Longitudinal Effects of Age at Onset and First. Drinking Situations on Problem Drinking^. Lynn A. Warner, M.S.W., Ph.D.'* and Helene R. White, Ph.D.^. 'School of ...
SUBSTANCE USE & MISUSE Vol. 38. No. 14. pp. 1983-2016, 2003
Longitudinal Effects of Age at Onset and First Drinking Situations on Problem Drinking^ Lynn A. Warner, M.S.W., P h . D . ' * and Helene R. White, Ph.D.^ 'School of Social Work, and ^Center of Alcohol Studies, Rutgers University, New Brunswick. New Jersey. USA
ABSTRACT The purpose of this study was to describe aspects of the first alcoholusc experience, and examine the predictive relations among age of first use, context of alcohol use initiation, and problem drinking with and without controls for psychosocial risk factors. Data were from the Rutgers Health and Human Deveiopnient Project, a five-wave, prospective study of substance-use behaviors in a community sample. Respondents, who were first interviewed at age 12 (1979-81) and
^A previous version of this article was presented as a poster al the Research Society on Alcoholism Twenty-Fourth Anntia! Scientific Meeting (June 2001), Montreal. Canada. *Correspondence: Lynn A. Warner, M.S.W.. Ph.D., School of Social Work, Rutgers University, 536 George Street, New Brunswick, NJ 08901, USA; Fax: (732) 932-8181; E-mail: lywarner(flirci.rutgers,edu. 1983 DOI: 10.l08i/JA-120025t23 Copyright © 2003 by Marcel Dekker. Inc.
1082-6084 tPrint); 1532-2491 (Online) ; www.dekker.com
1984
Warner and White most recently at age ."^0 or 31 (1999-2000) {N = i7ll reported on their first drinking expericnee. and on a range of known risk factors for alcohoi abuse. Most alcohol initialion occurred during a family gathering. Regardless of initiation context, youth who drank at an early age were more likely than youth who initiated later to become problem drinkers, although the risk was relatively greater for the youth who first drank outside a family gathering, Ba.sed on multivariate logistic regressions, feeling drunk al initiation was the only onset-related variable significantly associated with problem drinking; other significant risks factors included male gender, delinquency, and family history of alcoholism. Because most initiation occurs at a family gathering, alcoholism prevention research may benefit from examining the role that drinking in family contexts eould play with regard to socializing young drinkers to less risky drinking behaviors in adulthood. In particular, further research focusing on the subjective effects experienced by youth when they first drink may be merited. Key Words: Drinking initiation Protective factors; Risk factors.
context;
First aleohol use;
INTRODUCTION Life course developmetit and developmental psychopathology theories increit.singly emphasize the importance of both timing and cotitext of childhood and adolescent experiences for normal development (Boyce et a)., 1998; Cicchetti atid Rogosch. 1999). To the extent that common experiences, such as first alcohol use. occur at ages that are not normative and within contexts that reinforce negative alcohol-use behaviors, maladaptive learning patterns are developed. As a result, unsuccessful transitions to adult roles and compromised physical and mental health are more likely outcomes. Researchers in the area of substance use have increasingly recommended the application of developmental theories and methodologies to the study of substance misuse (Cicchetti and Rogosch. 1999; Flay et al.. 1999; Petraitis et al.. 1995; Zucker, 1994). However, the evaluation of age at onset along with eontextual infltiences on the developtnent of problems related to alcohol use has not occurred. Several studies have demonstrated signiticant associations between age of first use and subsequent alcohol-related problems and disorder (DeWit et al.. 2000; Grant and Dawson. 1997; Guo et al.. 2000; Nelson et al.. 1998). although there is debate in the literature about the strength
Longitudinal Effects on Problem Drinking
1985
of the association when other factors are controlled, and the extent to which age at tirst use interacts with other predictors. In a birth cohort of New Zealand children, alcohol use before age 6 compared to after age 13 increased the likelihood of having drinking problems at age 15, even controlHng for parental alcohol use and other risks (Fergusson et al., 1994). Additionally. Hawkins and colleagues found that early use of aleohol almost entirely mediated other risk factors for alcohol misuse (Hawkins et al., 1997). On the other hand, a study that examined a smaller set of predictors found friends' drinking behavior was more important than youth's own early drinking in predicting alcohol misuse (Griffin et al.. 2000). McGue and colleagues' results indicate that early age at first drink may be a reflection of underlying pathological behavior that causes alcoholism, rather than a direct cause itself (McGuc ct al.. 2001). Similarly, based on analyses of data provided by Norwegian youth. Pedersen and Skrondal suggest that the role of age of onset may be overestimated whenever analyses do not account for common risk factors predicting both onset and consumption (Pedersen and Skrondal. 1998). Finally, results from the longitudinal Rutgers Health and Human Development Project have shown that age at first use of alcohol does not predict later intensity of alcohol use or alcohol problems when other substance use is included in the models (Labouvie et al.. 1997). Interpreting these results in the context of other studies of age at onset, Labouvie and White argue that age of onset may be a useful indicator of risk when the period of observation is limited to adolescence or young adulthood, but that over a longer period current behaviors and experiences assume significant relationships with alcohol misuse while age at onset recedes in importance (Labouvie and White, 2002). Thus, attention to developmental processes over a longer period of time may provide a more accurate understanding of the role of age at first use. When context effects are studied, the outcome of interest is usually current substance use or drinking patterns, rather than the development of harmful substance-use behavior. Assessing contextual influences on long-term, alcohol-use behaviors is further complicated by the different levels at which context can be measured, and the difficulty of adequately measuring and analyzing ail dimensions within a single study. In a comprehensive review of risk and protective factors for substance abuse. Hawkins., Catalano. and Miller identified studies that highlight proximal, or micro-level influences on youth's drinking, including peer, friend, and parent attitudes and drinking behaviors (Hawkins et al., 1992). Distal social influences include laws and norms, as well as effects that stem from economic deprivation and social disorganization at the neighborhood level. An increasing number of cross-sectional studies of alcohol
