Hospitalizations, Nursing Home Admissions, and Deaths Attributable ...

Epidemiology/Health Services/Psychosocial Research O R I G I N A L

A R T I C L E

Hospitalizations, Nursing Home Admissions, and Deaths Attributable to Diabetes LOUISE B. RUSSELL, PHD1,2 ELMIRA VALIYEVA, PHD1,2,3,4 SHEILA H. ROMAN, MD5

LEONARD M. POGACH, MD6 DONG-CHURL SUH, PHD3,4 MONIKA M. SAFFORD, MD7

OBJECTIVE — To estimate all-cause hospitalizations, nursing home admissions, and deaths attributable to diabetes using a new methodology based on longitudinal data for a representative sample of older U.S. adults. RESEARCH DESIGN AND METHODS — A simulation model, based on data from the National Health and Nutrition Examination Survey (NHANES) I Epidemiologic Followup Study, was used to represent the natural history of diabetes and control for a variety of baseline risk factors. The model was applied to 6,265 NHANES III adults aged 45–74 years. The prevalence of risk factors in NHANES III, fielded in 1988 –1994, better represents today’s adults. RESULTS — For all NHANES III adults aged 45–74 years, a diagnosis of diabetes accounted for 8.6% of hospitalizations, 12.3% of nursing home admissions, and 10.3% of deaths in 1988 – 1994. For people with diabetes, diabetes alone was responsible for 43.4% of hospitalizations, 52.1% of nursing home admissions, and 47% of deaths. Adjusting for related cardiovascular conditions, which may provide more accurate estimates of attributable risks for people with diabetes, increased these estimates to 51.4, 57.1, and 56.8%, respectively. CONCLUSIONS — Risks of institutionalization and death attributable to diabetes are large. Efforts to translate recent trials of primary prevention into practice and continued efforts to prevent complications of diabetes could have a substantial impact on hospitalizations, nursing home admissions, and deaths and their societal costs. Diabetes Care 28:1611–1617, 2005

T

he prevalence of diabetes, 8.7% in 2002 (1), is rising. Although longer survival due to better treatment has contributed to higher prevalence, the sharpest increase has occurred among people aged 30 –39 years, indicating that incidence is also rising. Growing numbers

of people are thus at risk of diabetes complications, particularly cardiovascular disease (2), and the disability and death that accompany them. Recent trials (3–9) have shown that diabetes complications can be prevented or delayed by new interventions for hyperglycemia, hyperten-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Institute for Health, Rutgers University, New Brunswick, New Jersey; the 2Department of Economics, Rutgers University, New Brunswick, New Jersey; the 3Ernest Mario School of Pharmacy, Rutgers University, Piscataway, New Jersey; the 4School of Public Health, University of Medicine and Dentistry of New Jersey, Piscataway, New Jersey; the 5Department of Medicine, Johns Hopkins University, Baltimore, Maryland; 6New Jersey Veterans’ Administration Health Care System, East Orange, New Jersey; and the 7 Deep South Center on Effectiveness at the Birmingham VA Medical Center, University of Alabama at Birmingham, Birmingham, Alabama. Address correspondence and reprint requests to Louise B. Russell, PhD, Institute for Health, Rutgers University, 30 College Ave., New Brunswick, NJ 08901. E-mail: [email protected]. Received for publication 22 December 2004 and accepted in revised form 4 April 2005. Abbreviations: ADA, American Diabetes Association; NHANES, National Health and Nutrition Examination Survey; NHDS, National Hospital Discharge Survey; NHEFS, NHANES I Epidemiologic Followup Study. A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion factors for many substances. © 2005 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005

