GJ, Williams J, Pulkkinen A, Canick JA et al. Prenatal screening for Down's syndrome with use o f maternal serum markers [see comments]. N Engl J Med 1992 Aug 27 327:588-593. (18)

Benacerraf BR. The second-trimester fetus with Down syndrome: detection using sonographic features. Ultrasound Obstet Gynecol 1996 Feb 7:147-155.

(19) Taipale P, Hiilesmaa V, Salonen R, Ylostalo P. Increased nuchal translucency as a marker for fetal chromosomal defects [see comments]. N Engl J Med 1997 Dec 4 337:1654-1658. (20) Verp MS, Bombard AT, Simpson JL, Elias S. Parental decision following prenatal diagnosis of fetal chromosome abnormality. Am J Med Genet 1988 Mar 29:613-622. (21) Hosmer JrD, Lemeshow S. Applied logistic regression. New York: John Wiley & Sons, Inc., 1989. (22) Chiang CL. Competing risks in mortality analysis. Annu Rev Public Health 1991 12:281-307. (23) Tsiatis A. A nonidentifiability aspect o f the problem o f competing risks. Proc Natl Acad Sci U S A 1975 Jan 72:20-22. (24) Yashin Al, Manton KG, Stallard E. Dependent competing risks: a stochastic process model. Journal of Mathematical Biology 1986; 24(2): 119-140. (25)

Heckman JJ, Honore B.E. The identifiability of the competing risks model. Biometrika 1989; 76(2):325-330.


141

CHAPTER

6

- RESULTS

This chapter details the results o f the empirical analyses for estimating the effects o f advanced maternal age on the following sets o f outcomes: 1 ) low birth weight and preterm delivery, 2) Down syndrome and 3) cause-specific infant mortality. In addition, the distribution o f maternal age at first birth and births o f second or higher order are presented for the four ethnic groups in the study: African Americans, Mexican Americans, non-Hispanic whites and Puerto Ricans. The chapter is organized as follows: Sections 6 .1 reports the results o f the analyses o f the effects o f age on low birth weight and preterm delivery that pertain to first births and Section 6.2 reports the results for births o f second or higher order. The analyses are stratified by ethnicity and are carried on with and without adjustment for socioeconomic and behavioral correlates o f delayed childbearing including prenatal care. This set o f analyses allows investigation o f the hypotheses on interactions between maternal age with birth order and ethnicity as outlined in Chapter 5 (Hypotheses 1-3). After a brief description o f maternal age distribution for first births in the data set (Section 6.1.1), Section 6.1.2 reports the results o f the analyses o f the effects o f maternal age at first births on low birth weight. Crude (unadjusted) effects o f age on very low (< 1.5 kg) and moderately low (1.5-2.5kg) birth weight are compared with the effects o f age adjusted for maternal education, marital status, smoking and


142 prenatal care. In this set o f analyses, maternal ages o f 35 years and older are compared with the reference group o f 20-34 years o f age. This set o f analyses are then followed by a more detailed analysis o f the effects o f age comparing the effects o f maternal age 30-34, 35-39, and 40 years o f age and greater with the reference group o f 20-29 years o f age (Section 6.1.3). Finally, in order to assess the effect o f prenatal care, the analyses are done with and without adjustment for prenatal care (Section 6.1.4). Sections 6.1.5 to 6.1.7 report the same set o f results for preterm delivery. Section 6.2 is organized similarly to Section 6 . 1 and reports the results for births o f second or higher order. Section 6.3 reports the results o f the analysis comparing the effects o f maternal age on the risk o f Down syndrome across ethnic and education groups. The interest here is primarily to define the interactions between maternal age, ethnicity and education. As noted in chapter 5, the effects o f maternal age on the risk for Down syndrome might vary by ethnicity and education (Hypothesis 4). Specifically, the risk might be greater for African Americans and Mexican Americans due to lesser access to or use o f prenatal diagnosis services and selective termination o f affected fetuses. The age-related risk o f Down syndrome might also be greater for women with lower levels o f education within a given ethnic group due to lesser use o f prenatal diagnosis. Section 6.4 cause-specific infant mortality reports the results o f investigation with regard to the differential effects o f advanced age on cause-specific infant mortality (Chapter 5, Hypothesis 5). Specifically, the aim is to investigate the


143

hypothesis that advanced maternal age increases the risk associated with biologically related causes o f infant mortality: infant deaths due to congenital anomalies and abnormalities o f the placenta (placenta previa and abruptio placentae). On the other hand, advanced maternal age might be associated with lower risks o f infant mortality due to the set o f causes that are related to childcare practices such as infant deaths due to unintentional and intentional injuries. This association might be due to the effect o f birth timing itself or to the effect o f unmeasured socioeconomic factors associated with delayed childbearing or both.

6

.1

6.1.1

Effects o f age on low birth weight and preterm delivery - First births First births - maternal age distribution Table 6.1(a) shows distribution o f maternal age at first birth for four ethnic

groups in the United States for the years 1989-1991. The results are consistent with previous data (Chapter 2) that document delayed childbearing to be more pronounced for non-Hispanic whites as compared with other groups. For first births, the proportions o f infants bom to women 35 years o f age and greater were 5.6 % (95 % binomial confidence interval [C.I.], 5.6-5.7%) for non-Hispanic whites, 3.2 % (95 % C.I., 3.1-3.2%) for African Americans, 1.9 % (95 % C.I., 1.8-2.1%) for Puerto Ricans and 1.4 % (95 % C.I., 1.4-1.5%) for Mexican Americans in the period 1989-1991.


144

6.1.2 First births - low birth weight Table 6.2(a) shows the association o f delayed childbearing with low birth weight. For all four racial/ethnic groups, maternal age > 3 5 years at first birth was associated with statistically significant and substantial increases in the risks for low (< 2.5 kg) birth weight (Table 6.2(a)). Compared with mothers 20-34 years o f age, mothers > 3 5 years were at 82 % increased risk o f very low (< 1.5 kg) birth weight [Table 6.2(a) - Mantel-Haenszel Odds Ratio: 1.82, 95 % C.I.: 1.75-1.90] and 55 % increased risk o f moderately low (1.5-2.5 kg) birth weight [Mantel-Haenszel Odds Ratio: 1.55, 95 % C.I.: 1.52-1.58]. The effects o f maternal age > 35 years on low birth weight were fairly similar in direction and in magnitude among the four racial/ethnic groups particularly after adjustment for other sociodemographic and reproductive history characteristics (Tables 6.2(a) and 6.4(a)). However, Puerto Ricans and particularly African Americans have two-three fold higher ‘base line’ risks o f low birth weight. Therefore, equivalent odds ratios for the effect o f delayed childbearing in African Americans and Puerto Ricans signify greater differences in terms o f risk differences or absolute elevations in risk for low birth weight. This can be seen in Table 6.2(c) where risk differences in very low, moderately low and low birth weight that were associated with advanced maternal age at first birth are shown for all four ethnic groups. The highest risk differences in very low and moderately low birth weight, and hence the


145 highest overall risk differences in low birth weight, pertained to African Americans and Puerto Ricans and the lowest to non-Hispanic whites. Specifically, in the case o f low birth weight, the risk differences for low birth weight were 5.3 % (95 % C.I., 4.7-6.0) for African Americans, 4.3 % (95 % C.I., 1.7-6.9) for Puerto Ricans, 3.7 % (95 % C.I., 2.8-4.5) for Mexican Americans and 2.6 % (95 % C.I., 2.4-2.7) for non-Hispanic whites. Table 6.3(a) shows the low birth weight estimates for the attributable fractions among the exposed and the population attributable risk percentages for the population as a whole related to advanced maternal age at first birth. Estimates o f the attributable fractions among the exposed (AFe) suggest that about 40 % o f all low birth weight and 50 % o f veiy low birth weight infants in women with maternal age > 35 years are due to the ‘exposure’ o f advanced maternal age. On the other hand, the population attributable risk percentages (PAR.%) estimated for the ‘population’ o f first births in the period 1989-1991 suggest that only 3-4 % o f all low birth weight and 4-5% of all very low birth weight first births were due to women with advanced maternal age. This is because despite recent trends, mothers > 35 years still comprised a small minority o f all mothers at first births (about 5% or less - see Table 6.1(a)) in this period particularly for groups other than non-Hispanic whites. On the other hand, note that these analyses only included first births and do not capture the impact o f delayed childbearing on birth outcomes for higher parities (see Tables 6.3(b) and 6.3(c) and below for the discussion o f births second or higher order). Furthermore, with the


146 continuing trends towards delayed childbearing, these estimates of the populationlevel impact o f delayed childbearing are probably low by a factor o f about 40-50 % for the current levels o f advanced maternal age. Hence the estimates o f age effect suggest that with the current maternal age distribution, approximately 6 - 8 % o f low birth weight and 8 - 1 0 % o f very low birth first births are due to women with advanced maternal age. Table 6.4(a) presents results o f logistic regression models o f low birth weight that adjusted the effect o f maternal age > 3 5 years at first birth for maternal education, marital status, prenatal care and tobacco use. For all groups, the adjusted effects o f advanced maternal age remained significant and o f the same or increased magnitude after adjusting for other factors. Adjustment for other factors significantly increased the risk associated with advanced maternal age for non-Hispanic whites by about 10 % but not for others. The adjusted odds ratio for low birth weight for non-Hispanic whites was 1.73 (95 % C.I., 1.69-1.76) as compared with their crude odds ratio o f 1.59 (1.55-1.62). Similarly, the adjusted odds ratio for very low birth weight was 1.94 (95 % C.I., 1.85-2.04) as compared with their crude odds ratio of 1.81 (1.72-1.90) and for moderately low birth weight, the adjusted odds ratio was 1.68 (95 % C.I., 1.641.72) as compared with their crude odds ratio o f 1.54 (1.51-1.58).


147 6.1.3

First births - low birth weight - dose effect Table 6.2(e) shows the results o f a more detailed analysis o f the effect o f

delayed childbearing on low birth weight. For this set o f analyses, the effect of maternal age groups o f 30 years and greater were analyzed for three age groups: 3034 years, 35-39 years and 40 years of age and older as compared with a reference group o f mothers 20-29 years o f age. For all four ethnic groups a ‘dose effect’ was present such that the odds o f very low, moderately low and hence overall low birth weight tended to increase with advancing age. Specifically, compared with women 20-29 years o f age, women who were 30-34 years, 35-39 years and 40 years of greater at the time o f their first births had progressively higher risks o f low birth weight. There were a few departures from this pattern, which occurred in the case o f the smallest subcells representing the more rare outcome o f very low birth weight in the smallest age category o f women 40 years and older in the smaller ethnic subgroups, i.e., Puerto Ricans and Mexican Americans, where the odds ratios were estimated with less precision. However, even in these subgroups by for most o f the estimates showed a pattern consistent with a dose effect and one that was fairly consistent with the magnitude o f age-related increases observed for the larger subgroups o f African Americans and non-Hispanic whites. Overall, the four ethnic groups showed fairly similar rates o f increase in the odds o f very low, moderately low and low birth weight with increasing maternal age beyond 30 years.


148 For African Americans, compared to women 20-29 years o f age, the odds o f low birthweight increased from 1.32 (95% C.I., 1.28-1.35) in women 30-34 years, to 1.64 (95% C.I., 1.57-1.72) in women 35-39 years, to 1.83 (95% C.I., 1.64-2.05) in women 40 years o f age and greater. Similarly, for non-Hispanic whites, compared to women 20-29 years o f age, the odds o f low birth weight increased from 1.21 (95% C.I., 1.19-1.22) in women 30-34 years to 1.58 (95% C.I., 1.55-1.61) in women 35-39 years to 1.97 (95% C.I., 1.87-2.05) in women 40 years o f age and greater. The effects o f advancing maternal age were greater for all four ethnic groups in the case o f very low birth weight infants, who are almost always very or moderately preterm and/or intrauterine growth retarded, as compared with the moderately low birth weight infonts, who are moderately preterm or term and/or intrauterine growth retarded. For African Americans, the odds ratio for women 30-34 years as compared with women 20-29 years o f age was 1.44 (95 % C.I., 1.36-1.52) for very low birth weight as compared with 1.28 (95 % C.I., 1.24-1.32) for moderately low birth weight. Similarly, the odds ratios for women 35-39 years and 40 years and above were 1.85 (95 % C.I., 1.70-2.02) and 2.03 (95 % C.I., 1.65-2.50) for very low birth weight as compared with 1.56 (95 % C.I., 1.49-1.65) and 1.74 (95 % C.I., 1.54-1.98), respectively for moderately low birth weight. Table 6.4(c) shows the results o f the logistic regression models o f low birth weight that adjusted the effect o f maternal ages 30-34, 35-39 and > 40 years, as compared with 20-29 years o f age at first birth for maternal education, marital status,


149 prenatal care and tobacco use. For all groups, the adjusted effects o f advanced maternal age tended to be greater by about 10-20 % than the unadjusted effects (Table 2c), suggesting that as reported in previous literature (Chapter 2) women who delay childbearing tend to be o f higher socioeconomic status and hence, other than their age profile, at lower risk for adverse birth outcomes such as low birth weight (Chapter 3). For African Americans, the adjusted odds ratios for women 30-34, 35-39 and 40 years and above for low birth weight were 1.46 (95 % C.I., 1.42-1.50), 1.82 (95 % C.I., 1.74-1.91), and 1.97 (95 % C.I., 1.76-2.20) as compared with the corresponding unadjusted effects o f 1.32 (95 % C.I., 1.28-1.35), 1.64 (95 % C.I., 1.57-1.72), and 1.83 (95 % C.I., 1.64-2.05). Similarly, for non-Hispanic whites, the adjusted odds ratios for women 30-34, 35-39 and 40 years and above for low birth weight were 1.39 (95 % C.I., 1.37-1.41), 1.82 (95 % C.I., 1.78-1.86), and 2.20 (95 % C.I., 2.09-2.32) as compared with the corresponding unadjusted effects o f 1.21 (95 % C.I., 1.19-1.22), 1.58 (95 % C.I., 1.55-1.61), and 1.97 (95 % C.I., 1.87-2.07).

6

.1.4

First births - low birth weight —effect o f prenatal care Table 6.4(e) shows the results o f the logistic regression models o f low birth

weight that adjusted the effect of maternal ages 30-34, 35-39 and > 40 years, as compared with 20-29 years o f age at first birth for maternal education, marital status, and tobacco use but not for prenatal care. The aim o f this set o f analyses was to assess the effect o f prenatal care. As discussed in Chapter 5, the effects o f prenatal care in an observational study such as the one undertaken here might be due to adverse or


150 favorable selection or any direct benefits o f prenatal care and hence the sum o f any selection or direct treatment effects. A comparison o f the estimates o f age effect obtained from the fully adjusted models that did control for prenatal care (Table 6.4(c)) with the estimates from the models that did not control for prenatal care (Table 6.4(e)) shows that the inclusion o f prenatal care as a right h an d side control variable had very little effect o n the estimates of age effects. For African Americans, the odds ratios for women 30-34, 3539 and 40 years and above for low birth weight in models that did not control for prenatal care (Table 6.4(e)), were 1.45 (95 % C.I., 1.41-1.49), 1.81 (95 % C.I., 1.731.90), and 1.98 (95 % C.I., 1.77-2.22) as compared with the adjusted odds ratios in models that did control for prenatal care (Table 6.4(c)), which were 1.46 (95 % C.I., 1.42-1.50), 1.82 (95 % C.I., 1.74-1.91), and 1.97 (95 % C.I., 1.76-2.20) for women 30-34, 35-39 and 40 years and above, respectively. Similarly, for non-Hispanic whites, the odds ratios for women 30-34, 35-39 and 40 years and above for low birth weight in models that did not control for prenatal care (Table 6.4(e)), were 1.36 (95 % C.I., 1.34-1.38), 1.79 (95 % C.I., 1.75-1.83), and 2.21 (95 % C.I., 2.10-2.32), as compared with the adjusted odds ratios in models that did control for prenatal care (Table 6.4(c)), which were 1.39 (95 % C.I., 1.37-1.41), 1.82 (95 % C.I., 1.78-1.86), and 2.20 (95 % C.I., 2.09-2.32) for women 30-34, 35-39 and 40 years and above, respectively.


151

6.1.5

First births - preterm delivery Table 6.5(a) shows the association o f delayed childbearing with preterm

delivery. For all four racial/ethnic groups, maternal age > 35 years at first birth was associated with statistically significant and substantial increases in the risks for preterm (< 37 weeks completed gestation) delivery. Compared with mothers 20-34 years o f age, mothers > 35 years were at

66

% increased risk o f very preterm (< 32

weeks) delivery [Table 6.5(a) - Mantel-Haenszel Odds Ratio: 1.66, 95 % C.I.: 1.591.72] and 39 % increased risk o f moderately preterm (32-37 weeks) delivery [MantelHaenszel Odds Ratio: 1.39, 95 % C.I.: 1.36-1.41]. Note that the effects o f advanced age on preterm delivery tended to be less than were the case for low birth weight (Tables 6.2(a) and 6.5(a)) suggesting that advanced age tends to both increase the proportion o f infants who are bom preterm and the proportion o f infants with intrauterine growth retardation. The effects o f maternal age > 3 5 years on preterm delivery were fairly similar in direction and in magnitude among the four racial/ethnic groups particularly after adjustment for other sociodemographic and reproductive history characteristics (Tables 6.5(a) and 6.7(a)). Puerto Ricans and particularly African Americans have two-three fold higher ‘base line’ risks o f preterm delivery, however. Therefore, equivalent odds ratios for the effect o f delayed childbearing on preterm delivery in African Americans and Puerto Ricans signify greater differences in terms o f risk


152 differences or absolute elevations in risk o f preterm delivery. This can be seen in Table 6.5(c) where risk differences in very preterm and moderately preterm delivery that were associated with advanced maternal age at first birth are shown for all four ethnic groups. The highest risk differences in very preterm and moderately preterm delivery, and hence the highest overall risk differences in preterm delivery, pertained to African Americans and Puerto Ricans and the lowest to non-Hispanic whites. Specifically, in the case o f preterm delivery, the risk differences for preterm delivery were 4.9 % (95 % C.I., 4.1-5.6) for African Americans, 5.2 % (95 % C.I., 2.3-8.1) for Puerto Ricans, 4.4 % (95 % C.I., 3.4-5.5) for Mexican Americans and 2.7 % (95 % C.I., 2.6-2.9) for non-Hispanic whites. Table 6 .6 (a) shows the preterm delivery estimates for the attributable fractions among the exposed and the population attributable risk percentages for the population as a whole related to advanced maternal age for births o f second or higher order. Estimates o f the attributable fractions among the exposed (AFe) suggest that about 30 % o f all preterm delivery and 39 % o f very preterm delivery in women with maternal age > 3 5 years were due to the ‘exposure’ o f advanced maternal age. On the other hand, the population attributable risk percentages (PAR%) estimated for the ‘population’ o f first births in the period 1989-1991 suggest that only 3 % o f all preterm delivery and 4 % o f all very preterm delivery first births were due to women with advanced maternal age. This is because despite recent trends, mothers > 3 5 years still comprised a small minority o f all mothers at first births (about 5% or less - see


153

Table 6.1(a)) in this period particularly for groups other than non-Hispanic whites. On the other hand, note that these analyses only included first births and do not capture the impact o f delayed childbearing on preterm delivery for births o f second or higher order (see Tables 6.6(b) and 6.6(c) and below for the discussion o f preterm delivery in births o f higher parity). Furthermore, with the continuing trends towards delayed childbearing, these estimates o f the population-level impact o f delayed childbearing on preterm delivery are probably low by a factor o f about 40-50 % for the current levels o f advanced maternal age (NCHS, 1997). Hence the estimates o f age effect suggest that with the current maternal age distribution, approximately 6 % o f preterm delivery and 8 % o f very preterm delivery first births are due to women with advanced maternal age. Table 6.7(a) presents results o f logistic regression models o f preterm delivery that adjusted the effect o f maternal age > 3 5 years at first birth for maternal education, marital status, prenatal care and tobacco use. For all groups, the adjusted effects o f advanced maternal age remained significant and o f the same or slightly increased magnitude after adjusting for other factors. Adjustment for other factors significantly increased the risk associated with advanced maternal age for non-Hispanic whites by about 7 % but for other ethnic groups the increases tended to be less and did not reach statistical significance due to the smaller number o f infants in ethnic groups other than non-Hispanic whites. The adjusted odds ratio for preterm delivery for nonHispanic whites was 1.50 (95 % C.I., 1.47-1.53) as compared with their crude odds


154 ratio o f 1.42 (1.40-1.45). Similarly, the adjusted odds ratio for very preterm delivery was 1.78 (95 % C.I., 1.70-1.86) as compared with their crude odds ratio o f 1.64 (1.57-1.72) and for moderately preterm delivery, the adjusted odds ratio was 1.45 (95 % C.I., 1.42-1.48) as compared with their crude odds ratio o f 1.38 (1.361.41).