1986
Warner and White
consumption in nations other than the United States underscore the wide variability in the role alcohol plays in social life, drinking patterns, and the consequences of drinking that are considered to be problematic (Bennet et al.. 1998; Smart and Ogborne. 2000; Vega et al,. 2002). Furthermore, drinking patterns vary across cultural groups within countries. Although drinking behaviors and attitudes of persons who have influence on youth's development have been studied, it is interesting to note that we know very little about the contexts in which youths actually drink. A relatively small body of research has examined modeling effects when youth drink as part of religious observances within family settings, or as part of a cultural tradition. Bales was among the first to propose that this type of early drinking, because it is minimal and invested with a religious purpose, is not likely to increase risk for alcoholism (Bales, 1946). More recent research in this area has focused on religiosity and suggests that it may serve as a protective factor for youth at risk of drinking at early ages (Heath et al., 1999; Kendler et al.. 1999; Mason and Windle, 2002), although it is not clear if drinking at family gatherings was controlled in these studies. One study that directly examined youth's alcohol use at home found high probabilities of alcohol use for seventh grade students who had been allowed to drink at home when they were fifth graders (Jackson et al.. 1999), However, data were not available to estimate the development or persistence of alcohol-related problems for those children as they aged, or to understand the nature ofthe occasions when youth were allowed to drink. The primary aim of this research is to examine the context in which drinking first occurs to better understand the nature of the relationship between age at onset of alcohol use and subsequent problem drinking. Of particular interest is the possibility that drinking at relatively early ages in the context of family gatherings may help distinguish early onset drinkers whose use escalates into problems from those who never develop alcohol problems, or mature out of them. For example, research has identified two main categories of adolescent drinkers: one whose drinking develops in ways that are not harmful, and another who are more prone to alcohol problems because of a lack of control and impuisivity (Labouvie and White. 2002; Weber et al., 1989). Although the specific reasons that alcohol problems develop may not be generalizable. the categorization is a useful departure point for understanding why some youth who drink develop problems and some don't. Pedersen and Skrondal have argued that it is important to differentiate between "normal" early onset and "problem prone" early onset (Pedersen and Skrondal, 1998). To the extent that drinking for the first time at a family gathering is
Longitudinal Effects on Problem Drinking
1987
"normal." we would expect the family initiator to have lower risk for problem drinking, other factors being equal. Therefore, this study will (a) characterize the first-use experience in terms of age of onset and onset context (e.g., initiation at, vs. outside a family gathering); (b) identify aspects of the first drinking experience that predict alcohol problems; and (c) examine whether effects of the first-use experience remain when risk factors from key domains are included in the analyses. The domains of risk factors were seleeted because prior research has established their significanl associations with alcohol-related problems or disorder, and because of their substantive relationship with aspects ofthe first-use experience and broader social context. Specifically, individuallevel factors, such as degree of attachment to social institutions, inciuding religion and school (Mason and Windle, 2001. 2002; Brown et al.. 2001, Hops et al.. 1999; Miller et al., 2000; Park et al., 2001), as well as tendencies toward delinquency (White, 1997), sensation seeking (Bates et al., 1994; Comeau ct al., 2001; Donohew et al., 1999; Greene et al., 2000). or depression (Deykin et al., 1987; White, 1992; Tsehann, 1994) may be associated with youth's attitudes about or reasons for using alcohol. Family history of alcoholism might infiuence expectations about the effect of alcohol (Brown. 1999; Wiers. 2000), or actual reactions to it. Alcohol use among parents, siblings, and friends provide opportunities to drink and to model drinking behavior (Barnes et al.. 1997; Weinberg et al., 1994; White et al., 1991; Welte et a!., 1999; Wood et al., 2001). Finally, aspects ofthe parent-child relationship, particularly those that suggest parents know and care about what their children are doing, influence decisions to drink and to spend time with youth who do (Barnes and Farrell. 1992). It is also likely that opportunities to model drinking behavior and characteristics of the parent-child relationship vary across families in which youth drink at family gatherings. For example, the long-term consequences of drinking within a family context may be harmful when the family is comprised of members who themselves are prone to excessive alcohol consumption, whereas there may be no or protective effects when it occurs in families without negative alcohol-use patterns.