sion, and hyperlipidemia. Other trials in high-risk populations have suggested that lifestyle interventions may delay or prevent onset of type 2 diabetes (10 –12). As background for prevention and control efforts, this report presents new estimates of all-cause hospitalizations, nursing home admissions, and deaths attributable to diabetes. We used a methodology based on the first and third National Health and Nutrition Examination Surveys (NHANES), two of the surveys that provide data for national health goals and clinical guidelines. Epidemiological methodologists note that cohort studies are more reliable than cross-sectional data for establishing the existence and magnitude of etiologic relationships (13–14). We used the NHANES I Epidemiologic Followup Study (NHEFS), a cohort study that tracked outcomes over 2 decades (1971–1992), to develop a simulation model that projects diabetes outcomes. We applied the simulation model to adults aged 45–74 years from NHANES III. Fielded in 1988 –1994, NHANES III better represents today’s adults (e.g., more obesity, fewer smokers). Our methodology allowed us to make estimates for diabetes alone and for diabetes together with major cardiovascular risk factors that cluster with diabetes. The estimates contribute to a more complete picture of diabetes’ impact on the population as a whole and on people with diabetes. RESEARCH DESIGN AND METHODS — The NHANES, large national surveys conducted by the National Center for Health Statistics, collect health information for representative samples of noninstitutionalized Americans. The simulation model used here is based on equations that link risk factors measured at baseline in NHANES I to outcomes recorded during NHEFS follow-up to reproduce the natural course of diabetes. At four follow-ups (1982–1984, 1986, 1987, and 1992), survey participants (or proxies for deceased or incapacitated subjects) were interviewed about all stays in health facilities since baseline or 1611

Attributable risk of diabetes

Table 1—Characteristics of NHANES III adults aged 45–74 years, by age and sex Aged 45–64 years Weighted means (see RESEARCH DESIGN AND METHODS)

Men

Sample size (n) Risk factors Age (years) Black race (%) Current smoker (%) Former smoker (%) Serum cholesterol (mg/dl) Systolic blood pressure (mmHg) BMI ⬎27.8 kg/m2 for men, ⬎27.3 for women (%) BMI ⬍19 kg/m2 (%) Serum albumin (g/dl) Diet Fiber (grams daily) Fish/shellfish (weekly servings) Fruits/vegetables (weekly servings) Exercise (regular recreational activity) (%)* Alcohol consumption† None (%) 1–3 drinks daily, almost daily (%) 4 or more drinks daily, almost daily (%) Any amount, once a week or less often (%) Chronic conditions‡ Diabetes (%) Bronchitis/emphysema, chronic cough, asthma, pleurisy (%) Heart attack (%) Heart failure (%) Arthritis or gout (%) Fracture of spine, hip, or wrist (%) Stroke or polio/paralysis (%) Malignant tumor (%)

Aged 65–74 years

Women

Men

Women

1,959 (206)

2,142 (268)

1,084 (160)

1,080 (213)

53.5 (55.7) 8.6 (11.5) 30.4 (16.7) 42.9 (60.1) 216 (214) 128 (133) 40.3 (72.9) 0.9 (0.7) 4.2 (4.1)

54.1 (56.9) 10.7 (20.9) 24.6 (23.5) 25.6 (31.7) 225 (242) 126 (136) 46.6 (69.4) 3.3 (3.8) 4.1 (4.0)

69.2 (69.2) 8.0 (11.4) 18.3 (10.6) 56.8 (62.4) 212 (210) 138 (138) 43.0 (47.1) 1.7 (1.5) 4.1 (4.1)

69.3 (69.3) 8.9 (18.5) 13.1 (9.2) 32.3 (40.3) 233 (235) 137 (143) 41.9 (69.2) 3.7 (0.4) 4.0 (3.9)

19.2 (18.4) 1.5 (1.3) 28.0 (28.0) 57.2 (48.3)

14.5 (15.1) 1.5 (1.4) 31.0 (28.8) 49.8 (32.2)

18.8 (18.5) 1.5 (1.6) 31.9 (36.8) 63.3 (60.1)

15.7 (15.7) 1.4 (1.5) 34.8 (31.3) 56.2 (41.8)

39.7 (55.2) 38.5 (21.4) 5.2 (1.6) 16.5 (21.8)

58.5 (80.8) 22.4 (8.5) 0.6 (0.7) 18.4 (10.0)