6.1.6

First births - preterm delivery - dose effect Table 6.5(e) shows the results o f a more detailed analysis o f the effect o f

delayed childbearing on preterm delivery. For this set o f analyses, the effect o f maternal age groups o f 30 years and older were analyzed for three age groups: 30-34 years, 35-39 years and 40 years and greater as compared with a reference group o f mothers 20-29 years o f age. For all four ethnic groups a ‘dose effect’ was present such that the odds o f very preterm (< 32 weeks), moderately preterm (32-37) and hence overall preterm (< 37 weeks) delivery tended to increase with advancing age. Specifically, compared with women 20-29 years o f age, women who were 30-34 years, 35-39 years and 40 years o f age and older at the time o f their first births had progressively higher risks o f preterm delivery. The four ethnic groups showed fairly similar rates o f increase in the odds o f very preterm and moderately preterm delivery and hence overall preterm delivery with increasing maternal age beyond 30 years o f age. For African Americans, compared to women 20-29 years o f age, the odds o f preterm delivery increased from


155 1.18 (95% C.I., 1.15-1.21) in women 30-34 years, to 1.44 (95% C.I., 1.38-1.50) in women 35-39 years, to 1.50 (95% C.I., 1.35-1.67) in women 40 years o f age and older. S imilarly, for non-Hispanic whites, compared to women 20-29 years o f age, the odds o f preterm delivery increased from 1.13 (95% C.I., 1.12-1.14) in women 30-34 years to 1.41 (95% C.I., 1.39-1.44) in women 35-39 years to 1.58 (95% C.I., 1.511.66) in women 40 years o f age and older. The effects o f advancing maternal age were greater for all four ethnic groups in the case o f very preterm delivery, as compared with the moderately preterm delivery. For African Americans, the odds ratio for women 30-34 years as compared with women 20-29 years o f age was 1.29 (95 % C.I., 1.23-1.35) for very preterm delivery as compared with 1.14 (95 % C.I., 1.11-1.18) for moderately preterm delivery. Similarly, the odds ratios for women 35-39 years and 40 years and above were 1.63 (95 % C.I., 1.51-1.76) and 1.75 (95 % C.I., 1.45-2.11) for very preterm delivery as compared with 1.36 (95 % C.I., 1.29-1.43) and 1.40 (95 % C.I., 1.241.59), respectively for moderately preterm delivery. Table 6.7(c) shows the results o f the logistic regression models o f preterm delivery that adjusted the effect o f maternal ages 30-34, 35-39 and > 40 years, as compared with 20-29 years o f age at first birth for maternal education, marital status, prenatal care and tobacco use. For all groups, the adjusted effects o f advanced maternal age tended to be greater by about 10 % than the unadjusted effects (Table 6.5(c)), suggesting that as reported in previous literature (Chapter 2) women who


156

delay childbearing tend to be o f higher socioeconomic status and hence, other than their age profile, at lower risk for adverse birth outcomes such as preterm delivery (Chapter 3). For African Americans, the adjusted odds ratios for women 30-34, 35-39 and 40 years and above for preterm delivery were 1.31 (95 % C.I., 1.28-1.35), 1.61 (95 % C.I., 1.54-1.68), and 1.62 (95 % C.I., 1.45-1.81) as compared with the corresponding unadjusted effects o f 1.18 (95 % C.I., 1.15-1.21), 1.44 (95 % C.I., 1.38-1.50), and 1.50 (95 % C.I., 1.35-1.67). Similarly, for non-Hispanic whites, the adjusted odds ratios for women 30-34, 35-39 and 40 years and above for preterm delivery were 1.25 (95 % C.I., 1.23-1.26), 1.55 (95 % C.I., 1.53-1.58), and 1.69 (95 % C.I., 1.62-1.77) as compared with the corresponding unadjusted effects o f 1.13 (95 % C.I., 1.12-1.14), 1.41 (95 % C.I., 1.39-1.44), and 1.58 (95 % C.I., 1.51-1.66).

6.1.7

First births - preterm delivery - eflect o f prenatal care Table 6.7(e) shows the results o f the logistic regression models o f preterm

delivery that adjusted the effect o f maternal ages 30-34, 35-39 and > 40 years, as compared with 20-29 years o f age at first birth for maternal education, marital status, and tobacco use but not for prenatal care. The aim o f this set o f analyses was to assess the effect o f prenatal care. As noted above and discussed in more detail in Chapter 5, the effects o f prenatal care in an observational study such as the one undertaken here might be due to adverse or favorable selection or any direct benefits o f prenatal care


157 on reducing the risk o f preterm delivery and hence the sum o f any selection or direct treatment effects. A comparison o f the estimates o f age effect obtained from the fully adjusted models that did control for prenatal care (Table 6.7(c)) with the estimates from the models that did not control for prenatal care (Table 6.7(e)) shows that the inclusion o f prenatal care as a right hand side control variable had very little effect on the estimates o f age effects. For African Americans, the odds ratios for women 30-34, 3539 and 40 years and above for preterm delivery in models that did not control for prenatal care (Table 6.7(e)), were 1.30 (95 % C.I., 1.27-1.33), 1.59 (95 % C.I., 1.521.66), and 1.62 (95 % C.I., 1.46-1.81) as compared with the adjusted odds ratios in models that did control for prenatal care (Table 6.7(c)), which were 1.31 (95 % C.I., 1.28-1.35), 1.61 (95 % C.I., 1.54-1.68), and 1.62 (95 % C.I., 1.45-1.81) for women 30-34, 35-39 and 40 years and above, respectively. Similarly, for non-Hispanic whites, the odds ratios for women 30-34, 35-39 and 40 years and above for preterm delivery in models that did not control for prenatal care (Table 6.7(e)), were 1.22 (95 % C.I., 1.21-1.23), 1.53 (95 % C.I., 1.50-1.55), and 1.69 (95 % C.I., 1.62-1.77), as compared with the adjusted odds ratios in models that did control for prenatal care (Table 6.4(c)), which were 1.25 (95 % C.I., 1.23-1.26), 1.55 (95 % C.I., 1.53-1.58), and 1.69 (95 % C.I., 1.62-1.77) for women 30-34, 35-39 and 40 years and above, respectively.


158 6.2

Effects o f age on low birth weight and preterm delivery —Births o f second

or higher order 6.2.1

Births o f second or higher order - maternal age distribution Table 6.1(b) shows distribution o f maternal age for births o f second or higher

order for four racial/ethnic groups in the United States for the years 1989-1991. The results show that ethnic differences in the proportion o f infants bom to mothers with advanced maternal age are less for births o f second or higher order. For births o f second or higher order, the proportions o f infants bom to women 35 years o f age and greater were 13.4 % (95 % C.I., 13.3-13.4%) for non-Hispanic whites, 10.1% (95 % C.I., 10.0-10.2%) for African Americans, 10.1% (95 % C.I., 10.0-10.1%) for Mexican Americans and 7.5 % (95 % C.I., 7.2-7.7%) for Puerto Ricans in the period 19891991.

6.2.2

Births o f second or higher order - low birth weight Table 6.2(b) shows the association of advanced maternal age in births of

second or higher order with low birth weight. For all four racial/ethnic groups, maternal age >35 years for births o f second or higher order was associated with statistically significant increases in the odds of low (< 2.5 kg) birth weight (Table 6.2(b)). However, these increases tended to be less than were the case for first births (Table 6.2(a)). Compared with mothers 20-34 years o f age, mothers > 3 5 years were at 34 % increased risk o f very low (< 1.5 kg) birth weight [Table 6.2(b) - Mantel-


159 Haenszel Odds Ratio: 1.34, 95 % C.I.: 1.31-1.39] and 15 % increased risk o f moderately low (1.5-2.5 kg) birth weight [Mantel-Haenszel Odds Ratio: 1.15, 95 % C.I.: 1.14-1.17] for births o f second or higher order. As noted above, for first births, mothers > 3 5 years were at 82 % increased risk o f very low (< 1.5 kg) birth weight [Table 6.2(a) - Mantel-Haenszel Odds Ratio: 1.82, 95 % C.I.: 1.75-1.90] and 55 % increased risk o f moderately low (1.5-2.5 kg) birth weight [Mantel-Haenszel Odds Ratio: 1.55, 95 % C.I.: 1.52-1.58]. The effects o f maternal age > 3 5 years tended to be slightly more for Mexican Americans and Puerto Ricans as compared w ith non-Hispanic whites and African Americans. (Tables 6.2(b) and 6.4(b)). In addition, Puerto Ricans and African Americans have greater ‘base line’ risks o f low birth weight. Therefore, equivalent odds ratios for the effect o f delayed childbearing in African Americans and Puerto Ricans signify greater differences in terms o f risk differences or absolute elevations in risk. The net effect o f these factors are seen in Table 6.2(d) where risk differences in very low, moderately low and low birth weight that were associated with advanced maternal age for births o f second or higher order are shown for all four ethnic groups. The highest risk differences in very low and moderately low birth weight, and hence the highest overall risk differences in low birth weight, pertained to Puerto Ricans, African Americans and Mexican Americans and the lowest to non-Hispanic whites. Specifically, in the case o f low birth weight, the risk differences for low birth weight were 1.4 % (95 % C.I., 1.2-1.6) for African Americans, 1.7 % (95 % C.I., 1.1-


160 2.4) for Puerto Ricans, 1.4 % (95 % C.I., 1.3-1.6) for Mexican Americans and 0.5 % (95 % C.I., 0.4-0.5) for non-Hispanic whites. Note that for all four ethnic groups the risk differences associated with advanced age were four to five fold greater in first births as compared with births o f second or higher order (Tables 6.2(c) and 6.2(d)). For African Americans, the risk difference for low birth weight that was associated with maternal age > 35 years was 1.4 % (95 % C.I., 1.2-1.6) in births o f second or higher order as compared with 5.3 % (95 % C.I., 4.7-6.0) in first births. Similarly, for non-Hispanic whites, the risk difference for low birth weight that was associated with maternal age > 35 years was 0.5 % (95 % C.I., 0.4-0.5) in births o f second or higher order as compared with 2.6 % (95 % C.I., 2.4-2.7) in first births. Table 6.3(b) shows the low birth weight estimates for the attributable fractions among the exposed and the population attributable risk percentages for the population as a whole related to advanced maternal age for births o f second or higher order. Estimates of the attributable fractions among the exposed (AFe) suggest that about 14-30 % o f all low birth weight and 18-45 % of very low birth weight infants in women with maternal age > 3 5 years are due to the ‘exposure’ o f advanced maternal age for births o f second or higher order. On the other hand, the population attributable risk percentages (PAR%) estimated for the ‘population’ o f births o f second or higher order in the period 1989-1991 suggest that only 2-4 % o f all low birth weight and 57% o f all very low birth weight in births o f second or higher order were due to women with advanced maternal age. This is because despite recent trends in delayed


161 childbearing, mothers > 3 5 years still comprised a small minority o f all mothers for births o f second or higher order (about 10-13% —see Table 6.1(b)) in this period. This may in part be due to the decrease in higher, particularly fourth or greater, parity births in the United States (Chapter 2). Table 6.3(c) shows the low birth weight estimates for the population attributable risk percentages for the population as a whole related to advanced maternal age for all births. The population attributable risk percentages estimated for all births in the period 1989-1991 suggest that 5-6 % o f all low birth weight and 510% o f all very low birth weight births were due to women with advanced maternal age. As noted above, this relatively modest effect o f maternal age is in part due to the fact that despite recent trends in delayed childbearing, mothers >35 years still comprised a small minority o f all mothers in this period. With the continued trends towards delayed childbearing in the 1990s (Chapter 2), the estimates o f population attributable risk percentage for the years 1989-1991 suggest that with the current maternal age distribution (NCHS, 1997), approximately 10% of all low birth weight and 15% o f all very low birth births might be due to women with advanced maternal age. Table 6.4(b) presents results o f logistic regression models o f low birth weight that adjusted the effect o f maternal age > 35 years for births o f second or higher order for maternal education, marital status, prenatal care and tobacco use. For all groups, the adjusted effects o f advanced maternal age remained significant and o f the same or


162 increased magnitude after adjusting for other factors. Adjustment for other factors most s ig n i f i c a n t l y increased the risk associated with advanced maternal age for nonHispanic whites and to a lesser extent for African Americans but not for others. The adjusted odds ratio for low birth weight for non-Hispanic whites was 1.52 (95 % C.I.. 1.46-1.59) as compared with their crude odds ratio o f 1.35 (1.30-1.40). For African Americans, the adjusted odds ratio for low birth weight was 1.32 (95 % C.I., 1.241.40) as compared with their crude odds ratio o f 1.22 (1.15-1.29).

6.2.3

Births of second or higher order - low birth weight - dose effect Table 6.2(f) shows the results o f a more detailed analysis o f the effect o f

delayed childbearing on low birth weight for births o f second or higher order. For this set o f analyses, the efiect o f maternal age groups o f 30 years and greater were analyzed for three age groups: 30-34 years, 35-39 years and 40 years o f age and older as compared with a reference group o f mothers 20-29 years o f age. For all four ethnic groups a ‘dose effect’ was present such that the odds o f very low, moderately low and hence overall low birth weight tended to increase with advancing age. Specifically, compared with women 20-29 years o f age, women who were 30-34 years, 35-39 years and 40 years o f greater at the time o f their second or higher order birth had progressively higher risks o f low birth weight. There were a few departures from this pattern, which occurred in the case o f the unadjusted odds ratios for non-Hispanic white women 30-34 years o f age (Table 6.2(f)). However, in the case o f adjusted


163

estimates, all four groups showed progressively higher odds o f low birth weight as maternal age increased beyond 30 years. For African Americans, compared to women 20-29 years o f age, the unadjusted odds o f low birth weight increased from 1.13 (95% C.I., 1.11-1.14) in women 30-34 years, to 1.17 (95% C.I., 1.14-1.19) in women 35-39 years, to 1.23 (95% C.I., 1.17-1.29) in women 40 years o f age and greater. Similarly, for nonHispanic whites, compared to women 20-29 years o f age, the odds o f low birth weight increased from 0.84 (95% C.I., 0.83-0.85) in women 30-34 years to 1.02 (95% C.I., 1.01-1.04) in women 35-39 years to 1.36 (95% C.I., 1.32-1.41) in women 40 years o f age and older. The effects o f advancing maternal age on very low, moderately low and hence the overall odds o f low birth weight were about three to four fold greater in the case o f first births as compared with births o f second or higher order for all four ethnic groups. (Tables 6.2(e) and 6.2(f)). For African Americans, compared to women 20-29 years o f age, the odds o f low birth weight in first births were 1.32 (95% C.I., 1.281.35), 1.64 (95% C.I., 1.57-1.72), and 1.83 (95% C.I., 1.64-2.05) as compared with 1.13 (95% C.I., 1.11-1.14), 1.17 (95% C.I., 1.14-1.19), and 1.23 (95% C.I., 1.171.29) for births o f second or higher order in women 30-34 years, 35-39 years, and 40 years o f age and older respectively. Similarly, for non-Hispanic whites, compared to women 20-29 years o f age, the odds o f low birth weight in first births were 1.21 (95% C.I., 1.19-1.22), 1.58 (95% C.I., 1.55-1.61), and 1.97 (95% C.I., 1.87-2.07) as


164 compared with 0.84 (95% C.I., 0.83-0.85), 1.02 (95% C.I., 1.01-1.04), and 1.36 (95% C.I., 1.32-1.41) for births o f second or higher order in women 30-34 years, 3539 years, and 40 years o f age and older respectively. Table 6.4(d) shows the results o f the logistic regression models of low birth weight that adjusted the effect o f maternal ages 30-34, 35-39 and > 40 years, as compared with 20-29 years o f age for births o f second or higher order for maternal education, marital status, prenatal care and tobacco use. For all groups, the adjusted effects o f advanced maternal age tended to be greater by about 10-20 % than the unadjusted effects (Table 6.2(d)), suggesting that as reported in previous literature (Chapter 2) women who delay childbearing tend to be o f higher socioeconomic status and hence, other than their age profile, at lower risk for adverse birth outcomes such as low birth weight (Chapter 3). For African Americans, the adjusted odds ratios for women 30-34, 35-39 and 40 years and above for low birth weight were 1.28 (95 % C.I., 1.26-1.30), 1.39 (95 % C.I., 1.35-1.42), and 1.48 (95 % C.I., 1.41-1.56) as compared with the corresponding unadjusted effects o f 1.13 (95 % C .I„ 1.11-1.14), 1.17 (95 % C.I., 1.14-1.19), and 1.23 (95 % C.I., 1.17-1.29). Similarly, for non-Hispanic whites, the adjusted odds ratios for women 30-34, 35-39 and 40 years and above for low birth weight were 1.12 (95 % C.I., 1.11-1.14), 1.43 (95 % C.I., 1.40-1.45), and 1.82 (95 % C.I., 1.76-1.89) as compared with the corresponding unadjusted effects o f 0.84 (95 % C.I., 0.83-0.85), 1.02 (95 % C.I., 1.01-1.04), and 1.36 (95 % C.I., 1.32-1.41).


165 6.2.4

Births o f second or higher order - low birth weight - efiect o f prenatal care Table 6.4(f) shows the results o f the logistic regression models o f low birth

weight that adjusted the effect o f maternal ages 30-34, 35-39 and > 40 years, as compared with 20-29 years o f age for births o f second or higher order for maternal education, marital status, and tobacco use but not for prenatal care. The aim o f this set o f analyses was to assess the effect o f prenatal care. As noted above and more fully discussed in Chapter 5, the effects o f prenatal care in an observational study such as the one undertaken here might be due to adverse or favorable selection or any direct benefits o f prenatal care and hence the sum o f any selection or direct treatment effects. A comparison o f the estimates o f age effect obtained from the fully adjusted models that did control for prenatal care (Table 6.4(d)) with the estimates from the models that did not control for prenatal care (Table 6.4(f)) shows that, as was the case for first births, the inclusion o f prenatal care as a right hand side control variable had very little effect on the estimates o f age effects for births o f second or higher order. For African Americans, the odds ratios for women 30-34, 35-39 and 40 years and above for low birth weight in models that did not control for prenatal care (Table 6.4(f)), were 1.25 (95 % C.I., 1.23-1.27), 1.35 (95 % C.I., 1.32-1.38), and 1.45 (95 % C.I., 1.38-1.52) as compared with the adjusted odds ratios in models that did control for prenatal care (Table 6.4(d)), which were 1.28 (95 % C.I., 1.26-1.30), 1.39 (95 % C.I., 1.35-1.42), and 1.48 (95 % C.I., 1.41-1.56) for women 30-34, 35-39 and 40


166 years and above, respectively. Similarly, for non-Hispanic whites, the odds ratios for women 30-34, 35-39 and 40 years and above for low birth weight in models that did not control for prenatal care (Table 6.4(f)), were 1.10 (95 % C.I., 1.08-1.11), 1.40 (95 % C.I., 1.37-1.42), and 1.81 (95 % C.I., 1.75-1.87), as compared with the adjusted odds ratios in models that did control for prenatal care (Table 6.4(c)), which were 1.12 (95 % C.I., 1.11-1.14), 1.43 (95 % C.I., 1.40-1.45), and 1.82 (95 % C.I., 1.761.89) for women 30-34, 35-39 and 40 years and above, respectively.