METHOD Satnple
^
Data are from the Rutgers Health and Human Development Project (HHDP), a five-wave, prospective study of substance-use behaviors in a
1988
Warner and White
community sample that is most representative of white adolescents living in a metropolitan, working- and middle-class environment (Pandina et al., 1984). Adolescents aged 12, 15. and 18 were recruited between 1979 and 1981 from a random selection of telephone numbers in New Jersey. The initial recruitment was conducted by trained staff of the Eagleton Institute and the Eagleton Poll at Rutgers University to identify households with an eligible adolescent. Eligibility criteria included birth year, absence of a language barrier, absence of serious physical and mental limitations, and not being institutionalized at the time of initial contact. After the initial anonymous telephone survey, field staff comprised primarily of college-educated male and female interviewers of mixed races visited adolescents in their homes to recruit them and obtain their and their parents' consent to participate in a longitudinal study about health and human development. Because the project design called for the sample to be evenly divided by gender and age. a quota sampling procedure guided the initial recruitment phase. Overall. 46% of those asked to participate agreed to enroll in the study. Compared to youth who refused, participants had higher levels of parental income and education. At the time of initial data collection, the sample was comparable to the population of the state with regard to religion (50% Catholic, 30% Protestant, 9% Jewish, and 11% an "other" or no religion), and median family iticome (between $20,000 and $29,000). although a slightly higher proportion of persons in the sample were white (89yo), compared to the state census (83%) (US Bureau of the Census. 1981). Patterns of substance use were also comparable between the HHDP respondents and same-age peers in nationally representative samples, living in the northeastern part of the United States at that time (Johnston et al., 2000a. 2000b: Substance Abuse and Mental Health Services Administration. 1996). All participants were reinterviewed at three points over a 13-year period (Time 2 and Time 3 interviews occurred at 3-year intervals, followed by a 7-year interval). The longitudinal retention rate between Time 1 and 4 was 9 1 % . Only the youngest respondents were reinterviewed after an additional 5-year interval (during 1999/2000), such that their ages at the five data collection points were 12. 15. 18, 25, and 30, or 31 years old. Because most of the youth who were 15 and 18 years old had first tried alcohol before the time of their first interview, and the recall period for first use was comparatively shorter for the youth who were 12 years old at first interview, analyses were limited to the youngest participants who provided data at all follow-up periods (374 out of 447). Analysis of
Longitudinal (Effects on Problem Drinking
1989
attrition effects is presented in the Results section. Because we aim to understand the risk of developing alcohol-related problems, we further restricted the sample to those who reported ever drinking (H = 371) at Time I, 2, or 3. The respondents who began drinking between Times 3 and 4 were not asked questions about first-use context.
Survey Protocol At recruitment, parents and participants completed self-report questionnaires in their homes and mailed them back to the study team. Subsequently, arrangements were made for study participants to go to the project site, the Center of Alcohol Studies at Rutgers University, for a day of data collection. Data collection for all follow-up surveys occurred at the project site as well, although some participants were interviewed in their own homes. Not counting breaks, participants spent between four and six hours on a single day completing self-report, interviewer administered questionnaires, cognitive/behavioral tasks, and physical health assessments at their own pace. The data in this study eome from self-report questionnaires, which have been shown to be valid indicators of alcohol use in nonclinical samples (e.g.. Harrison et al., 1998; Rouse et al., 1985; Smith et al.. 1995). Trained interviewers supervised all aspects associated with data collection.