47.1 (61.7) 31.6 (19.4) 6.8 (3.1) 14.4 (15.8)

70.3 (90.2) 15.3 (2.5) 1.3 (0) 13.1 (7.3)

8.0 (100) 12.3 (19.8) 7.6 (15.5) 3.4 (7.5) 23.5 (25.6) 11.5 (13.2) 2.3 (4.7) 2.5 (4.7)

7.9 (100) 16.4 (23.4) 2.8 (11.5) 2.2 (9.6) 33.1 (49.4) 7.9 (11.5) 2.0 (6.9) 5.8 (3.8)

11.2 (100) 18.0 (15.0) 15.6 (21.2) 7.0 (14.1) 42.5 (45.3) 12.2 (17.9) 5.2 (11.6) 7.5 (8.5)

12.6 (100) 18.0 (22.6) 7.2 (21.6) 6.3 (21.1) 49.3 (59.2) 15.4 (10.5) 5.3 (15.8) 9.4 (7.8)

Data are overall values (values for people with diabetes). *Omitted category is combinations with only moderate or little activity of either type. †Omitted category is nondrinkers. ‡Omitted category is people with no chronic condition.

previous follow-up; their reports were matched with institutional records, which were also searched for additional stays. Deaths were confirmed by death certificate. The simulation model is applied to NHANES III adults aged 45–74 years to estimate the attributable risks of diabetes. The simulation model has four components: the cohort dataset, which contains baseline information about NHANES III adults aged 45–74 years; the mortality submodel, which projects allcause mortality; and the hospitalization and nursing home submodels, which project admissions. Each submodel uses separate regressions, estimated from NHEFS, for men aged 45– 64 years at baseline, women aged 45– 64 years, men aged 65–74 years, and women aged 1612

65–74 years; NHEFS excluded people older than 74. The multivariable regressions relate admissions and mortality to all clinical risk factors shown by multiple studies to be statistically significantly related to disease and/or death (Table 1). The effects of diabetes in the model are thus its net effects after adjustment for the effects of all other risk factors. Diabetes is a statistically significant determinant of all three outcomes. NHANES I defined diabetes as present if an adult reported having received a doctor’s diagnosis of diabetes (fasting blood glucose was not measured). Type 2 diabetes, which accounts for 90 – 95% of diabetes, was not distinguished from type 1. Diagnostic criteria, and the medical management of diabetes, changed little from the 1970s to the early

1990s (15); newer antiglycemic agents were not yet available and antihypertensive and anticholesterolemic medications only became widespread in the later 1980s. Thus, equations fitted to NHEFS data reasonably reflect the natural history of diabetes and can be applied to NHANES III adults, who were diagnosed by the same criteria. In-sample and outof-sample tests of the model are presented in (16 –17). For this report, the model was updated to incorporate data for 1988 –1992, the final NHEFS follow-up (18). Cohort dataset The cohort dataset includes all NHANES III respondents aged 45–74 years who were examined by a physician (n ⫽ 6,265). Since 84% had complete informaDIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005

Russell and Associates

Men Aged 45–64 years

Simulation submodels The mortality submodel is based on Weibull hazard functions that relate survival to baseline risk factors. The hospital and nursing home submodels use negative binomial regressions to relate annual admissions to the same risk factors. (Regressions available on request.) Estimates of admissions are adjusted for survival using the mortality submodel.

Aged 65–74 years

Estimates We calculated baseline estimates of admissions and deaths by entering observed baseline risk factors for the 6,265 NHANES III adults aged 45–74 years into the model regressions and running the model. Following the traditional definition of attributable risk as the “most we can hope to accomplish in reducing the risk of the disease if we completely eliminate the exposure” (13), we estimated outcomes for the scenarios listed below and calculated attributable risks as differences between these scenarios and baseline. ●

●

Total

119.0 ⫺8.6% 1.42 ⫺12.3% 4.09 ⫺10.3%

Sample weights NHANES III was a stratified cluster sample. Since all individuals in the cohort received medical examinations, the “examined sample final weight” was used to derive population totals.