6.2.5

Births o f second or higher order - preterm delivery Table 6.5(b) shows the association o f delayed childbearing with preterm

delivery. Overall, the risks o f preterm delivery associated with advanced maternal age were much smaller than were the case for first births in all four ethnic groups. In particular, for African Americans, there was very little evidence o f any substantial risk for preterm delivery associated with advanced age for births o f second or higher order. For African Americans, compared with mothers 20-34 years o f age, mothers > 35 years only had a 4 % increased risk o f very preterm (< 32 weeks) delivery [Odds Ratio: 1.04, 95 % C.I.: 1.01-1.08] and moderately preterm (32-37 weeks) delivery [Odds Ratio: 1.04, 95 % C.I.: 1.02-1.06]. For Non-Hispanic whites, the effect o f advanced age on preterm delivery for births o f second or higher order was again modest with about an 11 % increased risk associated with maternal age > 35 years [Odds Ratio: 1.11, 95 % C.I.: 1.09-1.12], The greatest increase in risk associated with


167 advanced age for preterm delivery in births o f second or higher order was in the case o f Mexican Americans with about a 27 % increased risk in preterm delivery associated with maternal age > 35 years [Odds Ratio: 1.27, 95 % C.I.: 1.24-1.30]. Because o f the smaller number o f births to Puerto Rican mothers, the odds ratio for preterm delivery was estimated with less precision and did not reach statistical significance for Puerto Ricans [Odds Ratio: 1.10, 95 % C.I.: 0.55-2.18]. But again the point estimate suggests a modest effect if any o f advanced age on the risk o f preterm delivery for births o f second or higher order. As in first births, the effects o f advanced age on preterm delivery for births o f second or higher order tended to be less than were the case for low birth weight (Tables 6.2(b) and 6.5(b)) suggesting that advanced age tends to both increase the proportion o f infants who are bom preterm and the proportion o f infants with intrauterine growth retardation. Table 6.5(d) shows the risk differences in very preterm and moderately preterm delivery that were associated with advanced maternal age for births o f second or higher order. The highest risk differences in very preterm and moderately preterm delivery, and hence the highest overall risk differences in preterm delivery, pertained to Mexican Americans and the lowest to African Americans and non-Hispanic whites. Specifically, the risk differences for preterm delivery were 2.2 % (95 % C.I., 2.0-2.5) for Mexican Americans, 0.7 % (95 % C.I., 0.6-0.8) for non-Hispanic whites and 0.6 % (95 % C.I., 0.4-0.9) for African Americans. Again the relatively small number o f


168 infants bom to Puerto Rican mothers resulted in wide confidence intervals for the estimates o f risk difference for Puerto Rican [1.1 % (95 % C.I., -6.8-8.9)]. Table 6.6(b) shows the preterm delivery estimates for the attributable fractions am ong

the exposed and the population attributable risk percentages for the population

as a whole related to advanced maternal age for births o f second or higher order. Estimates o f the attributable fractions among the exposed (AFe) suggest that for births o f second or higher order about 10 % o f preterm delivery for non-Hispanic white, and 21 % o f preterm delivery for Mexican American women with maternal age > 3 5 years were due to the ‘exposure’ o f advanced maternal age. On the other hand, as noted above and reflected in the very small values o f AFe (Table 6.6(b)), there was little evidence o f any substantial risk associated with advanced age for African Americans or Puerto Ricans. For all four groups, therefore, advanced age was associated with at most a very modest increase in the risk o f preterm delivery for births o f second or higher order. The population attributable risk percentages (PAR%) estimated for the ‘population’ o f births o f second or higher order in the period 1989-1991 suggest that only 1 % o f all preterm delivery and 2 % o f all very preterm delivery in births o f second or higher order were due to women with advanced maternal age. This is because o f the modest effects o f advanced age on the risk o f preterm delivery for births o f second or higher order and the fact that despite recent trends, mothers > 3 5


169 years still comprised a small minority o f all mothers at births o f second or higher order (10-15% - see Table 6.1(b)). Table 6.6(c) shows the preterm delivery estimates for the population attributable risk percentages for the population as a whole related to advanced maternal age for all births. The population attributable risk percentages estimated for all births in the period 1989-1991 suggest that 2-4 % o f all preterm delivery and 3-6% o f all very preterm delivery births were due to women with advanced maternal age in the four ethnic groups in this study. As noted above, this relatively modest effect o f maternal age is in part due to the fact that despite recent trends in delayed childbearing, mothers > 3 5 years still comprised a small minority o f all mothers in this period. With the continued trends towards delayed childbearing in the 1990s (Chapter 2), the estimates o f population attributable risk percentage for the years 1989-1991 suggest that with the current maternal age distribution (NCHS, 1997), approximately 8% o f all preterm delivery and 12% o f all very low birth births might be due to women with advanced maternal age. Table 6.7(b) presents results o f logistic regression models o f preterm delivery that adjusted the effect o f maternal age > 35 years for births o f second or higher order for maternal education, marital status, prenatal care and tobacco use. For all groups, the effects o f advanced maternal age increased after adjusting for other factors (Tables 6.5(b) and 6.7(b)). The adjusted odds ratio for preterm delivery for nonHispanic whites was 1.26 (95 % C.I., 1.25-1.28) as compared with their crude odds


170 ratio o f 1.11 (1.10-1.12). For African Americans, the adjusted odds ratio for preterm delivery was 1.16 (95 % C.I., 1.14-1.19) as compared with their crude odds ratio o f 1.04 (1.03-1.06).

6.2.6 Births o f second or higher order - preterm delivery - dose effect Table 6.5(f) shows the results o f a more detailed analysis o f the effect o f delayed childbearing on preterm delivery for births o f second or higher order. For this set o f analyses, the effect o f maternal age groups o f 30 years and greater were analyzed for three age groups: 30-34 years, 35-39 years and 40 years o f age and older as compared with a reference group o f mothers 20-29 years o f age. For all four ethnic groups a ‘dose effect’ was present such that the odds of very preterm, moderately preterm and hence overall preterm delivery tended to increase with advancing age. Specifically, compared with women 20-29 years o f age, women who were 30-34 years, 35-39 years and 40 years o f age and older at the time o f their second or higher order birth had progressively higher risks o f preterm delivery. There were a few departures from this pattern, which occurred in the case o f the (unadjusted) odds ratios for non-Hispanic white or African American women 30-34 years o f age (Table 6.5(f)). However, in the case o f adjusted estimates discussed below, all four groups showed progressively higher odds o f preterm delivery as maternal age increased beyond 30 years (Table 6.7(d)).


171 For African Americans, compared to women 20-29 years o f age, the unadjusted odds o f preterm delivery increased from 0.98 (95% C.I., 0.97-0.99) in women 30-34 years to 1.02 (95% C.I., 1.00-1.04) in women 35-39 years, to 1.15 (95% C.I., 1.10-1.20) in women 40 years o f age and older. Similarly, for nonHispanic whites, compared to women 20-29 years o f age, the odds o f preterm delivery increased from 0.86 (95% C.I., 0.85-0.87) in women 30-34 years to 1.02 (95% C.I., 1.01-1.03) in women 35-39 years to 1.28 (95% C.I., 1.24-1.31) in women 40 years of age and older. The effects o f advancing maternal age on very preterm, moderately preterm and hence the overall odds o f preterm delivery were substantially greater in the case o f first births as compared with births o f second or higher order for all four ethnic groups. (Tables 6.5(e) and 6.5(f)). For African Americans, compared to women 20-29 years o f age, the odds o f preterm delivery in first births were 1.18 (95% C.I., 1.151.21), 1.44 (95% C.I., 1.38-1.50), and 1.50 (95% C.I., 1.35-1.67) as compared with 0.98 (95% C.I., 0.97-0.99), 1.02 (95% C.I., 1.00-1.04), and 1.15 (95% C.I., 1.10-1.20) for births o f second or higher order in women 30-34 years, 35-39 years, and 40 years o f age and older respectively. Similarly, for non-Hispanic whites, compared to women 20-29 years o f age, the odds o f preterm delivery in first births were 1.13 (95% C.I., 1.12-1.14), 1.41 (95% C.I., 1.39-1.44), and 1.58 (95% C.I., 1.51-1.66) as compared with 0.86 (95% C.I., 0.85-0.87), 1.02 (95% C.I., 1.01-1.03), and 1.28 (95%


172 C.I., 1.24-1.31) for births o f second or higher order in women 30-34 years, 35-39 years, and 40 years o f age and older respectively. Table 6.7(d) shows the results o f the logistic regression models o f preterm delivery that adjusted the effect o f maternal ages 30-34, 35-39 and > 40 years, as compared with 20-29 years o f age for births o f second or higher order for maternal education, marital status, prenatal care and tobacco use. For all groups, the adjusted effects o f advanced maternal age increased as compared with the unadjusted effects (Table 6.5(f)), suggesting that as reported in previous literature (Chapter 2) women who delay childbearing tend to be o f higher socioeconomic status and hence, other than their age profile, at lower risk for adverse birth outcomes such as preterm delivery (Chapter 3). For African Americans, the adjusted odds ratios for women 30-34, 35-39 and 40 years and above for preterm delivery were 1.09 (95 % C.I., 1.08-1.11), 1.17 (95 % C.I., 1.15-1.19), and 1.30 (95 % C.I., 1.25-1.36) as compared with the corresponding unadjusted effects o f 0.98 (95 % C.I., 0.97-0.99), 1.02 (95 % C.I., 1.00-1.04), and 1.15 (95 % C.I., 1.10-1.20). Similarly, for non-Hispanic whites, the adjusted odds ratios for women 30-34, 35-39 and 40 years and above for preterm delivery were 1.03 (95 % C.I., 1.02-1.04), 1.24 (95 % C.I., 1.22-1.25), and 1.49 (95 % C.I., 1.45-1.53) as compared with the corresponding unadjusted effects o f 0.86 (95 % C.I., 0.85-0.87), 1.02 (95 % C.I., 1.01-1.03), and 1.28 (95 % C.I., 1.24-1.31).


173 6.2.7 Births o f second or higher order - preterm delivery - effect o f prenatal care Table 6.7(f) shows the results o f the logistic regression models o f preterm delivery that adjusted the effect o f maternal ages 30-34, 35-39 and > 40 years, as compared with 20-29 years o f age for births o f second or higher order for maternal education, marital status, and tobacco use but not for prenatal care. The aim o f this set o f analyses was to assess the effect o f prenatal care. As noted above and more fully discussed in Chapter 5, the effects o f prenatal care in an observational study such as the one undertaken here might be due to adverse or favorable selection or any direct benefits o f prenatal care and hence the sum o f any selection or direct treatment effects. A comparison o f the estimates o f age effect obtained from the fully adjusted models that did control for prenatal care (Table 6.7(d)) with the estimates from the models that did not control for prenatal care (Table 6.7(f)) shows that, as was the case for first births, the inclusion o f prenatal care as a right hand side control variable had very little effect on the estimates o f age effects for births of second or higher order. For African Americans, the odds ratios for women 30-34, 35-39 and 40 years and above for preterm delivery in models that did not control for prenatal care (Table 6.7(f)), were 1.07 (95 % C.I., 1.06-1.09), 1.14 (95 % C.I., 1.12-1.16), and 1.28 (95 % C.I., 1.23-1.34) as compared with the adjusted odds ratios in models that did control for prenatal care (Table 6.7(d)), which were 1.09 (95 % C.I., 1.08-1.11), 1.17 (95 % C.I., 1.15-1.19), and 1.30 (95 % C.I., 1.25-1.36) for women 30-34, 35-39 and 40


174 years and above, respectively. Similarly, for non-Hispanic whites, the odds ratios for women 30-34, 35-39 and 40 years and above for preterm delivery in models that did not control for prenatal care (Table 6.7(f)), were 1.01 (95 % C.I., 1.00-1.01), 1.21 (95 % C.I., 1.20-1.23), and 1.48 (95 % C.I., 1.44-1.52), as compared with the adjusted odds ratios in models that did control for prenatal care (Table 6.7(d)), which were 1.03 (95 % C.I., 1.02-1.04), 1.24 (95 % C.I., 1.22-1.25), and 1.49 (95 % C.I., 1.451.53) for women 30-34, 35-39 and 40 years and above, respectively.

6.3

Effects of age on Down syndrome: Effect modification by socioeconomic factors

6.3.1

Down syndrome - interaction effects between ethnicity and age Table 6.8 relates results o f the analysis for comparing the effects o f delayed

childbearing on the birth prevalence o f Down syndrome among African Americans, Mexican Americans, and non-Hispanic whites in the United States for the years 19891991. The results showed the well-known, exponential rise o f the risk for Down syndrome with advancing maternal age for all three groups (Figure 6.1). However, there were significant interactions between the maternal age and race/ethnicity effects (Likelihood ratio x2, p < 0.0001). Specifically, the age-related increases in the odds o f Down syndrome were significantly greater for Mexican American and African American mothers as compared with non-Hispanic whites. Compared to the reference group o f women 25-29 years o f age, the odds ratios for Mexican American women


175 were 5.2 (95 % C.I., 3.8 - 6.9) for the 35-39 year age group, 19.4 (95 % C.I., 14.1-26.7) for the 40-44 year age group and 52.3 (95 % C.I., 25.2-108.4) for the 4549 year age group. The corresponding odds ratios were 3.9 (95 % C.I., 2.9-5.1), 11.6 (95 % C.I., 8.2 - 16.5) and 40.4 (95 % C.I., 16.4 - 99.5) for African Americans and 2.7 (95 % C.I., 2.5-3.0), 8.5 (95 % C.I., 7.5 - 9.7) and 22.5 (95 % C.I., 14.8 - 34.1) for non-Hispanic whites. Table 6.9 shows results o f the analysis to compare the effects and the population-level impact o f advanced maternal age (>35 years) o n the risk for Down syndrome among the three ethnic groups. The odds ratio and the population attributable risk o f Down syndrome due to maternal age > 3 5 years were highest for Mexican Americans [Odds ratio: 6.5, 95 % C.I., 5.4 - 7.7); Population attributable risk = 28 %], intermediate for African Americans [Odds ratio: 5.2, 95 % C.I., 4.36.3); Population attributable risk = 19 %] and lowest for non-Hispanic whites [Odds ratio: 3.2, 95 % C.I., 3.0-3.5; Population attributable risk = 17 % ]. Similarly, the odds ratio and the population attributable risk o f Down syndrome due to maternal age > 40 years were highest for Mexican Americans [Odds ratio: 14.5, 95 % C.I., 11.3 - 18.5); Population attributable risk = 14 %], intermediate for African Americans [Odds ratio: 11.3, 95 % C.I., 8.4-15.1); Population attributable risk = 7 %] and lowest for nonHispanic whites [Odds ratio: 7.3, 95 % C.I., 6.5-8.1; Population attributable risk = 7 %]. Advanced maternal age had a greater impact (higher population attributable risk) on birth prevalence o f Down syndrome for Mexican American and African American


176 mothers even though, as noted above and shown in Tables 6.1(a) and 6.1(b), a significantly lower proportion o f Mexican American and African American mothers were > 35 years o f age. Mantel-Haenszel analysis o f age-specific odds o f amniocentesis use for the three groups shows that African American and particularly Mexican American mothers were substantially less likely to use amniocentesis (Table 6.10). For all age groups but particularly for 35-39 and > 40 years age groups, African American and Mexican American mothers had amniocentesis rates that were about 2/3 to 1/3 o f the rates for their non-Hispanic white counterparts [p < 0.001]. The lowest odds o f amniocentesis use were for Mexican American mothers 35-39 (Odds Ratio: 0.30, 95% Cl, 0.29 - 0.31) and > 40 years o f age (Odds Ratio: 0.26, 95% Cl, 0.24 - 0.28).

6.3.2

Down syndrome - interaction effects between ethnicity and education Table 6.11 shows the results o f the analysis for estimating the effects o f

maternal education on the age-related increases in the birth prevalence o f Down syndrome for African Americans and non-Hispanic whites in the United States for the years 1989-1991. The results showed the well-known, exponential rise o f the risk for Down syndrome with advancing maternal age for all ethnicity and education groups (Figures 6.2 and 6.3). However, for non-Hispanic whites, the age-related increases in the odds o f Down syndrome were significantly greater for women with less than 12 years of education as compared with women with 12 years of education or more

R e p ro du ced with permission o f the copyright owner. Further reproduction prohibited without permission.

177 (Figure 6.2, Likelihood ratio test, p < 0.0001). For non-Hispanic women with less than 12 years o f education, compared to the reference group o f women 25-29 years o f age, the odds ratio for the 30-34 year age group was 1.6 (95 % C.I., 1.4-1.8), for the 35-39 year age group, 3.7 (95 % C.I., 3.2-4.3), for the 40-44 year age group, 11.9 (95 % C.I., 9.8-14.5), and for the 45-49 year age group, 34.6 (95 % C.I., 20.3-59.1). The corresponding odds ratios in non-Hispanic white women with 12 years o f education or more were 1.5 (95 % C.I., 1.3-1.6), 2.4 (95 % C.I., 2.1-2.7), 7.0 (95 % C.I., 5.98.3), and 12.6 (95 % C.I., 6.0-26.6). On the other hand, for African Americans, the age-related increases in the odds o f Down syndrome did not appear to be significantly different for women with less than 12 years o f education as compared with women with 12 years o f education or more (Figure 6.3, Likelihood ratio test, p=0.68). For African American women with less than 12 years o f education, compared to the reference group of women 2529 years of age, the odds ratios for the 30-34 year age group was 1.5 (95 % C.I., 1.02.2), for the 35-39 year age group, 4.8 (95 % C.I., 3.3-7.0), for the 40-44 year age group, 12.9 (95 % C.I., 7.9-21.0), and for the 45-49 year age group, 42.4 (95 % C.I., 13.2-136.0). The corresponding odds ratios in African American women with 12 years o f education or more were 1.0 (95 % C.I., 0.7-1.6), 3.0 (95 % C.I., 2.4-4.5), 10.9 (95 % C.I., 6.6-18.0), and 44.6 (95 % C.I., 10.8-185.1). Mantel-Haenszel analysis o f age-specific odds o f amniocentesis use by ethnicity and education showed that in both African Americans and non-Hispanic


178 whites, women with less than 12 years o f education were substantially less likely to use amniocentesis (Table 6.12). In particular, compared to African A m erican women with 12 or more years o f education, the odds o f amniocentesis for Âfrican American women with less than 12 years o f education who were 35-39 a n d 40 years o f age and older were 0.44 (96 % C.I., 0.42-0.47) and 0.40 (96 % C.I., 0.36-0.44) respectively. Similarly, for non-Hispanic whites, the odds o f amniocentesis for nonHispanic white women with less than 12 years of education who were 3 5 -3 9 and 40 years o f age and older were 0.62 (96 % C.I., 0.62-0.63) and 0.60 (96 % C M., 0.580.62) respectively. These results suggest that the difference noted above beetween the effects o f education on age-related increases in the birth prevalence o f D o w n syndrome in African Americans as compared with non-Hispanic whites maay not be due to differences in the effects o f education on the frequency o f am niocentesis use in African Americans as compared with non-Hispanic whites. Instead Africam Americans and non-Hispanic whites in the same education group might diflffer in the frequency of termination once the diagnosis is made after amniocentesis is administered.