Measures Alcohol Use-Related Problems At each assessment, respondents completed a self-report inventory of problems experienced as a result of aleohol during the last 3 years (ever at Time I), We used this information to construct a proxy measure of lifetime DSM-IV alcohol abuse and dependence, based on 16 and 14 items, respectively (American Psychiatric Association. 1994; Muthen et al.. 1993; Woody et al., 1993). Note that this measure includes problems experienced over a 3-year period rather than a i-year period as is used for actual DSM-IV diagnoses. Because the main aim of the current research is not to understand variations over time in severity of alcohol-use disorder, but to determine if initiation context is significant for developing alcohol-related problems, we treat this measure as a clinically relevant threshold rather than a marker of disorder. Therefore, participants were categorized either as problem drinkers
1990
Warner and White
(i.e.. met the proxy criteria for either abuse or dependence) and coded as I, otherwise coded 0. Aspects of First Use The HHDP respondents were asked the age at which they tried (more than a few sips) beer. wine, and hard liquor for the first time, and the age at which they first tried each type of alcohol outside a family gathering. Because respondents were asked about onset age at each of the first three data collection points, there were some inconsistent reports of the onset age from one interview occasion to another. In those cases we chose the age reported the first time the respondent indicated ever drinking. From these responses we created a four-level variable indicating "'early use at a family gathering." "early use outside a family gathering," "late use at a family gathering", and "late use outside a family gathering." Based on the onset age distribution (median = 10, mean = 10.7). a respondent is in the "early use" category if the first drink occurred before the age of 11. It is important to note that these analyses are based only on the youngest cohort respondents, who on average reported younger ages at onset than the older respondents, and also had smaller discrepancies in reported onset ages (Golub et al., 2000), Many respondents had two types of alcohol onset; the first time at a family gathering and the first time outside a farnily gathering. Therefore, we created another variable that represented the difference in years between the ages reported for inside and outside family gathering initiation: of those who first drank at a family gathering, 56% drank outside a family gathering within 5 years, while the other 44% did not drink outside a family gathering for 5 or more years. Information about the first-use experience includes whether the respondent felt "high or drunk" at alcohol initiation (in the form of a yes/no question), and what the experience was like (very pleasant, pleasant, no effect, unpleasant, very unpleasant) trichotomized into "not pleasant" (the reference group), "no efiect." and "pleasant." Risk Factors The HHDP data set contains information on a range of factors that research has shown to be associated with alcohol use outcomes (Petraitis et al.. 1995; Hawkins et al., 1992; for reviews). The definition and operationalization of the variables that were used in the multivariate models are described in Table 1. Specific measures fall into four domains
Longitudinal Effects on Problem Drinking
1991
of influence; family history, drinking among youth's friends and family (i.e., social influences), parental relationship, and intrapersonal. Due to concerns th;it the power to detect significant relationships between predictors and problem drinking would be limited in multivariate models, we used a two-step process to select risk factors. First, available variables from each ofthe four domains described above that have been identified in the literature as having strong associations with early alcohoi use or alcohol use-related problems were evaluated one at a time in a series of chi-square analyses for categorical variables and one-way analysis of variance (ANOVAs) for continuous variables. Variables significantly associated with alcohol use-related problems at p i J - i
u Ci£> g
Q~
>-™ ^ ^
- ^
>
U ^
-
.
fl^
^
-
O -
O
r
-
f
O O - C ^ = ° O o
l .Sf^±-a.i-ifloo
" ^ f l ^ oj-doj CkO C i GO ;yfl§ C M
'
^- ^ ,2J^ :2 ^ . £ °
d
Q ^
eft r \
-n--^.
^
cS,;^°'^« c ' ^ ^ r - o j - c ' ^ C OJ QQ oij ( ^ flkH^O^W j
"^
o
« fl ^ . d
-
O
O D . , i ? ! j
C f l O K f l ^ f l oJ-Cojn.J2ojJS OJJ e x OJ) ^ Q . Oi} Q., ^fljq^'fl^fl
C M
Q S
CM C M
d
•o
ot
^ 5 f l 0
=
^
^
T~ ^
' ^ i2
S |f
S -I .E Mu | . 2
7 - ^
^
flS
n d i o
o
^-—-S^-S
"^^
• •
•c^rto
4J d
^^'r,
^ E OJ
•o
^
^
o o -d P
S *
a
-So d >-: o •£ o -r. ' ^
O
'^
~ ^
•c
"
-s o
£ — t; o
_
c t ;
^
^
^
fcc
^
t: 00
00
d
g
,„
fl "
a >. c S -c 3 IaT -3 d
dd C c o • oJ K 1 -1 ••= fl ei
5 d &
?
* = 5
S
o d E EP:(J x; fl a u 2
^
^
P^c::SL.fl.
OJ >
^
> ' . ^ L L 1 "
.E
o x: •
t/,
ems
_
_ 2 o , >>