6.4

Effects o f age on cause-specific infant mortality Tables 6.13(a) and 6.13(b) show the results o f the set o f analyses tro examine

the effect o f delayed childbearing on the risk of infant death due to congenital anomalies, complications o f placenta, cord and membranes, sudden infant death


179 syndrome, intentional and unintentional injuries and all-cause mortality. Overall, advanced maternal age was associated w ith a slightly lower risk o f infant death [Hazard Ratio for all cause-mortality (M odel I), 0.93, p=0.04]. However, this survival advantage was entirely due to the effects o f sociodemographic and reproductive history confounders [Adjusted Hazard Ratio (Model II), 1.04, p = 0.2327]. Conditional on birth weight, however, there was about a 12 % lower risk o f infant mortality for older mothers [Birth weight-Adjusted Hazard Ratio (Model III), 0.88, p = 0.0004]. The effect o f advanced maternal age differed substantially across the various causes o f death. Infants bom to older mothers had lower risks o f death due to intentional and unintentional injuries and sudden infant death syndrome. The risk for unintentional injuries (accidents) for infants bom to mothers > 35 years was about 1/3 o f the risk for infants bom to mothers 20 —34 years o f age [Adjusted Hazard Ratio (Model III), 0.34, p = 0.001]. Similarly, the risk for intentional injuries (homicides) was also less than a third o f the risk for infants bom to mothers 20-34 years o f age [Hazard Ratio (Model I), 0.28, p = 0.0286]. After adjustment for the control variables in the model, the hazard ratio changed little but was no longer statistically significant at the 0.05 level [Adjusted Hazard Ratio (Model III), 0.39, p = 0.1063]; this was most likely due to the small number o f intentional injuries in the population as well as the underreporting o f intentional infant deaths which reduces the number o f events in the data and hence the precision o f the estimates. The risk for Sudden Infant Death


180 Syndrome was also substantially lower for infants bom to older mothers [Adjusted Hazard Ratio (Model HI), 0.55, p = 0.0001]. O n the other hand, the risk o f infant death due to congenital anomalies [Adjusted Hazard Ratio (Model II), 1.25, p = 0.0003] and particularly due to complications o f placenta, cord and membranes [Hazard Ratio, 1.68, p = 0.01] were higher for older mothers.


181

CHAPTER 7 - CONCLUSIONS Childbearing and child rearing are core family related behaviors, both in social science theory and in widely shared perceptions of family life l. They are also important for a wide range o f domestic policy issues such as employment, social security, education, parental benefits and poverty. While the United States total fertility rate has remained relatively unchanged over the past three decades, important changes have taken place in the United States fertility. This study focuses on a set o f issues related to the impact and implications o f one of the salient features o f U.S. fertility in the recent years, namely the persistent trend towards delayed childbearing in the past three decades. Specifically, the study assesses the individual effects and population level impact o f delayed childbearing on the risks for a set o f major adverse birth outcomes in the United States. In this final chapter after presenting a summary o f the major findings, results o f the study are discussed in the context o f their place in the literature that deals with the impact o f advancing maternal age on the risk o f adverse outcomes. Study results are interpreted using the biological and epidemiological evidence from previous literature. Contributions to the literature and implications o f the study findings for individual as well as societal concerns regarding the effects o f delayed childbearing on an important set o f adverse birth and infant outcomes are discussed.


The results o f the study are summariz e d and discussed in the same order as was reported in Chapter 6. The effects o f advancing maternal age on low birth weight and preterm delivery are presented first, which are then followed by a discussion o f the results for the effects o f age on Down syndrome and finally the effects o f age on cause-specific infant mortality. As is noted below, the study results are particularly important in terms o f their implications in the case o f Down syndrome, as well as other birth defects, most notably neural tube defects, which can be detected in utero. The results o f the study on the differences in the use o f prenatal diagnosis (amniocentesis) by ethnicity and education and their association with agerelated differences in Down syndrome fell within a broader literature that is concerned with socioeconomic determinants and consequences of technology access and use in health care. The results demonstrating substantial differentials by education and ethnicity in the age-related risk o f Down syndrome is presumably applicable to other birth defects that can be detected prenatally; which in turn implies a substantial socioeconomic disparity in the risk o f birth defects that is related to differential use o f prenatal diagnosis. Socioeconomic differences in prenatal diagnosis present unique features 2, that include but go beyond the issues related to access, cost, knowledge or effective use o f technology. Currently the only effective means o f prevention for birth defects that can be detected prenatally is “selective termination”; namely the abortion o f the affected fetus, which is related to a complex set o f personal and societal issues.


183

In discussing the results, perhaps it is most important to note at the outset that the review o f literature and the findings here imply that by far the majority o f women who choose to delay childbearing to their mid or even late thirties will be able to have a normal pregnancy and birth outcome. Nevertheless, the risks for several adverse outcomes do increase with age. In terms o f their impact on reproduction for an individual woman, the effects o f age on fecundity and risk o f miscarriage remain the most important age-related barriers to successful reproduction (Chapter 4). The effects on low birth weight and preterm delivery begin in the early thirties, are greater for first births and appear to continually increase with age. The effects o f age on the relative risk for low birth weight and preterm delivery for first births are comparable to those on fecundity (roughly about a 2 fold increase in relative risks as estimated by odds ratios from the 30s to the 40s compared to the 20s). However, the greatest effects o f age in terms o f absolute risk differences are its effects on fecundability and risk o f miscarriage. In interpreting the results o f the effects o f age on low birth weight and preterm delivery the statistical issue o f multiple comparisons problem is relevant. Since in reporting and comparing the results o f the effects o f age across ethnic and parity groups several comparisons were made the type I error rate needs to be adjusted accordingly to preserve the overall alpha level o f 0.05. However, the results here in general showed lack o f substantial differences across the four ethnic groups where the


184 multiple comparisons problem would pose the greatest problem particularly for smaller minority groups where the confidence intervals were wider due to lesser precision o f estimates that were based on smaller numbers o f births. Hence the results suggested not rejecting the null hypothesis and in any case showed an apparent lack o f any substantial differences in the effects o f age on the relative risk o f low birth weight and preterm birth across ethnic groups. Therefore, in practice, the multiple comparisons problem was less o f an issue here. Adjusting for the socioeconomic and behavioral factors that are associated with both delayed childbearing and the risk o f low birth weight and preterm delivery produced results that were in the predicted direction; i.e., adjustment resulted in increased estimates o f risk associated with age. However, the effects o f adjustment were fairly small in magnitude; i.e., adjustment increased the estimates o f the age effect on the order o f 10 % or less when comparing adjusted odds ratios with unadjusted ones. In particular, the effect o f adjusting for adequacy o f prenatal care on the estimates o f age effect was essentially negligible once the effects o f other factors, namely maternal education, marital status, smoking and in the case o f births o f second or higher order, parity and interval since termination o f last pregnancy and previous preterm were adjusted. This lack o f a substantial effect from the adjustment for the socioeconomic and the other control variables deserves comment. It should first be noted that consideration o f socioeconomic and other control variables were much more


18 5

important in the case o f cause-specific infant mortality and even more so in the case o f Down syndrome (see below). On the other hand the small magnitude o f the adjustment effect in the case o f low birth weight and preterm delivery outcomes probably reflects the following: i) there is significant but fairly moderate correlations between socioeconomic factors and delayed childbearing. While several studies including the estimates obtained in this study have shown significant correlations between advanced maternal age at first birth and socioeconomic and the other control factors, the size o f the correlations are not very large particularly for births o f second or higher order; ii) the effects o f advanced age on low birth weight and preterm delivery, while significant, are fairly modest; equivalent to an odds ratio o f about 1.6 or so for first births and about 1.3 for births o f second or higher order in the case o f low birth weight; iii) the effects of control variables on low birth weight and preterm delivery, including the effects o f the socioeconomic factors, are also moderate in magnitude and on the order o f an odds ratio o f 2 or so for most factors used here as control variables. While examining the precise contribution o f the above explanations may be important, it is perhaps more important to point out that what these results underscore is lack o f an adequate model o f causation and o f the contributions o f socioeconomic factors and prenatal care to the etiology and prevention o f low birth weight and preterm delivery. The results showed the effect o f age on low birth weight and preterm delivery to be fairly similar across ethnic groups in terms o f the age effects on relative risks.


186 However, in terms o f absolute risk differences, the effects o f age were greater for African Americans and Puerto Ricans who have higher baseline risks o f low birth weight and preterm delivery. The most substantial interaction between the effects o f age and other maternal attributes on low birth weight and preterm delivery was in the case o f birth order, however. Specifically, the effects o f age were much greater for first births than for births o f second or higher order in all four ethnic groups. This may be mediated at least in part by the effects o f age on hypertensive disorders during pregnancy and in particular preeclampsia that is far more common for first births 345 6

The results also suggest that advancing maternal age affects both gestational age and the rate of intrauterine growth. The effects o f age on low birth weight tended to be greater than its effects on preterm delivery (Chapter 6). This was the case for all the ethnic groups studied here and for both first births and for births o f second or higher order. Hence it appears that advancing maternal age increases the risk o f low birth weight by two distinct mechanisms, namely by shortening the period o f gestation, as well as by decreasing the rate of intrauterine growth. In terms o f magnitude and direction, the results obtained for the analyses o f the effects o f advancing maternal age on low birth weight and preterm delivery are consistent with the other population-based studies in the United States and Sweden with adequate power to examine the effects of age on low birth weight. In addition, the results obtained here provide estimates of the effects o f age on both low birth


187 weight and preterm delivery by birth order and ethnicity for the entire United States birth cohort. With current levels o f obstetric and neonatal care in the United States, by far most low birth weight infants will survive 7’8. Furthermore, advances in obstetric and neonatal care appear to have been accompanied by a decrease in the prevalence o f handicapping conditions among very low birth weight infants 9. However, the care o f these babies also incurs very high direct and indirect financial and human costs ,35 years o f age. This set o f findings is consistent with the hypothesis that ethnic differences in the impact o f advanced maternal age on the risk o f Down syndrome in the United States might reflect differences in their use o f prenatal diagnostic technologies. Review o f literature did not yield any evidence to suggest that there might be biological reasons for the observed age-related differences in the birth prevalence o f Down syndrome among the three ethnic groups. Previous studies have reported ethnic variations in gestational age-specific levels o f biochemical screening markers including alpha-fetoprotein, human chorionic gondatropin and unconjugated estriol I8'19. In one study 18, the same general pattern o f differences was observed for all three markers and the authors concluded that averaging the values for all ethnic groups tends to inappropriately lower the calculated Down syndrome risks for African American and Asian women. While another study 19 o f the ethnic differences in levels o f the biochemical screening markers also showed significant differences, the authors concluded that such differences would only be expected to have a minimal effect on the odds of detecting Down syndrome. The results o f neither study suggest that ethnic differences in the levels o f biochemical markers might be responsible for the substantial magnitude o f the age-related differences in the risk o f Down syndrome or


189 utilization

o f amniocentesis that were observed among the three ethnic groups in

this study. No other evidence was found to suggest that there might be biological reasons for the observed age-related differences in the birth prevalence o f Down syndrome among the three ethnic groups. Still, the possibility that the differences observed in this study might be to some extent biological in origin, or have some other explanation unrelated to prenatal diagnostic utilization, cannot be excluded. However, given the findings o f marked ethnic variations in amniocentesis use, differences in access to and/or decision to utilize prenatal diagnostic services among women in the three ethnic groups seem a more plausible explanation. The analyses reported here were based on birth certificate data that are likely to represent underestimates o f the true birth prevalence o f Down syndrome 16 due to incomplete case ascertainment and/or reporting. There may also be differential underdiagnosis and/or reporting by maternal ethnicity. However, a pattern o f ethnicity and age-specific biases in case ascertainment and/or reporting that would produce the results observed in this study seems unlikely. Specifically, differential misclassifications are unlikely to result in higher estimates o f age-related increases for Mexican Americans and African Americans. Such diagnostic and/or reporting biases would require substantial overestimation o f age-related increases in Down syndrome for Mexican Americans and African Americans and/or substantial underestimation o f age-related increases for non-Hispanic whites. There is no a priori


190 reason or empirical evidence from previous studies that would predict such a pattern o f differential misclassification. Therefore, even though, birth certificate data are likely to underestimate the true birth prevalence o f Down syndrome, there is no reason to believe that there is much greater underreporting / less complete case ascertainment for non-Hispanic white women >35 years o f age as compared with their African American and more so Mexican American counterparts. Birth certificate data may also represent underestimates o f amniocentesis utilization

and furthermore do not specify the indication for or the timing o f the

procedure. Although it is possible that underreporting o f prenatal diagnostic procedures may be correlated with ethnicity and/or age o f the mother, there is no evidence to suggest that such a phenomenon is operating to explain the ethnic u tilization

differences observed in the study. In contrast, previous data have identified

similar patterns o f utilization in the United States ,5’17-20 and abroad 21. Nevertheless, given the nature o f the data used for this study, the findings on the ethnic differences in amniocentesis utilization need to be confirmed and elaborated by other studies that can measure amniocentesis usage more precisely. Possible explanations for lower u tiliz a tio n o f amniocentesis for Mexican American and African American women also need further study. In order to have the optimum opportunity for benefiting from prenatal diagnostic technologies, women need to initiate prenatal care early in their pregnancies. Late initiation and/or lack o f prenatal care might be one impediment for African American and Mexican American


191 women’s access to prenatal diagnostic technologies “ . Since African American, and to a greater extent Mexican American women are less likely to initiate prenatal care during the first trimester o f their pregnancies, lower rates o f amniocentesis use among African American and Mexican American women might be in part related to their late initiation or lack of prenatal care. Whatever might be the role o f timing o f initiation o f prenatal care, however, issues relevant to access and utilization o f prenatal diagnostic technologies are clearly much more complex than timing or adequacy o f prenatal care alone. Important issues to consider comprise a host o f individual, family and system related factors. These factors include parental preferences, including cultural and religious beliefs and values, health insurance, differences in the content and quality o f prenatal care, and specific public and private policies regarding prenatal diagnostic and intervention services. There are relatively little data on ethnic, or in general socioeconomic, differences in access to or use o f prenatal diagnostic services: there are not sufficient data regarding how women in different ethnic groups reach their decisions about undergoing prenatal diagnosis, how they are informed or come to perceive their reproductive risks, and how they make their choices about continuation or termination of pregnancy in the event of discovering a fetal anomaly 2j. In addition, factors related to physicians and other health care providers who affect the decisions o f women to undergo a test or understand and trust its results also need to be considered.


192 The role o f socioeconomic and cultural factors as determinants of access to. and choices regarding, prenatal diagnostic services need further study. The results also showed a strong influence o f education on both the agerelated increases in the birth prevalence o f Down syndrome and age-specific odds o f amniocentesis use (Chapter 6). Again, as for the case o f ethnicity, it seems unlikely that the effects o f education found here were simply data artifacts. There is neither an a priori reason nor empirical evidence from previous studies to suggest that agerelated birth prevalence o f Down syndrome might be systematically underreported for more educated non-Hispanic white women. Hence the observed association between higher levels o f education and a lower rate of age-related increase in the birth prevalence o f Down syndrome appears to be a genuine observation. Furthermore, the results of comparing age and education-specific rates o f amniocentesis use were consistent with the hypothesis that the effect of education on age-related increase o f the birth prevalence o f Down syndrome might operate by the higher use o f prenatal diagnosis and selective termination among women with higher levels o f education. One possible explanation for the strong effect o f education on the use o f prenatal diagnostic services is issues related to access 2. More highly educated women who are also more likely to be in more affluent households tend to have better medical insurance and coverage for comprehensive prenatal care that includes access to a full range o f prenatal diagnostic services. More highly educated women are also more likely to be better informed about the availability and use of prenatal diagnostic


193 services 2. It is also possible that education might be correlated with cultural values and belief systems including religious beliefs that might in turn affect a woman's choices with regard to prenatal diagnostic services. In addition to access, knowledge and issues related to women’s preferences, two lines o f argument from the human capital literature could explain the education effect on prenatal diagnostic services and selective termination. First, more highly educated women have a greater opportunity cost o f time, a lower demand for quantity o f children 24 and might worry more about the impact o f a child with birth defect on their lives and careers2. Second, as Becker has argued the interaction between the quantity and quality o f children 25 implies that households with higher investments per child have a lower demand for quantity o f children and a higher demand for their quality. More highly educated women tend to belong to households with higher incomes. More highly educated women, as well as households with higher levels of incomes are more likely to invest more per child, and hence may be more likely to want fewer babies and also have a greater demand for birth o f a baby without any knowable major anomalies26. Future studies should address the mechanisms by which education, ethnicity and other socioeconomic factors such as income or cultural and religious beliefs affect the use o f prenatal diagnosis. In particular, the mechanisms by which socioeconomic factors affect the use o f prenatal diagnostic services can have very different implications for the presumed effects o f policy interventions that one might


consider. If socioeconomic differences in use o f prenatal diagnosis primarily reflect issues related to the financial costs o f obtaining these services such as health insurance coverage, paying for the abortion o f a fetus with birth defect or state-level abortion policies, the anticipated effects o f various policy options would be very different than they would be if the disparities in the use o f prenatal diagnosis were predominantly related to the knowledge o f the effective use o f services or to the women’s preferences such as their religious beliefs or their demand or desire for birth o f a baby without any known defects. A rather puzzling observation in the data was that the negative effect o f education on age-related increases in the birth prevalence of Down syndrome appeared to be operational for non-Hispanic whites but not for African Americans. This was even more difficult to interpret since the positive effect o f education on the rate o f amniocentesis use was similar in African Americans and non-Hispanic whites. One possibility o f course is that education has a different effect on the decision to terminate a fetus found to have Down syndrome in African Americans as compared with non-Hispanic whites. Unfortunately, this question cannot be investigated with the data sets that were available for this study since the data do not include indication for or timing o f amniocentesis or follow up data on the decisions that were made once the diagnosis o f an anomaly was made. Another possibility is that the rather crude classification o f education used here, 12 years o f education or greater compared with less than 12 years o f education,


195 represents quite heterogeneous groups o f women in African Americans and nonHispanic whites. Hence the results do no provide an accurate comparison o f education effects in African Americans vs. non-Hispanic whites. Still, it is difficult to explain why the effects o f education groups on the age-specific odds o f amniocentesis were similar for African Americans and non-Hispanic whites but not for age-related increases in birth prevalence o f Down syndrome without allowing for different termination rates once the anomaly is diagnosed. In general with regards to education, the study here leaves several questions unanswered. The study did not examine the various ways o f estimating the effects o f education in terms o f different functional forms of an interval scale measure or presence of any threshold effects. Furthermore, this study cannot disentangle the various mechanisms by which education might decrease the effect o f maternal age on the risk o f a birth defect. The analysis suggests that at least part o f the effect o f maternal education operates through the differential use o f prenatal diagnostic services with selective termination o f the fetuses affected with major birth defects. However, this study cannot reach a definitive conclusion with regard to this question as direct evidence would require termination data, as well as data on age and education-specific rates o f amniocentesis, indications for amniocentesis and use of any other prenatal diagnostic services. The effect o f paternal as compared with maternal education was not addressed in this study. The study by Parker and colleagues 27 suggests that maternal education


196 might be the best single predictor o f birth outcome. However, the effects o f paternal vs. maternal education, as well as the effect o f income vs. education have not been studied in the specific context o f the parental choices with regard to prenatal diagnostic and intervention services. As explained above, the sign and magnitude o f such effects, as well as the mechanisms by which they operate to produce the observed choices o f parents with regard to prenatal diagnostic and intervention services have important implications for predicting the presumed effects o f policy choices in the field o f prenatal diagnosis and intervention.

Advanced maternal age was associated with a lower risk o f infant death when all causes were combined. In contrast to the case o f low birth weight, however, the result o f the analysis of the effects o f age on the hazard o f cause-specific infant mortality was quite sensitive to adjustments for sociodemographic factors that are associated with delayed childbearing. The overall survival advantage for the infants bom to older mothers was entirely due to the effects o f sociodemographic and reproductive history factors associated with delayed childbearing. Furthermore, the effect o f advanced maternal age differed substantially across the various causes o f death. The results were consistent with the hypothesis that advanced maternal age increases the risk associated with biologically related causes o f infant mortality: infant deaths due to congenital anomalies and abnormalities of the placenta. On the other hand, advanced maternal age was associated with lower risks o f infant mortality


197 due to the set o f causes that are related to childcare practices such as infant deaths due to unintentional and intentional injuries and Sudden Infant Death Syndrome. These associations might be due to the effect o f birth timing itself or to the effect o f unmeasured socioeconomic factors associated with delayed childbearing or both. In particular, the effect o f advanced maternal age to lower the risk o f Sudden Infant Death syndrome might be related to differences in breast feeding or infant sleep position between younger and older mothers as supine sleep position and breast feeding are associated with lower risk o f SIDS.

In conclusion, the majority o f women who delay their childbearing to their mid or even late thirties are likely to have a normal pregnancy and birth outcome. However, the risk o f several adverse pregnancy and birth outcomes do increase substantially with age and a considerable minority o f women who delay their childbearing to their mid or late thirties can be expected to have either difficulty to conceive or to successfully carry a normal pregnancy to term. The major adverse birth outcomes include low birth weight, preterm delivery and birth defects the risk o f all o f which, and in particular some o f the birth defects, do increase with age. Birth defects constitute one of the leading causes o f infant mortality in the United States 28 and the fifth leading cause o f years o f potential life lost29. Most children with birth defects do not die in infancy and require special medical care in addition to special education, rehabilitation and other nonmedical services.


198 Interventions are often not fully corrective making birth defects a major cause o f childhood and adult disability. Using a human capital approach, Waitzman and colleagues 28 estimated the direct costs o f medical care, developmental and special education services and the indirect costs o f lost work and household productivity due to cerebral palsy and 17 o f the most important structural birth defects. Their results implied that the combined costs o f the 18 conditions in the United States were $8 billion in 1992 29. Conditions with the highest per case and total costs included Down syndrome [$ 451,000 per case and $ 1.8 billion total costs] and spina bifida [$ 294,000 per case and $ 489

m illio n

total costs]. These estimates did not include time

costs o f the family or the psychosocial costs o f illness. For these and other reasons, the above human capital estimates may underestimate the level that public is willing to pay to prevent these conditions. The advent of modem prenatal diagnosis in the past three decades allows several birth defects to be detected in utero and gives parents the option o f selective termination once the diagnosis o f an abnormally is made. There appear to be strong socioeconomic differences in access to or use o f prenatal diagnostic services and selective termination. The reasons for these differences are incompletely understood at best and need further study. These differences appear to result in substantial socioeconomic differences in the birth prevalence o f birth defects, which imply a widening gap in socioeconomic disparities in birth outcomes to women who give birth at older ages. These socioeconomic differences in the risk o f birth defects also


199 imply that the burden o f care for inlants with birth defects might fell disproportionately on families with fewer resources as well as on publicly funded health care.

Reference List (1) Casterline JB, Lee RD, Foote KA. Fertility in the United States: New patterns, new theories. New York: Population Council, 1996. (2) Kolker A. Advances in prenatal diagnosis. Social-psychological and policy issues. Inti J o f Technology Assessment in Health Care 1989; 5:601-617. (3) Roberts JM. Preeclampsia: what we know and what we do not know. Semin Perinatol 2000; 24:24-28. (4) Dekker GA. Risk factors for preeclampsia. Clin Obstet Gynecol 1999 Sep 42:422-435. (5) Dekker G A Sibai BM. The immunology of preeclampsia. Semin Perinatol 1999 Feb 23:24-33. (6) Redman CW, Sacks GP, Sargent IL. Preeclampsia: an excessive maternal inflammatory response to pregnancy. Am J Obstet Gynecol 1999 Feb 180:499-506. (7) Lee KS, Paneth N, Gartner LM, Pear 1man MA, Gruss L. Neonatal mortality: an analysis of the recent improvement in the United States. Am J Public Health 1980 Jan 70:15-21. (8) Lee KS, Khoshnood B, Hsieh H, Kim BI, Schreiber MD, Mittendorf R. Which birth weight groups contributed most to the overall reduction in the neonatal mortality rate in the United States from I960 to 1986? Paediatric & Perinatal Epidemiology 1995; 9(4):420-430. (9) Lee KS, Kim BI, Khoshnood B, Hsieh HL, Chen TJ, Herschel M et al. Outcome of very low birth weight infants in industrialized countries: 1947-1987. American Journal o f Epidemiology 1995; 141(12):! 188-1193. (10) Lewit EM, Monheit AC. Expenditures on health care for children and pregnant women. 95-114. 1992. The Future o f Children. Ref Type: Serial (Book,Monograph) (1 1) Khoshnood B, Lee KS, Corpuz M, Koetting M, Hsieh HI, Kim BI. Models for determining cost of care and length o f stay in neonatal intensive care units. Int J Technol Assess Health Care 1996 Winter 1996; 12:62-71. (12) Lewit EM, Baker LS, Corman H, Shiono PH. The direct cost of low birth weight. Future of Children 1995; 5(l):35-56.


200 (13) Barker DJ, Fall CH. Fetal and infant origins o f cardiovascular disease. [Review]. Archives o f Disease in Childhood 1993; 68(6):797-799. (14) Barker DJ. The fetal and infant origins of disease. [Review]. European Journal of Clinical Investigation 1995; 25(7):457-463. (15) Bishop J, Huether C, Torfs C, Lorey F, Deddens J. Epidemiologic study of Down syndrome in a racially diverse California population, 1989-1991. American Journal o f Epidemiology 1997; 145(2):134-147. (16) Centers for Disease Control and Prevention. Down syndrome prevalence at birth—United States, 1983-1990. MMWR 1994; 43:617-622. (17) Krivchenia E, Huether CA, Edmonds LD, May DS, Guckenberger S. Comparative epidemiology of Down syndrome in two United States population, 1970-1989. Am J Epidemiol 1993 Apr 15 137:815-828. (18) O'Brien J, Dvorin E, Drugan A, Johnson M, Yaron Y, Evans M. Race-ethnicity-specific variation in multiple-marker biochemical screening: alpha-fetoprotein, hCG, and estriol. Obstetrics & Gynecology 1997; 89(3):355-358. (19) Benn P, Clive J, Collins R. Medians for second-trimester maternal serum alpha-fetoprotein, human chorionic gonadotropin, and unconjugated estriol; differences between races or ethnic groups. Clinical Chemistry 1997; 43(2):333-337. (20) Sokal DC, Byrd JR, Chen AT, Goldberg MF, Oakley GPJ. Prenatal chromosomal diagnosis. Racial and geographic variation for older women in Georgia. JAMA 1980 Sep 19 244:13551357. (21) Halliday J, Lumley J, Watson L. Comparison of women who do and do not have amniocentesis or chorionic villus sampling. Lancet 1995; 345:704-709. (22) Khoshnood B, Pryde P, Wall S, Singh J, Mittendorf R, Lee KS. Ethnic differences in the impact of advanced maternal age on the birth prevalence of Down syndrome in the United States. Am J Publ Hlth. In press. (23) Pryde PG, Drugan A, Johnson MP, Isada NB, Evans MI. Prenatal diagnosis: choices women make about pursuing testing and acting on abnormal results. Clinical Obstetrics & Gynecology 1993; 36(3):496-509. (24) Michael RT. Education and the Derived Demand for Children. In: Schultz T, editor. Economic of the Family: Marriage, Children and Human Capital. Chicago: The University of Chicago Press, 1974: 120-156. (25) Becker GS. A Treatise on the Family. Enlarged ed. Cambridge, MA: Harvard University Press, 1991. (26) Katz VL. Two trends in middle-class birth in the United States. Human Nature 4(4): 1993-382.


201 (27) Parker JD, Schoendorf KC, Kiely JL. Associations between measures o f socioeconomic status and low birth weight, small for gestational age, and premature delivery in the United States. Ann Epidemiol 1994; 4:271-278. (28) Waitzman NJ, Romano PS, Scheffler RM. Estimates of the economic costs o f birth defects. Inquiry 1994; 31:188-205. (29) Centers for Disease Control and Prevention. Economic costs o f birth defects and cerebral palsy-United States, 1992. Mortality and Morbidity Weekly Report 1995; 44:694-699.


202

APPENDIX A - TABLES AND FIGURES IN CHAPTER 2


CBR

TFR

30

25

S 20

2.5

I

o o o 15 On

d“ i O

OO

CO

t WO

NO

O

d" VO

ON d" NO

VO

O 4—4 i- *

WO 00 0

00 0

O 4 JD

VI

CD -*-•

^ -1

— 4

----

rZ

r— 4

WO 00 0

d ;

On

WO

4— 4

cm

4— 4

4— 4

•—
0 tc *3 (D

CM

I WO

1

CO

CD

00

o

wo

WO CM

NO

ON

CO OO NO

WO d" NO

0 0 00 d CM

wo NO NO CM CM

ON

Ou

CM

OO

0 0

CD CO

ON 4—1

ID'’ 4— 4

O CO CO 0 » < 4— 1

NO WO WO

00 CM wo

00 CM WO

wo

0 00 CO

wo ON 00 c-

0 0 NO r-

'“ H1

1—1

VO c~ On 4—1

c~ rOn 4—4

0 wo CO 00 ’—l

0

O CM

O On C "OO ’ 1

CM r~ On

CO c ON

d rOn 4—4

wo

NO

wo

4—1

fH

wo

^Z ,

cm

V ™4

CO ON WO

c -;

On

cm

4—1

cm

wo

NO 4—4

00

00

c -‘ 0

vd 0 4—1

d-

d

cm

1— 1

4—
i-c 3

CO

E

® CN

o

O oo

®

Cjtipjaj o g io a d s-sS y


o


Figure 2.7 - Shifts in age-specific fertility, parity > 2, United States, 1975, 1980, 1985, 1997 (Series 1-4)

& 60 Series 1 Series2 Series3 Series4

15-19

20-24

25-29

30-34

40-44 215

Maternal age (years)

35-39

216

APPENDIX B - TABLES AND FIGURES IN CHAPTER 6



Table 6.1(a) - Total number and proportion of live births by maternal age among four racial/ethnic groups - First Births - United States, 1989 - 1991 ( N = 3,346,822) Maternal Age (years) Racial/Ethnic Group

Number

35

African American

356,401

30.6 %(30.4 - 30.7)*

3.2% (3.1 ■■3.2)

Mexican American

288,865

34.3% (34.1 -34.5)

1.4 %(1.4 ■■1.5)

Non-Hispanic White

2,670,280

13.5% (13.5-13.6)

5.6 %(5.6 ■-5.7)

31,276

38.2 %(37.6 - 38.7)

1.9 %(1.8 ■ -2.1)

Puerto Rican * 95 %confidence interval

217

Table 6 .1(b) - Total number and proportion o f live births by maternal age among four racial/ethnic groups - Parity > 1 - United States, 1989 - 1991 ( N = 5,119,037)______________________________ Maternal Age (years) Racial/Ethnic Group

Number

35

African American

663,934

7.6 % (7.5 - 7.7)*

10. 1% ( 1 0 . 0 - 1 0 . 2 )

M exican American

545,669

6.4 % (6.4 - 6.5)

10. 1% ( 1 0 . 0 - 1 0 . 1 )

3,826,059

2.5 % (2.4 - 2.5)

13. 4% ( 1 3 . 3 - 1 3 . 4 )

51,451

10.1% ( 9 . 8 - 1 0 . 3 )

7.5 % (7.2 - 7.7)

Non-Hispanic White

Puerto Rican * 95 % confidence interval

Table 6.2(a) - Odds ratios^ for the association o f delayed childbearing ( maternal age > 35 years) with low (< 2.5 kg) birth weight among four racial / ethnic groups - First Births - United States, 1989 - 1991

VLBW* ( < 1 . 5 kg)

MLBW* ( 1 .5 - 2 .5 kg)

LBW* ( < 2.5 kg)

1 .8 0 (1 .6 3 - 1.99)**

1.54 (1.45 - 1.63)

1.61 (1.53 - 1.70)

M exican Americans

2 .1 7 ( 1 .6 8 - 2 .7 9 )

1.70(1.51 -1 .9 1 )

1 .7 7 (1 .5 9 - 1.97)

Non-Hispanic Whites

1.81 (1.72 - 1.90)

1.54 (1.51 - 1.58)

1.59(1.55 - 1.62)

Puerto Ricans

2 .7 4 ( 1 .6 4 - 4 .5 7 )

1.48(1.11 - 1.96)

1.66 (1 .2 8 -2 .1 4 )

A H **

1 .8 2 (1 .7 5 - 1.90)

1.55 (1.52 - 1.58)

1.60 (1 .5 7 - 1.63)

Racial/Ethnic Group African Americans

* Reference group for calculation o f odds ratios: mothers 20 - 34 years o f age. * VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight ** 95 % confidence interval ** The odds ratios for All are the Mantel-Haenszel estimates o f the combined odds ratios for all four racial/ethnic groups.


Table 6.2(b) - Odds ratios^ for the association o f delayed childbearing ( maternal age > 35 years) with low (< 2.5 kg) birth weight among four racial / ethnic groups - Parity > 1 - United States, 1989 - 1991

VLBW * ( < 1.5 kg)

MLBW* ( 1 .5 - 2 .5 kg)

LBW* ( < 2 .5 kg)

1.20 (1.15 - 1.25)**

1.13(1.10-1.15)

1.14(1.12-1.16)

M exican Americans

1 . 6 0 ( 1 . 4 7 - 1.75)

1. 37(1.31 - 1.42)

1.41 ( 1 . 3 6 - 1.46)

Non-Hispanic Whites

1. 29(1. 25 - 1.34)

1.11 ( 1 . 0 9 - 1 . 1 2 )

1.13(1.12-1.15)

Puerto Ricans

1 . 6 7 ( 1 . 4 0 - 1.99)

1 . 1 8 ( 1 . 0 8 - 1.29)

1 . 2 6 ( 1 . 1 6 - 1.36)

AH* *

1 . 2 9 ( 1 . 2 6 - 1.32)

1.14(1.12-1.15)

1.16(1.15-1.17)

Racial/Ethnic Group African Americans

^ Reference group for calculation o f odds ratios: mothers 20 - 34 years o f age. * VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight ** 95 % confidence interval ** The odds ratios for All are the Mantel-Haenszel estimates o f the combined odds ratios for all four racial/ethnic groups.

220


Table 6.2(c) - Risk differences * for the association o f delayed childbearing ( maternal age > 35 years) with low (< 2.5 kg) birth weight among four racial / ethnic groups - First Births - United States, 1989 - 1991

VLBW * ( < 1 . 5 kg)

MLBW * ( 1 . 5 - 2 . 5 kg)

LBW* ( < 2 . 5 kg)

African Americans

1.1 (1.0 - 2.1)

3.9 ( 3 . 3 - 4 . 5 )

5.3 ( 4 . 7 - 6 . 0 )

M exicans

0.8 ( 0 . 4 - 1 . 1 )

2.9 (2.1 - 3 . 7 )

3.7 (2.8 - 4.5)

Non-Hispanic Whites

0.5 (0.5 - 0.6)

2.1 ( 2 . 0 - 2 . 2 )

2.6 (2.4 - 2.7)

Puerto Ricans

1.7 ( 0 . 4 - 3 . 0 )

2.8 (0.4 - 5.2)

4.3 ( 1 . 7 - 6 . 9 )

Racial/Ethnic Group

* Expressed per 100 compared with the reference group o f mothers 20 - 34 years o f age in the same ethnic group. * VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight ** 95 % confidence interval

221

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission

Table 6.2(d) - Risk differences * for the association o f delayed childbearing ( maternal age > 35 years) with low (< 2.5 kg) birth weight among four racial / ethnic groups - Parity > 1 - United States, 1989 - 1991

V L B W ’1' ( < 1.5 kg)

MLBW * ( 1 . 5 - 2 . 5 kg)

LBW* ( < 2 . 5 kg)

African Americans

0.4 (0.3 - 0.5)

1.1 ( 0 . 9 - 1 . 3 )

1.4 ( 1 . 2 - 1 . 6 )

M exicans

0.3 (0.3 - 0.4)

1.1 ( 1 . 0 - 1 . 3 )

1.4 ( 1 . 3 - 1 . 6 )

Non-H ispanic W hites

0.1 (0.1 - 0 . 2 )

0.3 (0.3 - 0.4)

0.5 (0.4 - 0.5)

Puerto Ricans

0.7 ( 0 . 4 - 1 . 0 )

1.1 ( 0 . 5 - 1 . 7 )

1.7 (1.1 - 2 . 4 )

Racial/Ethnic Group

* Expressed per 100 compared with the reference group o f mothers 20 - 34 years o f age in the same ethnic group. * V LBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight ** 95 % confidence interval

222


T ab le 6 .2 (e ) - O d d s ratios^* for th e a sso c ia tio n o f d elayed ch ild b ea rin g w ith lo w birth w eig h t (< 2.5 kg) a m o n g four racial / eth n ic grou p s - First B irth s - U n ited S tates, 1 9 8 9 - 1991 V L B W * ( < 1.5 k g) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years

1 . 4 4 ( 1 . 3 6 - 1 .5 2 )* *

1.85 ( 1 . 7 0 - 2 . 0 2 )

2 .0 3 ( 1. 65 - 2 . 5 0 )

M ex ica n -A m erica n s

1 . 4 9 ( 1 . 3 0 - 1. 71)

2.30(1.87-2.84)

2.42(1.47-3.98)

N o n -H isp a n ic W h ites Puerto R ica n s

1.23 ( 1 . 1 9 - 1. 28) 1.13 ( 0 . 8 7 - 1 .4 8 )

1 . 7 7 ( 1 . 6 8 - 1 .8 6 )

2.24(1.99-2.51)

2 .3 2 ( 1 . 6 4 - 3 . 2 8 )

1.93 ( 0 . 7 9 - 4 . 7 1 )

A frican A m erica n s

M L B W * ( 1 . 5 - 2 . 5 kg) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years


1 . 2 8 ( 1 . 2 4 - 1. 32)

1. 56 ( 1 . 4 9 - 1 .6 5 )

1 . 7 4 ( 1 . 5 4 - 1. 98)


1 . 2 0 ( 1 . 1 3 - 1. 28)

1. 68 ( 1. 5 2 - 1. 85)

1 . 5 4 ( 1 . 1 9 - 1. 98)

N o n -H isp a n ic W h ites

1 . 2 0 ( 1 . 1 9 - 1 .2 2 )

1. 54 (1.51 - 1. 58)

1.91(1.81-2.02)

Puerto R ica n s

1 . 2 0 ( 1 . 0 7 - 1. 35)

1.31 ( 1 . 0 7 - 1 . 6 1 )

2. 5 8 ( 1 . 7 7 - 3 . 7 7 )

L B W * ( < 2 .5 k g ) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years


1 . 3 2 ( 1 . 2 8 - 1. 35)

1 . 6 4 ( 1 . 5 7 - 1 .7 2 )

1.83 ( 1 . 6 4 - 2 . 0 5 )


1.24(1.17-1.31)

1 . 7 7 ( 1 . 6 2 - 1 .9 4 )

1.66(1.32-2.09)


1.21 ( 1 . 1 9 - 1 . 2 2 )

1.58 ( 1. 55 - 1. 61)

1.97(1.87-2.07)

Puerto R ica n s

1.19(1.07-1.33)

1.48 ( 1. 23 - 1 .77)

2. 51 ( 1 . 7 6 - 3 . 5 8 )

*

R eferen ce group for ca lcu la tio n o f o d d s ratios: m others 2 0 - 2 9 years o f age.

* V L B W = v ery lo w birth w eig h t; M L B W = m od erately lo w birth w eight; L B W = lo w birth w eigh t * * 9 5 % c o n fid e n c e interval

223

T ab le 6 .2 (f) - O d d s ratios^

for th e a sso c ia tio n o f d elayed ch ild b earin g w ith lo w birth w eigh t (< 2 .5 kg)

a m o n g fou r racial / eth n ic grou p s - Parity > 1 - U n ited States, 1 9 8 9 - 1991 _________________________________________________________ V L B W * ( < 1 . 5 kg) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years

1 . 2 0 ( 1 . 1 6 - 1 .2 4 )* *

1.24(1.19-1.31)

1 . 2 9 ( 1 . 1 6 - 1.44)


1 . 4 0 ( 1 . 3 0 - 1. 50)

1.73 ( 1 . 5 7 - 1 .9 0 )

1.93 ( 1 . 6 0 - 2 . 3 3 )


0.91 ( 0 . 8 8 - 0 . 9 4 )

1 . 2 0 ( 1 . 1 5 - 1. 25)

1 . 6 2 ( 1 . 4 9 - 1. 75)

1.36(1.17-1.58)

1.74(1.42-2.12)

2.07(1.42-3.02)


Puerto R ica n s

M L B W * ( 1 . 5 - 2 . 5 kg) 3 0 -3 4 years

3 5 - 3 9 years

> 4 0 years


1.11(1.09-1.13)

1.15(1.12-1.17)

1.21 ( 1 . 1 4 - 1 . 2 8 )


1 . 0 8 ( 1 . 0 4 - 1. 12)

1.34 ( 1 . 2 8 - 1 .4 0 )

1 . 6 7 ( 1 . 5 4 - 1.82)


0 .8 3 (0 .8 2 - 0 .8 4 )

1 . 0 0 ( 0 . 9 8 - 1. 01)

1 . 3 2 ( 1 . 2 7 - 1. 37)

Puerto R ica n s

1.09(1.02-1.16)

1 . 1 6 ( 1 . 0 5 - 1 .2 8 )

1.41 ( 1 . 1 7 - 1 . 7 1 )

L B W * ( < 2 .5 k g ) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years


1.13(1.11-1.14)

1.17(1.14-1.19)

1.23 ( 1 . 1 7 - 1.2 9 )

M ex ica n -A m er ica n s

1.12(1.09-1.16)

1.40 ( 1 . 3 4 - 1. 45)

1. 72 ( 1 . 5 9 - 1. 85)


0 .8 4 (0 .8 3 - 0 .8 5 )

1 . 0 2 ( 1 . 0 1 - 1 .0 4 )

1 . 3 6 ( 1 . 3 2 - 1. 41)

P uerto R ica n s

1.13 ( 1 . 0 6 - 1 . 2 0 )

1.25 ( 1 . 1 4 - 1. 36)

1.51 ( 1 . 2 7 - 1. 80)

9 R eferen ce group for ca lcu la tio n o f o d d s ratios: m others 2 0 - 2 9 years o f age. * V L B W = very lo w birth w eigh t; M L B W = m od erately lo w birth w eigh t; L B W = lo w birth w eig h t

** 9 5 % c o n fid e n c e interval


T ab le 6 .3 (a ) - A ttrib u table fraction s o f lo w birth w eig h t a m o n g th e e x p o sed (A F e ) and attributable fraction s in th e p o p u la tio n (A F p ) a sso cia ted w ith d ela y ed ch ild b ea rin g ( > 3 5 y e a r s ) am o n g four ra cia l/eth n ic g ro u p s - First B irth s - U n ited S tates, 1 9 8 9 - 1991________________________________

V L B W * ( < 1.5 k g )

M L B W * (1. 5 - 2. 5 kg)

L B W * (< 2 .5 k g )

R a cia l/E th n ic G roup

AFe*

A Fp**

AFe*

A F p **

AFe*

A Fp**

A frica n A m erica n s

0 .4 4

0 .0 3

0 .3 5

0 .0 2

0 .3 8

0 .0 3

0 .5 4

0 .0 2

0.4 1

0.0 1

0 .4 3

0 .0 2


0 .4 5

0 .0 5

0 .3 5

0 .0 3

0 .3 7

0 .0 4

Puerto R ica n s

0 .6 4

0 .0 5

0 .3 2

0.0 1

0 .4 0

0 .0 2

M e x ic a n A m erica n s

•* VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight *AFe = (RR - 1 ) / RR where RR is the estimated relative risk as approximated by the odds ratio. **AFp = Pc (RR - 1 ) / RR = Pc AFe where Pc is the exposure rate among cases.

225


T a b le 6 .3 (b ) - A ttrib u tab le fra ctio n s o f lo w birth w e ig h t a m o n g th e ex p o sed (A F e) and attributable fra ctio n s in th e p o p u la tio n (A F p ) a sso c ia ted w ith d ela y ed ch ild b ea rin g ( > 35 y e a r s ) am on g fou r ra cia l/eth n ic grou p s - Parity > 1 - U n ited States, 1 9 8 9 - 1991

V L B W * ( < 1.5 k g )

L B W * ( < 2 . 5 kg)

M L B W * (1. 5 - 2. 5 kg)

R a cia l/E th n ic G roup

AFe*

A F p **

A F e*

A Fp**

AFe*

A F p **


0. 1 8

0 .0 2

0.12

0 .0 2

0.14

0 .0 2

M e x ic a n A m erica n s

0 .4 2

0 .0 7

0 .2 7

0 .0 4

0 .3 0

0 .0 4


0 .2 6

0 .0 5

0. 11

0 .0 2

0.14

0 .0 2

P uerto R ica n s

0 .4 5

0 .0 6

0 .2 2

0 .0 2

0 .2 6

0 .0 3

'* VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight *AFe = (RR - 1 ) / RR where RR is the estimated relative risk as approximated by the odds ratio. **AFp = Pc (RR - 1 ) / RR = Pc AFe where Pc is the exposure rate among cases.

226


Table 6.3(c) - Attributable fractions o f low birth weight in the population (AFp) associated with delayed childbearing ( > 35 y e a r s) among four racial/ethnic groups - All Births - United States, 1 9 8 9 - 1991 VLBW ( < 1 . 5 kg)

M LBW ( 1 . 5 - 2. 5kg)

LBW ( < 2.5 kg)

AFp*

AFp*

AFp*

African Americans

0.05

0.04

0.05

M exican Americans

0.09

0.05

0.06

Non-H ispanic Whites

0.10

0.05

0.06

Puerto Ricans

0.11

0.03

0.05

Racial/Ethnic Group

* VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight *AFp = Pc (RR - 1 ) / RR = Pc AFe where Pc is the exposure rate among cases.

227

Table 6.4(a) - Logistic regression analysis o f the association between delayed childbearing (> 35 y e a r s) and risk o f low (< 2.5 kg) birth w eight am ong four racial/ethnic groups First births - United States, 1989 - 1991_________________________________

Adjusted Odds Ratios * (95 % confidence intervals)

Racial/Ethnic Group

VLBW * ( < 1 . 5 kg)

MLBW* ( 1 . 5 - 2 . 5 kg)

LBW* ( < 2.5 kg)

African Americans

1. 80( 1. 63 - 1.99)

1.61 ( 1 . 5 2 - 1 . 7 1 )

1.67 ( 1 . 5 9 - 1.76)

M exican Americans

2.18(1.69-2.81)

1.73 ( 1 . 5 4 - 1.95)

1.80(1.62-2.01)

N on-H ispanic White;

1. 94( 1. 85 - 2 . 0 4 )

1.68 ( 1 . 6 4 - 1.72)

1.73 ( 1 . 6 9 - 1.76)

•Puerto Ricans

2.69(1.60-4.52)

1 . 4 9 ( 1 . 1 2 - 1.98)

1.66(1.29-2.15)

* Adjusted odds ratios obtained for mothers > 3 5 years o f age in each racial/ethnic group compared to the mothers in the sam e racial/ethnic group 20 - 34 year o f age - separate logistic m odels which included maternal age, education, marital status, prenatal care and sm oking were estimated for each o f the racial/ethnic groups. *VLBW =very low birth weight; M LBW =moderately low birth weight; LBW =low birth weight


Table 6.4(b ) - L ogistic regression analysis o f the association betw een delayed childbearing (> 35 y e a r s) and risk o f low (< 2.5 kg) birth w eight am ong four racial/ethnic groups Parity > 1 - U nited States, 1989 - 1991 Adjusted O dds Ratios * (95 % confidence intervals)

R acial/E thnic Group

VLBW * ( < 1 . 5 kg)

M LBW * ( 1 . 5 - 2 . 5 kg)

LBW * ( < 2.5 kg)

African A m ericans

1 . 3 2 ( 1 . 2 4 - 1.40)

1.27 ( 1 . 2 4 - 1.31)

1 . 2 9 ( 1 . 2 5 - 1.32)

M exican A m ericans

1 . 7 7 ( 1 . 5 9 - 1.96)

1.44 ( 1 . 3 7 - 1.51)

1.50 (1.43 - 1.56)

N on -H isp anic White;

1 . 5 2 ( 1 . 4 6 - 1.59)

1.39 ( 1 . 3 7 - 1.41)

1.41 ( 1 . 3 9 - 1.44)

.Puerto R icans

1.82(1.36-2.43)

1.33 ( 1 . 1 6 - 1.52)

1 . 4 0 ( 1 . 2 4 - 1.58)

* Adjusted odds ratios obtained for mothers > 35 years o f age in each racial/ethnic group compared to the m others in the sam e racial/ethnic group 2 0 - 3 4 year o f age - separate logistic m odels w hich included maternal age, education, marital status, prenatal care and sm oking were estim ated for each o f the racial/ethnic groups. 229

*V L B W =very low birth weight; M LBW =m oderately low birth weight; L B W =low birth w eight


T a b le 6 .4 (c )-L o g is tic regression a n a ly sis o f th e a sso c ia tio n b etw een d elayed ch ild b earin g and risk o f

A d ju sted O d d s R a tio s* - V L B W * ( < 1.5 k g) 3 0 -3 4 years

3 5 - 3 9 years

> 4 0 years

1 . 5 2 ( 1 . 4 4 - 1 .60)**

1.95 ( 1. 78 - 2 . 1 2 )

2.09(1.69-2.57)


1.55 ( 1. 35 - 1. 77)

2.37(1.93 -2.92)

2.46(1.50-4.05)


1.41 ( 1 . 3 6 - 1 .4 6 )

2.0 1 ( 1 . 9 1 - 2 . 1 2 )

2 .4 4 ( 2 . 1 7 - 2 . 7 4 )

P u erto R ica n s

1 .2 2 (0 .9 3 - 1. 59)

2 .3 5 ( 1 . 6 6 - 3 . 3 3 )

1.91 (0 .7 8 - 4 . 6 6 )

A frica n A m erican s

A d ju sted O d d s R a tio s* - M L B W * ( 1 . 5 - 2 . 5 kg) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years


1. 43 ( 1 . 3 9 - 1 .4 8 )

1. 75 ( 1 . 6 6 - 1. 85)

1. 89 ( 1 . 6 7 - 2 . 1 5 )


1 . 2 7 ( 1 . 2 0 - 1. 35)

1 . 7 7 ( 1 . 6 1 - 1 .9 6 )

1.60(1.24-2.06)


1 .3 8 (1 .3 6 - 1 .4 0 )

1. 77 ( 1. 73 - 1 .8 2 )

2.14(2.03 -2.26)

P u erto R ica n s

1 . 3 2 ( 1 . 1 7 - 1. 49)

1 . 4 0 ( 1 . 1 4 - 1. 73)

2.64(1.81 -3.87)

A d ju sted O d d s R a tio s* - L B W * ( < 2 .5 k g) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years

.A frican A m erica n s

1 . 4 6 ( 1 . 4 2 - 1 .5 0 )

1. 82 ( 1 . 7 4 - 1. 91)

1.97(1.76-2.20)


1.31 ( 1 . 2 4 - 1. 39)

1.87(1.71 - 2 . 0 5 )

1.72(1.37-2.17)


1 . 3 9 ( 1 . 3 7 - 1. 41)

1.82 ( 1. 78 - 1. 86)

2.20(2.09-2.32)

Puerto R ica n s

1.31(1.17-1.46)

1. 57 ( 1. 31 - 1. 88)

2.58 (1.80 - 3 . 6 9 )

A d ju sted o d d s ratios ob tain ed for m oth ers > 3 0 years o f age in ea ch racial/eth n ic group com pared to th e m oth ers in th e sa m e ra cia l/eth n ic group 2 0 - 2 9 years o f age; separate lo g istic m o d els w h ich in clu d ed m aternal age, ed u ca tion , m arital status, prenatal care and sm o k in g w ere estim ated for each racial/eth n ic group. * V L B W = v ery lo w birth w eig h t; M L B W = m oderately lo w birth w eigh t; L B W = lo w birth w eig h t. ** 95 % c o n fid en ce interval

230


T ab le 6 .4 (d )-L o g istic regression a n a ly sis o f th e a sso c ia tio n b etw een d elayed ch ild b earin g and risk o f lo w (< 2 .5 k g ) birth w eig h t a m o n g four racial/eth n ic grou p s-P arity > 1-U n ited States, 19 8 9 - 1 9 9 1 A d ju sted O d d s R a tio s* - V L B W * ( < 1.5 k g) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years

1 . 3 2 ( 1 . 2 8 - 1. 37)**

1.43 ( 1 . 3 6 - 1. 50)

1 . 5 0 ( 1 . 3 5 - 1. 68)


1.51 ( 1 . 4 0 - 1. 63)

1. 87 (1 .6 9 - 2 .0 7 )

2 .0 9 ( 1. 73 - 2. 5 3 )


1.15 ( 1 . 1 2 - 1 . 1 9 )

1. 54 (1 .4 7 - 1 .60)

1.97(1.81 - 2 . 1 4 )

P uerto R ica n s

1 . 4 8 ( 1 . 2 7 - 1. 73)

1.83 (1 .4 8 - 2 . 2 5 )

2.09(1.42-3.07)


A d ju sted O d d s R a tio s* - M L B W * ( 1 . 5 - 2 . 5 kg) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years


1 . 2 6 ( 1 . 2 4 - 1 .2 8 )

1 . 3 6 ( 1 . 3 3 - 1 .4 0 )

1.45 ( 1. 38 - 1. 54)


1. 17 ( 1. 13 - 1 . 2 1 )

1 .4 6 ( 1 . 3 9 - 1. 53)

1. 82 ( 1 . 6 6 - 1. 98)


1. 12 ( 1 . 1 0 - 1 . 1 3 )

1 . 4 0 ( 1 . 3 8 - 1 .43)

1. 78 ( 1 . 7 2 - 1. 85)

Puerto R ica n s

1 . 2 4 ( 1 . 1 5 - 1. 33)

1 . 3 0 ( 1 . 1 7 - 1 .4 4 )

1 . 5 4 ( 1 . 2 7 - 1. 88)


3 5 -3 9 years

> 4 0 years


1 . 2 8 ( 1 . 2 6 - 1. 30)

1 . 3 9 ( 1 . 3 5 - 1 .42)

1 . 4 8 ( 1 . 4 1 - 1. 56)


1.2 2 ( 1 . 1 8 - 1 .2 6 )

1. 52 ( 1 . 4 6 - 1. 59)

1.87(1.72-2.02)


1.12(1.11 - 1 . 1 4 )

1.43 ( 1 . 4 0 - 1. 45)

1 . 8 2 ( 1 . 7 6 - 1. 89)

P uerto R ica n s

1 . 2 7 ( 1 . 1 9 - 1. 36)

1.38 ( 1 . 2 6 - 1. 52)

1.65 ( 1 . 3 8 - 1. 97)

A d ju sted o d d s ratios ob tain ed for m oth ers > 3 0 years o f a g e in ea ch racial/eth n ic group com pared to th e m oth ers in the sa m e ra cial/eth n ic group 2 0 - 2 9 years o f age; separate lo g istic m o d els w h ich in clu d ed m aternal ag e, ed u ca tio n , m arital status, prenatal care and sm o k in g w ere estim ated for each racial/eth n ic group. * V L B W = very lo w birth w eig h t; M L B W = m oderately lo w birth w eigh t; L B W = lo w birth w eig h t. ** 9 5 % c o n fid en ce interval

Table 6.4(e)-Logistic regression analysis o f the association between delayed childbearing and risk o f A d ju sted O d d s R a tio s

- V L B W * ( < 1.5 k g )

3 0 -3 4 years

3 5 - 3 9 years

> 4 0 years

1 . 5 2 ( 1 . 4 4 - 1. 61)**

1.96(1.80-2.14)

2.12(1.72-2.62)


1 . 5 4 ( 1 . 3 4 - 1 .7 7 )

2.36(1.92-2.91)

2.46(1.50-4.05)


1 . 3 9 ( 1 . 3 4 - 1 .4 4 )

2. 01 (1. 91 - 2 . 1 1 )

2.50(2.22-2.80)

P u erto R ica n s

1 .2 2 (0 .9 3 - 1 .6 0 )

2.42(1.71 -3.43)

1. 96 ( 0 . 8 0 - 4 . 7 9 )


A d ju sted O d d s R a tio s9 - M L B W * ( 1 . 5 - 2. 5 kg) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years


1 . 4 2 ( 1 . 3 7 - 1. 46)

1 . 7 4 ( 1 . 6 5 - 1. 83)

1.90(1.67-2.15)


1 . 2 4 ( 1 . 1 7 - 1. 32)

1 . 7 2 ( 1 . 5 6 - 1 .9 0 )

1.56(1.21-2.01)


1 .3 5 ( 1. 3 3 - 1 .3 7 )

1.75 (1.71 - 1 .7 9 )

2.14(2.03-2.26)

Puerto R ica n s

1.31(1.16-1.47)

1 . 4 0 ( 1 . 1 4 - 1 .7 2 )

2.64(1.81 -3.87)

A d ju sted O d d s R a tio s9 - L B W * ( < 2 .5 k g) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years


1 . 4 5 ( 1 . 4 1 - 1.49)

1.81 ( 1. 73 - 1 .90)

1.98(1.77-2.22)


1 . 2 8 ( 1 . 2 1 - 1. 36)

1. 82 ( 1 . 6 6 - 1 .9 9 )

1.69(1.34-2.12)


1 . 3 6 ( 1 . 3 4 - 1. 38)

1. 79 ( 1. 75 - 1. 83)

2.21 ( 2 . 1 0 - 2 . 3 2 )

P uerto R ica n s

1 . 3 0 ( 1 . 1 6 - 1. 45)

1 . 5 7 ( 1 . 3 1 - 1. 88)

2.59(1.81 -3.70)

to th e m others in th e sa m e racial/eth n ic group 2 0 - 2 9 years o f age; separate lo g istic m o d els w h ich in clu d ed m aternal a g e, ed u ca tio n , m arital status and sm o k in g w ere estim ated for each eth n ic group. * V L B W = v e ry lo w birth w eigh t; M L B W = m od erately lo w birth w eigh t; L B W = lo w birth w eig h t ** 9 5 % c o n fid e n c e interval


T ab le 6 .4 (f)-L o g istic regression a n a ly sis o f th e a sso cia tio n b etw een d ela y ed ch ild b earin g and risk o f lo w (< 2 .5 k g) birth w eig h t-P a rity > 1-U n ited S tates, 1 9 8 9 - 1 9 9 1 [prenatal care n ot controlled]________ A d ju sted O dds R a tio s* - V L B W * ( < 1.5 k g) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years

A frican A m erican s

1 . 3 0 ( 1 . 2 6 - 1. 35)

1 . 4 0 ( 1 . 3 3 - 1. 47)

1.48 ( 1. 33 - 1. 65)


1 . 5 0 ( 1 . 3 9 - 1 .6 2 )

1.86(1.68-2.06)

2 .0 8 ( 1 . 7 2 - 2 . 5 2 )


1.13 ( 1 . 1 0 - 1 . 1 7 )

1. 52 ( 1. 45 - 1. 58)

1.96 (1.81 - 2 . 1 3 )

Puerto R ica n s

1 . 4 6 ( 1 . 2 5 - 1 .7 0 )

1.79(1.45 -2 .2 0 )

2.04(1.39-2.99)

A d ju sted O d d s R a tio s* - M L B W * ( 1 . 5 - 2. 5 kg) 3 0 -3 4 years

3 5 -3 9 years

> 4 0 years


1.23 (1. 21 - 1. 25)

1.33 (1 .2 9 - 1. 36)

1.42 ( 1. 35 - 1. 51)

M ex ica n -A m erica n s N o n -H isp a n ic W h ites

1.14(1.10-1.18) 1.0 9 ( 1 . 0 7 - 1. 10)

1 . 4 2 ( 1 . 3 6 - 1 .4 9 ) 1 . 3 7 ( 1 . 3 5 - 1 .4 0 )

1.78 ( 1. 63 - 1 .94) 1 . 7 7 ( 1 . 7 1 - 1. 84)

Puerto R ica n s

1.19(1.11-1.28)

1.25 ( 1. 12 - 1. 38)

1 . 4 6 ( 1 . 2 0 - 1. 78)


3 5 -3 9 years

> 4 0 years


1.25 ( 1. 23 - 1 .2 7 )

1.35 ( 1. 32 - 1. 38)

1.45 ( 1 . 3 8 - 1. 52)


1.19(1.16-1.23)

1 . 4 9 ( 1 . 4 3 - 1. 56)

1.83 ( 1 . 6 9 - 1. 98)


1.10(1.08-1.11)

1.4 0 ( 1 . 3 7 - 1 .4 2 )

1.81 ( 1. 75 - 1. 87)

Puerto R ica n s

1.23(1.15-1.31)

1.33 (1.21 - 1 .4 6 )

1 . 5 7 ( 1 . 3 1 - 1.87)

A d ju sted od d s ratios ob tain ed for m oth ers > 3 0 years o f a g e in ea ch racial/ethn ic group com pared to th e m others in th e sa m e ra cial/eth n ic group 2 0 - 2 9 years o f age; separate lo g istic m o d els w h ich in clu d ed m aternal a g e, ed u ca tio n , m arital status and sm o k in g w ere estim ated for each eth n ic group. * V L B W = very lo w birth w eigh t; M L B W = m oderately lo w birth w eigh t; L B W = lo w birth w eig h t * * 9 5 % co n fid e n c e interval


Table 6.5 (a ) - Odds ratios^

for the association o f delayed childbearing ( maternal age > 35 years) with

preterm (< 37 w eeks) delivery am ong four racial / ethnic groups - First births - United States, 1989 - 1991

R acial/Ethnic Group V-Preterm* (< 32 w eeks) M-Preterm* (3 2 -3 7 w eek s) Preterm* (< 37 w eeks) A frican A m ericans

1.64 (1. 50 - 1.79)**

1.37 (1.30 - 1.45)

1.45 (1. 38 - 1.52)


2.00(1.63 -2.45)

1 . 5 0 ( 1 . 3 6 - 1.66)

1.58 ( 1 . 4 4 - 1.73)

N on-H isp anic W hites

1 . 6 4 ( 1 . 5 7 - 1.72)

1.38 (1.36 - 1.41)

1 . 4 2 ( 1 . 4 0 - 1.45)

Puerto R icans

1. 92( 1. 21 - 3 . 0 7 )

1.52 (1.18 - 1.95)

1.60(1.27-2.00)

A ll"

1 . 6 6 ( 1 . 5 9 - 1.72)

1 . 3 9 ( 1 . 3 6 - 1.41)

1.43 (1.41 - 1.45)

* R eference group for calculation o f odds ratios; mothers 20 - 34 years o f age ’* V-Preterm = very preterm delivery; M-Preterm = moderately preterm delivery ** 95 % con fid en ce interval "

The odds ratios for A ll are the M antel-H aenszel estim ates o f the com bined odds ratios for all

four ethnic groups.

234


Table 6.5 (b ) - O dds ratios^" for the association o f delayed childbearing ( maternal age > 35 years) with preterm (< 3 7 w eek s) delivery am ong four racial / ethnic groups - Parity > 1 - U nited States, 1989 - 1991

R acial/E thnic Group V-Preterm* ( < 32 w eek s) M-Preterm* (3 2 -3 7 w eek s) Preterm* ( < 37 w eek s) A frican A m ericans

1. 0 4 ( 1 . 0 1 - 1.08)

1 . 0 4 ( 1 . 0 2 - 1.06)

1 . 0 4 ( 1 . 0 3 - 1.06)


1 . 3 7 ( 1 . 2 8 - 1.46)

1 . 2 5 ( 1 . 2 2 - 1.29)

1 . 2 7 ( 1 . 2 4 - 1.30)

N on-H isp anic W hites

1.12(1.09-1.15)

1.11(1.09-1.12)

1.11 ( 1 . 1 0 - 1 . 1 2 )

Puerto R icans

0 .6 7 (0 .0 0 - 4 .0 5 )

1.18(0.57-2.41)

1.10(0.55-2.18)

All”

1.11(1.09-1.13)

1.11(1.10-1.12)

1.11(1.10-1.12)

* Reference group for calculation o f odds ratios: mothers 20 - 34 years of age * V-Preterm = very preterm delivery; M-Preterm = m oderately preterm delivery ** 95 % con fid en ce interval ”

The odds ratios for A ll are the M antel-H aenszel estim ates o f the com bined odds ratios for all

four ethnic groups.

235


T ab le 6 .5 (c ) - R isk d ifferen ces *

for the asso cia tio n o f d elayed childbearing ( maternal age > 35 years)

w ith preterm (< 3 7 w eek s) d elivery am o n g four racial / eth n ic grou p s - First B irths - U nited States, 1 9 8 9 - 1991

V -Preterm * ( < 32 w eek s)

M -Preterm * (3 2 -3 7 w eek s)

Preterm* ( < 3 7 w eek s)

A frican A m ericans

1.8 ( 1 . 4 - 2 . 2 )

3. 4 ( 2 . 7 - 4 . 1 )

4 .9 (4.1 - 5 . 6 )

M ex ica n s

1.1 ( 0 . 7 - 1 . 6 )

3.5 ( 2 . 5 - 4 . 4 )

4. 4 ( 3 . 4 - 5 . 5 )


0 .6 (0 .5 - 0 .7 )

2 .2 (2.1 - 2 . 4 )

2 .7 (2 .6 - 2 .9 )

Puerto R ican s

1.5 ( 0 . 0 8 - 2 . 9 )

3. 9 ( 1 . 2 - 6 . 7 )

5. 2 (2.3 - 8 . 1 )

R acial/E th n ic Group

* E xpressed per 100 com pared w ith th e reference group o f m others 2 0 - 34 years o f age in th e sam e eth n ic group. * V-Preterm = very preterm delivery; M -Preterm = m oderately preterm delivery ** 9 5 % c o n fid en ce interval

236


Table 6.5(d) - Risk differences * for the association o f delayed childbearing ( maternal age > 35 years) with preterm ( < 3 7 weeks) delivery among four racial / ethnic groups - Parity > 1 - United States, 19 8 9 - 1991

Racial/Ethnic Group V-Preterm* ( < 32 weeks)

M-Preterm* (32-37 weeks) Preterm* ( < 37 weeks)

African Americans

0.2 ( 0 .0 - 0 .3 )

0.5 (0.3 - 0.8)

0.6 ( 0 .4 - 0 .9 )

Mexicans

0.4 (0.3 - 0.5)

1.9 (1.6 -2 .1 )

2.2 (2.0 - 2.5)

Non-Hispanic Whites

0.1 ( 0 .0 - 0 .1 )

0.6 (0.5 - 0.7)

0.7 (0.6 - 0.8)

-0.6 (- 3 .4 -2 .1 )

1.6 (-6 .0 - 9 .3 )

1.1 ( -6 .8 -8 .9 )

Puerto Ricans

* Expressed per 100 compared with the reference group of mothers 20 - 34 years o f age in the same ethnic group. * V-Preterm = very preterm delivery; M-Preterm = moderately preterm delivery ** 95 % confidence interval

237


T ab le 6 .5 (e ) - O d d s ratios*" for th e a sso c ia tio n o f d elayed ch ild b earin g w ith preterm ( < 3 7 w eek s) d eliv ery a m o n g four racial / eth n ic grou p s - First B irths - U n ited States, 1 9 8 9 - 1991 V -Preterm * ( < 32 w eek s) A frican A m erica n s M ex ica n -A m erica n s N o n -H isp a n ic W h ites Puerto R ica n s

30-34 years 1.29 (1.23 - 1.35)** 1.27(1.13 - 1.42) 1.13 (1.10 -1 .1 7 ) 1 .2 4 (1 .0 0 - 1.52)

35-39 years 1.63 (1.51 - 1.76) 1 .8 7 (1 .5 7 -2 .2 3 ) 1 .5 7 (1 .5 0 - 1.64) 1 .9 0 (1 .4 0 -2 .5 8 )

> 40 years 1.75 (1.45 -2 .1 1 ) 2.03 (1 .3 4 -3 .0 7 ) 1.95 (1 .7 6 -2 .1 7 ) 2 .5 7 (1 .3 6 - 4 .8 7 )

M -Preterm * (32-37 w eek s) A frican A m erica n s M ex ica n -A m erica n s N o n -H isp a n ic W h ites P u erto R ica n s

30-34 years 1.14(1.11 -1 .1 8 ) 1 .1 1 (1 .0 6 -1 .1 7 ) 1 .1 3 (1 .1 2 -1 .1 4 ) 1 .1 0 (0 .9 9 - 1.22)

35-39 years 1 .3 6 (1 .2 9 - 1.43) 1 .5 0 (1 .3 8 - 1.63) 1 .3 8 (1 .3 6 - 1.41) 1.37(1.15 - 1.63)

> 40 years

1 .4 0 (1 .2 4 - 1.59) 1.57 (1.28 - 1.91) 1 .5 2 (1 .4 4 - 1.59) 1.54(1.03 -2 .3 1 )

Preterm ( < 37 w eek s)

African Americans M ex ica n -A m erica n s N o n -H isp a n ic W h ites Puerto R ica n s

30-34 years 1 .1 8 (1 .1 5 -1 .2 1 ) 1 .1 4 (1 .0 8 -1 .1 9 ) 1.13 (1.12 -1 .14) 1 .1 2 (1 .0 2 - 1.24)

35-39 years 1 .4 4 (1 .3 8 - 1.50) 1.56 (1.45 - 1.69) 1.41 (1 .3 9 - 1.44) 1 .4 7 (1 .2 6 - 1.72)

> 40 years 1.50 (1.35 - 1.67) 1.64 (1 .3 6 - 1.97) 1.58 (1.51 - 1.66) 1 .7 4 (1 .2 2 -2 .4 9 )

R eferen ce group for ca lcu la tio n o f o d d s ratios: m others 2 0 - 2 9 years o f age. * V -Preterm = v ery preterm d elivery; M -Preterm = m od erately preterm d eliv ery * * 9 5 % c o n fid e n c e interval 238


T ab le 6 .5 (f) - O d d s ratios*

for th e a sso cia tio n o f d elayed ch ild b earin g w ith preterm (< 3 7 w eek s)

d e liv e iy a m o n g four racial / eth n ic grou p s - Parity > 1 - U nited States, 1 9 8 9 - 1991________________ V -Preterm * ( < 32 w eek s) A frican A m erica n s M ex ica n -A m erica n s N o n -H isp a n ic W h ites Puerto R ica n s

30-34 years 1.03(1.01 - 1.06) 1.04 (0 .9 9 - 1.10) 0.80 (0.78 - 0.82) 1.07 (0 .9 6 - 1.20)

35-39 years 1.04 (1.01 - 1.08) 1.33 (1.23 - 1.43) 1.01 (0 .9 8 - 1.04) 1 .2 0 (1 .0 2 - 1.42)

> 40 years 1.10(1.01 -1.20) 1 .6 5 (1 .4 4 - 1.89) 1.27 (1.19 - 1.36) 1.71 (1 .2 8 -2 .3 0 )

M -Preterm * (32-37 w eek s) A frican A m erica n s M ex ica n -A m erica n s N o n -H isp a n ic W h ites Puerto R ica n s

30-34 years 0.96 (0.95 - 0.98) 1.01 (0 .9 9 - 1.03) 0.87 (0.86 - 0.88) 1.00 (0.95 - 1.06)

35-39 years 1 .0 2 (0 .9 9 - 1.04) 1 .2 1 (1 .1 7 -1 .2 4 ) 1.02 (1.01 - 1.03) 1 .1 9 (1 .1 0 -1 .2 9 )

> 40 years

1 .1 5 (1 .1 0 -1 .2 1 ) 1.50(1.41 - 1.59) 1 .2 7 (1 .2 4 - 1.31) 1 .4 9 (1 .2 8 - 1.73)

Preterm ( < 37 w eek s) A frican A m erica n s M ex ica n -A m erica n s N o n -H isp a n ic W h ites Puerto R ica n s

30-34 years 0.98 (0.97 - 0.99) 1.02 (0 .9 9 - 1.04) 0.86 (0.85 - 0.87) 1.01 (0 .9 6 - 1.06)

35-39 years 1.02 (1 .0 0 - 1.04) 1.2 2 (1 .1 9 - 1.26) 1.02(1.01 - 1.03) 1 .1 9 (1 .1 1 -1 .2 8 )

> 40 years

1 .1 5 (1 .1 0 -1 .2 0 ) 1 .5 2 (1 .4 4 - 1.61) 1.28 (1.24 - 1.31) 1.5 4 (1 .3 4 - 1.77)

* R eferen ce group for ca lcu la tio n o f od d s ratios: m others 2 0 - 2 9 years o f age. * V -Preterm = v ery preterm d elivery; M -Preterm = m od erately preterm d elivery * * 9 5 % co n fid e n c e interval

239

Table 6.6(a) - Attributable fractions o f preterm delivery among the exposed (AFe) and attributable fractions in the population(AFp) associated with delayed childbearing ( > 35 y ea rs) among four racial/ethnic groups First births - United States, 1989 - 1991

V-Preterm *(< 32 weeks)

M-Preterm*(32-37 weeks)

Preterm* ( < 37 weeks)

APjT

AFe*

AFp™

AFfl*

AFp™

0.39

0.03

0.27

0.02

0.31

0.02

0.50

0.02

0.34

0.01

0.37

0,01

Non-Hispanic Whites

0.39

0.04

0.28

0.02

0.30

0.03

Puerto Ricans

0.48

0.03

0.34

0.01

0.37

0.02

Racial/Ethnic Group

af/

African Americans M exican Americans

* V-Preterm = very preterm delivery; M-Preterm = moderately preterm delivery ÂFe = (RR - 1 ) / RR where RR is the estimated relative risk as approximated by the odds ratio. ^ A F p = Pc (RR - 1) / RR = Pc AFe where Pc is the exposure rate among cases.


Table 6.6(b) - Attributable fractions o f preterm delivery among the exposed (AFe) and attributable fractions in the population(AFp) associated with delayed childbearing ( > 35 years) among four racial/ethnic groups Parity > 1 - United States, 1989 - 1991___________________________________________________________________




Racial/Ethnic Group

AFe*

AFp™

AF/

AFp™

AFe*

AFp”

African Americans

0.04

0.00

0.04

0.00

0.04

0.00

Mexican Americans

0.27

0.04

0.20

0.03

0.21

0.03

Non-Hispanic Whites

0.11

0.02

0.10

0.01

0.10

0.01

Puerto Ricans

Of*

Qf f

0.01***

0.09

0.01

0

1

5

***

* V-Preterm = very preterm delivery; M-Preterm = moderately preterm delivery 9AFe = (RR - 1 ) / RR where RR is the estimated relative risk as approximated by the odds ratio. ^ A F p = Pc (RR - 1 ) / RR = Pc AFe where Pc is the exposure rate among cases. ** point estimates for the odds ratios less than one and not significant at alpha=0.05. *** point estimates for the odds ratios not significant at alpha=0.05.

241

Table 6.6(c) - Attributable fractions o f preterm delivery in the population(AFp) associated with delayed childbearing ( > 35 y ea rs) among four racial/ethnic groups - All Births - United States, 1989 - 1991




AFp*

AFp*

AFp^

African Americans

0.03

0.02

0.02

M exican Americans

0.06

0.04

0.04

Non-Hispanic Whites

0.06

0.03

0.04

Puerto Ricans

0.03

0.02

0.03

Racial/Ethnic Group

* V-Preterm = very preterm delivery; M-Preterm = moderately preterm delivery ÂFp = Pc (RR - 1 ) / RR = Pc AFe where Pc is the exposure rate among cases.


Table 6.7(a) - Logistic regression analysis o f the association between delayed childbearing (> 35 years ) and risk o f preterm (< 37 w eeks) delivery among four racial/ethnic groups - First births - United States, 1989 - 1991

Adjusted Odds Ratios ^ (95 % confidence intervals)

Racial/Ethnic Group

V-Preterm* ( < 32 w eeks)

M-Preterm* (32-37 w eeks)

Preterm* ( < 37 w eeks)

African Americans

1.71 (1 .5 6 - 1.87)

1.45 (1 .3 7 - 1.53)

1.52 (1 .4 5 - 1.60)

M exicans

2.03 ( 1 .6 5 - 2 .4 9 )

1.55 (1 .4 0 - 1.71)

1.63 (1.48 - 1.78)

Non-Hispanic Whites

1 . 7 8 ( 1 . 7 0 - 1.86)

1.45 ( 1 . 4 2 - 1.48)

1 . 5 0 ( 1 . 4 7 - 1.53)

Puerto Ricans

1.89(1.18-3.03)

1 . 5 2 ( 1 . 1 9 - 1.96)

1.60(1.28-2.01)

* Adjusted odds ratios obtained

for mothers > 35 years o f age in each racial/ethnic group compared to the

mothers in the same racial/ethnic group 20 - 34 year o f age - separate logistic models which included maternal age, education, marital status, prenatal care and smoking were estimated for each o f the racial/ethnic groups. * V-Preterm = very preterm delivery; M-Preterm = moderately preterm delivery

243


Table 6.7(b) - Logistic regression analysis o f the association between delayed childbearing (> 35 y e a r s) and risk o f preterm (< 37 w eeks) delivery among four racial/ethnic groups - Parity > 1 - United States, 1989 - 1991

Adjusted Odds Ratios

9

(95

% confidence

intervals)

V-Preterm* ( < 32 w eeks)

M-Preterm* (32-37 w eeks)

Preterm* ( < 37 w eeks)

African Americans

1.16(1.11-1.22)

1.16(1.13-1.19)

1.16(1.14-1.19)

M exican Americans

1.44(1.33 - 1.55)

1.29(1.25 - 1.33)

1.31 ( 1 . 2 7 - 1.35)

Non-H ispanic Whites

1.35 ( 1 . 3 0 - 1.39)

1.25 (1.23 - 1.27)

1 . 2 6 ( 1 . 2 5 - 1.28)

Puerto Ricans

1.24 ( 0 . 9 7 - 1.58)

1.36(1.23 - 1.52)

1.35 (1.23 - 1.49)

Racial/Ethnic Group

* Adjusted

odds ratios obtained for mothers > 35 years o f age in each racial/ethnic group compared to the

mothers in the same racial/ethnic group 20 - 34 year o f age - separate logistic models which included maternal age, education, marital status, prenatal care and smoking were estimated for each o f the racial/ethnic groups. * V-Preterm = very preterm delivery; M-Preterm = moderately preterm delivery

to *


Table 6.7(c) - Logistic regression analysis o f the association between delayed childbearing and risk o f preterm (< 37 weeks) delivery among four racial/ethnic groups - First births - United States, 1989 - 1991 V-Preterm* ( < 32 weeks) 30-34 years

35-39 years

> 40 years

1 .4 3 (1 .3 7 -1 .5 0 )* *

1.83 (1 .6 7 - 1.97)

1.87(1.55 - 2.26)

Mexican-Americans

1 .3 4 (1 .2 0 -1 .5 0 )

1 .9 7 (1 .6 5 -2 .3 4 )

2 .0 6 (1 ,3 6 -3 .1 2 )

Non-Hispanic Whites

1.33 (1 .2 9 - 1.37)

1 .8 2 (1 .7 4 - 1.91)

2 .1 5 (1 .9 4 -2 .3 8 )

Puerto Ricans

1 .3 6 (1 .1 0 - 1.68)

1 .9 7 (1 .4 4 -2 .6 8 )

2 .5 6 (1 .3 5 -4 .8 6 )

African Americans

M-Preterm* (32-37 weeks) 30-34 years

35-39 years

> 40 years

African Americans

1.27 (1.23 -1 .3 1 )

1.51 (1 .4 4 - 1.59)

1.51 (1 .3 4 -1 .7 1 )

Mexican-Americans

1.20(1.14- 1.26)

1.61 (1 .4 8 -1 .7 5 )

1 .6 2 (1 .3 2 -1 .9 8 )

Non-Hispanic Whites

1.23 (1 .2 2 -1 .2 5 )

1 .5 0 (1 .4 7 - 1.53)

1.61 (1 .5 3 -1 .6 9 )

Puerto Ricans

1 .1 7 (1 .0 5 -1 .3 1 )

1.44(1.21 - 1.72)

1 .5 6 (1 .0 4 -2 .3 4 )

Preterm ( < 37 weeks) 30-34 years

35-39 years

> 40 years

African Americans

1.31 (1 .2 8 - 1.35)

1.61 (1 .5 4 -1 .6 8 )

1 .6 2 (1 .4 5 - 1.81)

Mexican-Americans

1 .2 2 (1 .1 7 -1 .2 8 )

1 .6 7 (1 .5 5 - 1.80)

1.70(1.41-2.04)

Non-Hispanic Whites

1.25 (1.23 -1 .2 6 )

1.55 (1.53 - 1.58)

1 .6 9 (1 .6 2 - 1.77)

Puerto Ricans

1.21 (1 .1 0 -1 .3 3 )

1 .5 4 (1 .3 2 -1 .8 1 )

1.76(1.23 - 2.52)

* Adjusted odds ratios obtained for mothers > 30 years o f age in each racial/ethnic group compared to the mothers in the same racial/ethnic group 20 - 29 years o f age; separate logistic models which included maternal age, education, marital status, prenatal care and smoking were estimated for each racial/ethnic group. * VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight. * * 9 5 % confidence interval

245


Table 6.7(d) - Logistic regression analysis o f the association between delayed childbearing and risk o f preterm (< 37 weeks) delivery among four racial/ethnic groups - Parity > 1 - United States, 1989 -1991__________________________________________________________________ V-Preterm* ( < 32 weeks) 30-34 years

35-39 years

> 40 years

African Americans

1 ,1 6 (1 .1 3 -1 .1 9 )

1 .2 0 (1 .1 6 -1 .2 5 )

1 .2 7 (1 .1 7 -1 .3 8 )

Mexican-Americans

1.11 (1 .0 4 -1 .1 7 )

1 .3 9 (1 .2 9 - 1.50)

1 .7 0 (1 .4 8 -1 .9 5 )

Non-Hispanic Whites

1.03 (1.01 -1 .0 6 )

1,31 (1 .2 7 -1 .3 6 )

1 .5 5 (1 .4 5 - 1.66)

Puerto Ricans

1 .1 9 (1 .0 6 - 1.34)

1 .3 1 (1 .1 0 -1 .5 5 )

1 .7 4 (1 .2 9 -2 .3 5 )


35-39 years

> 40 years

African Americans

1 .0 7 (1 .0 6 -1 .0 9 )

1 .1 5 (1 .1 3 -1 .1 7 )

1.3 0 (1 .2 4 - 1.36)

Mexican-Americans

1 .0 6 (1 .0 4 -1 .0 9 )

1 .2 5 (1 .2 2 - 1.30)

1 .5 2 (1 .4 4 - 1.62)

Non-Hispanic Whites

1 .0 3 (1 .0 2 -1 .0 4 )

1.22(1.21 -1 .2 4 )

1.47(1.43 - 1.51)

Puerto Ricans

1 .1 0 (1 .0 4 -1 .1 6 )

1 .2 9 (1 .1 9 -1 .4 1 )

1.5 9 (1 .3 6 - 1.85)


35-39 years

> 40 years

African Americans

1 .0 9 (1 .0 8 - 1.11)

1 .1 7 (1 .1 5 -1 .1 9 )

1 .3 0 (1 .2 5 - 1.36)

Mexican-Americans

1 .0 7 (1 .0 4 -1 .0 9 )

1 .2 8 (1 .2 4 - 1.31)

1 .5 6 (1 .4 7 - 1.65)

Non-Hispanic Whites

1 .0 3 (1 .0 2 - 1.04)

1 .2 4 (1 .2 2 - 1.25)

1 .4 9 (1 .4 5 - 1.53)

Puerto Ricans

1 .1 2 (1 .0 6 -1 .1 8 )

1.30(1.21 -1 .4 1 )

1.63 (1 .4 2 - 1.88)

* Adjusted odds ratios obtained for mothers > 30 years o f age in each racial/ethnic group compared to the mothers in the same racial/ethnic group 20 - 29 years o f age; separate logistic models which included maternal age, education, marital status, prenatal care and smoking were estimated for each racial/ethnic group. * VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight. * * 9 5 % confidence interval

246


Table 6.7(e) - Logistic regression analysis of the association between delayed childbearing and risk o f preterm (< 37 weeks) delivery among four racial/ethnic groups - First births - United States, 1989- 1991 [prenatal care not controlled] V-Preterm* ( < 32 weeks) 30-34 years

35-39 years

> 40 years

African Americans

1.43(1.36- 1.50)**

1.82(1.69- 1.97)

1.89(1.56 - 2.28)

Mexican-Americans

1.32(1.18- 1.48)

1.94(1.63-2.31)

2.04(1.35-3.09)

Non-Hispanic Whites

1.30(1.26- 1.34)

1.80(1.72- 1.88)

2.18(1.97-2.42)

Puerto Ricans

1.36(1.10- 1.68)

2.01 (1.48-2.74)

2.62(1.38-4.96)


35-39 years

> 40 years

African Americans

1.25 (1.22- 1.29)

1.49(1.42- 1.57)

1.51 (1.34-1.71)

Mexican-Americans

1.16(1.11 -1.22)

1.55 (1.43- 1.69)

1.58 (1.29- 1.93)

Non-Hispanic Whites

1.20(1.19- 1.22)

1.48 (1.45 - 1.51)

1.60(1.53 - 1.69)

Puerto Ricans

1.16(1.04- 1.29)

1.43 (1.20- 1.70)

1.56(1.04-2.34)


35-39 years

> 40 years

African Americans

1.30(1.27-1.33)

1.59(1.52- 1.66)

1.62(1.46- 1.81)

Mexican-Americans

1.19(1.13-1.24)

1.62 (1.50- 1.75)

1.65(1.37- 1.99)

Non-Hispanic Whites

1.22(1.21 - 1.23)

1.53 (1.50- 1.55)

1.69(1.62-1.77)

Puerto Ricans

1 .2 0 (1 .0 9 - 1.32)

1.54 (1 .3 2 - 1.81)

1 .7 7 (1 .2 4 - 2 .5 3 )

*

Adjusted odds ratios obtained for mothers > 30 years o f age in each racial/ethnic group compared to the mothers in the same racial/ethnic

group 20 - 29 years o f age; separate logistic models which Included maternal age, education, marital status and smoking were estimated for each racial/ethnic group. * VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight. ** 95 % confidence interval

247


Table 6.7(f) - Logistic regression analysis o f the association between delayed childbearing and risk o f preterm (< 37 weeks) delivery among

foun‘acial/ethnicjrouj)Si-J|aritj£ >J^^^ V-Preterm* ( < 32 weeks) 30-34 years

35-39 years

> 40 years

1.14(1.11 -1.17)**

1 .1 8 (1 .1 4 -1 .2 3 )

1.25 (1.15 - 1.36)

Mexican-Americans

1 .1 0 (1 .0 4 -1 .1 7 )

1 .3 9 (1 .2 9 -1 .5 0 )

1 .7 0 (1 .4 8 -1 .9 5 )

Non-Hispanic Whites

1.01 (0 .9 9 - 1.04)

1 .2 9 (1 .2 5 - 1.33)

1 .5 5 (1 .4 4 - 1.66)

Puerto Ricans

1.17 (1 .0 4 - 1.32)

1 .2 8 (1 .0 8 - 1.52)

1.71 (1 .2 7 -2 .3 0 )

African Americans


35-39 years

> 40 years

African Americans

1 .0 5 (1 .0 4 -1 .0 7 )

1 .1 3 (1 .1 0 -1 .1 5 )

1.27(1.21 -1 .3 4 )

Mexican-Americans

1 .0 4 (1 .0 2 -1 .0 7 )

1.23 (1 .1 9 - 1.27)

1.50(1,41 - 1.59)

Non-Hispanic Whites

1.00 (0 .9 9 -1 .0 1 )

1 .2 0 (1 .1 8 - 1.22)

1 .4 6 (1 .4 2 -1 .5 0 )

Puerto Ricans

1 .0 8 (1 .0 2 -1 .1 4 )

1 .2 6 (1 .1 6 - 1.37)

1.53 (1.31 - 1.79)


35-39 years

> 40 years

African Americans

1 .0 7 (1 .0 6 -1 .0 9 )

1 .1 4 (1 .1 2 -1 .1 6 )

1 .2 8 (1 .2 3 - 1.34)

Mexican-Americans

1.05 (1.03 -1 .0 7 )

1.25(1.21 - 1.29)

1.53(1.45- 1.62)

Non-Hispanic Whites

1.01 (1 .0 0 -1 .0 1 )

1.21 (1 .2 0 - 1.23)

1 .4 8 (1 .4 4 -1 .5 2 )

Puerto Ricans

1 .0 9 (1 .0 4 - 1.15)

1 .2 7 (1 .1 8 -1 .3 7 )

1 .5 8 (1 .3 7 -1 .8 2 )

* Adjusted odds ratios obtained for mothers > 30 years o f age in each racial/ethnic group compared to the mothers in the same racial/ethnic group 20 - 29 years o f age; separate logistic models which included maternal age, education, marital status and smoking were estimated for each racial/ethnic group. * VLBW = very low birth weight; MLBW = moderately low birth weight; LBW = low birth weight. ** 95 % confidence interval

248


Table 6.8 - Risk o f Down syndrome at birth by age o f the mother for three racial/ethnic groups* - United States, 1989 - 1991 (N = 9,603,060) African Americans

Mexican Americans

Non-Hispanic Whites

Maternal Age (years)

Odds Ratio*

95 % C.l.

Odds Ratio*

95 % C.l.

Odds Ratio*

9 5 % C.l.

< 15

1.7

0 .8 - 3 .4

1.6

0 .4 - 6 .5

0.9

0 .3 - 2 .9

1 5 -1 9

0.9

0 .7 - 1 .2

1.2

0 .8 - 1.6

0.9

0 .8 - 1.0

2 0 -2 4

0.9

0 .7 -1 .1

1.1

0 .9 - 1.5

0.9

0 .8 - 1.0

2 5 -2 9

1.0

Reference

1.0

Reference

1.0

Reference

3 0 -3 4

1.2

0 .9 - 1 .6

1.8

1 .3 -2 .4

1.5

1 .4 - 1.6

3 5 -3 9

3.9

2 .9 -5 .1

5.2

3.8 - 6.9

2.7

2 .5 - 3 .0

4 0 -4 4

11.6

8 .2 -1 6 .5

19.4

1 4 .1 -2 6 .7

8.5

7.5 - 9.7

4 5 -4 9

40.4

1 6 .4 -9 9 .5

52.3

25.2 - 108.4

22.5

14.8-34.1

* O dds ratios obtained from logistic models stratified by maternal race / ethnicity -the reference group for each racial/ethnic group were mothers 25-29 years o f age o f the same group. * African Americans - N = 1,627,808 Mexican Americans - N = 1,043,873 Non-Hispanic Whites - N = 6,886,379

249


Figure 6.1 - Risk for Down syndrome at birth by maternal age and race/ethnicity - United State, 1989- 1991 60

Likelihood Ratio Test, p < 0.0001

50

40

•8 TJ

30

o

20 ■*— Mexican Americans - • - African Americans

10

Non-Hispanic Whites

■■r: 13

17

22

27

32

37

42

47


250


Table 6.9 - Impact o f delayed childbearing on the risk for Down syndrome for three different racial/ethnic groups in the United States, 1989 - 1991 ________________________________________________________ ________ Maternal Age > 35 years

Maternal Age > 40 years

Maternal Race/Ethnicity

Odds Ratio*

PAR*

Odds Ratio**

PAR*

African Americans

5.2 (4.3-6.3)

19%

11.3 (8.4-15.1)

7%

Mexican Americans

6.5 (5.4-7.7)

28%

14.5 (11.3-18.5)

14%

Non-Hispanic Whites

3.2 (3.0-3.5)

17%

7.3 (6.5-8.1)

7%

♦Odds Ratio (95 % Confidence Interval) - reference group: Mothers < 35 years o f age **Odds Ratio (95 % Confidence Interval) - reference group: Mothers < 40 years o f age * Population Attributable Risk Percent


Table 6 . 1 0 - Odds o f amniocentesis use by age and race o f the mother - United States, 1989 - 1991

African Americans

Mexican Americans

Odds Ratio*

95 % C.l.

Odds Ratio*

95 % C.l.

40

0.43

0.40 - 0.45

0.26

0.24 - 0.28

A ll 9

0.60

0.60-0.61

0.50

0.50-0.51


* Reference group: Non-Hispanic White mothers

9

Mantel-Haenszel estimate o f the age-adjusted combined odds ratio o f amnicentesis use for African Americans

/ Mexican Americans as compared with Non-Hispanic Whites; test for heterogeneity o f odds ratios [significance o f maternal age/race interaction], p < 0.0001

252


Table 6.11 - Risk of Down syndrome at birth by age and education of the mother for African Americans and non-Hispanic whites United States, 1989 -19 9 1

African Americans M aternal Age

Education < 12 years

Non-Hispanic Whites

Education > 12 years

Education < 12 years

Education > 12 years

(years)

O dds Ratio*

9 5 % C .l.

O dds Ratio*

95 % C .l.

O dds Ratio*

95 % C .l.

O dds Ratio*

9 5 % C.l.

= 12 years 40 35 30 25

20 15 10 jr

5

0 13

17

22

27

32

37

42

47


254


Figure 6.3 - Risk for Down syndrome at birth by maternal age and education - African American - United State, 1989- 1991 Likelihood Ratio Test, p=0.68

—♦ ■Education < 12 years - « - Education >= 12 years

20

-

13

17

22

27

32

37


255


Table 6.12 - Odds o f amniocentesis use by maternal age and education for African Americans and non-Hispanic whites United States, 1989 - 1991 African Americans

non-Hispanic whites

Odds Ratio*

95 % C.l.

Odds Ratio*

95 % C.l.

40

0.40

0 .3 6 -0 .4 4

0.60

0.58 - 0.62

Al l *

0.65

0 .6 4 -0 .6 7

0.76

0 .7 6 -0 .7 7

M aternal Age (years)

* Reference group: M aternal education > 12 years * Mantel-Haenszel estimate o f the age-adjusted combined odds ratio o f amnicentesis use

256


Table 6.13(a) - Competing risks analysis o f the effect o f delayed childbearing (maternal age > 35 years) on cause-specific infant mortality - United States, 1990 (N = 1,823,401)#

Hazard Ratios *

Model 1 (age only)* (p-values**) Model 11 (+ SES)* p-values Model III (+ BW & SES)* p-values "

Accidents

Homicides

SIDS ™

0.31

0.28

0.62

(0.0002)

(0.0286)

(0.0001)

0.35

0.39

0.68

(0.0011)

(0.1135)

(0.0001)

0.34

0.39

0.55

(0.0010)

(0.1063)

(0.0001)

50 % random sample o f singleton births in the United States in 1990

f Hazard ratios estimated from Cox proportional hazards models assuming conditional independence among causes o f infant death - hazard ratios correspond to the effect o f maternal age > 35 years as compared with the reference group o f mothers 20-34 years o f age. Sudden Infant Death Syndrome * Model I estimates the crude effect o f age. Model II estimates the effect o f age adjusted for maternal education, marital status, race, parity, interpregnancy interval, prenatal care, tobacco and alcohol use. Model III estimates the effect o f age adjusted for birth weight, as well as, maternal education, marital status, race/ethnicity, parity, interpregnancy interval, prenatal care, tobacco and alcohol use. ** Wald chi-square p-values

257


Table 6.13(b) - Competing risks analysis o f the effect o f delayed childbearing (maternal age > 35 years) on cause-specific infant mortality - United States, 1990 (N = 1,823,401)#

Hazard Ratios *

Model 1 (age only)* (p-values**) Model II (+ SES)* p-values Model HI (+ BW & SES)* p-values

Congenital Anomalies

Complications o f Placenta

All Causes

1.14 (0.0278)

1.59 (0.0195)

0.93 (0.0429)

1.25 (0.0003) 1.01 (0.8810)

1.68 (0.0120) 1.28 (0.2255)

(0.2327) 0.88 (0.0004)

1.04

* 50 % random sample o f singleton births in the United States in 1990 A

Hazard ratios estimated from Cox proportional hazards models assuming conditional independence among causes o f infant death - hazard ratios correspond to the effect o f maternal age > 35 years as compared with the reference group o f mothers 20-34 years o f age. * Model 1 estimates the crude effect o f age. Model II estimates the effect o f age adjusted for maternal education, marital status, race, parity, interpregnancy interval, prenatal care, tobacco and alcohol use. Model 111 estimates the effect o f age adjusted for birth weight, as well as, maternal education, marital status, race/ethnicity, parity, interpregnancy interval, prenatal care, tobacco and alcohol use. ** Wald chi-square p-values

258