Women

130.1 1.62 4.56

Men

175.6 ⫺9.1% 4.88 ⫺6.2% 5.22 ⫺10.3%

Women

193.1 5.20 5.82

No % No % No % No % Baseline diabetes change Baseline diabetes change Baseline diabetes change Baseline diabetes change

184.0 ⫺4.8% 2.66 ⫺12.5% 10.02 ⫺8.2%

% change

193.3 3.04 10.92

No Baseline diabetes

85.4 ⫺11.5% 0.20 ⫺25.9% 0.81* ⫺12.0%

160.8 ⫺43.4% 2.03 ⫺52.1% 5.77 ⫺47.0%

96.6 0.27 0.92*

284.0 4.24 10.89

⫺8.0% ⫺22.9% ⫺11.4%

234.5 ⫺37.3% 5.19 ⫺32.9% 6.90 ⫺40.9%

105.7 0.74 4.90

373.9 7.73 11.67

114.8 0.96 5.53

197.3 ⫺29.4% 2.67 ⫺56.1% 10.84 ⫺42.5%

⫺46.2% 260.4 ⫺68.2% 1.16 ⫺52.7% 2.65 279.5 6.08 18.86

134.1 1.32 7.09 119.7 ⫺54.0% 0.32 ⫺72.4% 1.31 ⫺50.6%

249.1 4.15 15.00

Table 2—Attributable risks of diabetes for all people and diabetic subjects, per 1,000 people, at year 1 and by age and sex

Year 1 All people Hospital admissions Nursing home admissions Deaths People with diabetes Hospital admissions Nursing home admissions Deaths

*Like all NHANES, NHANES I surveyed noninstitutionalized people capable of traveling to the examination trailers, thus they were healthier than the general noninstitutionalized population. NHANES I women aged 45– 64 years initially had very low mortality; the projection equations reflect this early advantage. Their mortality soon regressed to the mean and to the national men-to-women ratio for this age-group (⬍2:1). Relative differences due to diabetes are not affected.


tion and only 13 individuals lacked data for more than five risk factors, all sample subjects were retained; the age-sex mean for each risk factor replaced missing values. Most risk factors, including diabetes, were defined identically in NHANES I and III.

All people with a diagnosis of diabetes, and thus with a baseline value of one for the dichotomous variable representing diabetes, had this value changed to zero. All other risk factors remained at baseline values. In addition to setting diabetes at zero, systolic blood pressure, serum cholesterol, and rates of cardiovascular conditions associated with diabetes (heart attack, heart failure, and stroke) were reduced in people with diabetes so that the means of these risk factors equaled those for people without diabetes. Since diabetes and other risk factors for heart disease are linked, the true attributable risk of diabetes includes higher 1613

1614

255.5 ⫺40.6% 11.69 ⫺47.5% 39.87 ⫺38.8% 346.8 ⫺31.3% 430.0 28.83 ⫺25.6% 22.25 68.96 ⫺34.9% 65.14 204.0 ⫺52.2% 454.2 3.81 ⫺70.5% 53.23 19.01 ⫺47.3% 119.35 219.7 ⫺42.5% 426.5 4.94 ⫺63.4% 12.93 32.20 ⫺46.5% 36.07

359.9 ⫺20.8% 504.4 26.17 ⫺50.8% 38.73 76.97 ⫺35.5% 106.00

194.5 8.71 27.81 ⫺4.6% 206.3 ⫺3.1% 9.43 ⫺5.4% 29.48 276.0 28.5 55.35 ⫺2.1% 289.1 ⫺7.8% 29.4 ⫺4.5% 58.52 348.1 27.00 72.91 147.3 ⫺9.5% 355.6 2.38 ⫺21.2% 29.23 11.90 ⫺9.0% 76.33 174.3 ⫺5.2% 162.7 3.31 ⫺13.4% 3.02 22.97 ⫺6.6% 13.07

Year 10 All people Hospital admissions 183.9 Nursing home admissions 3.82 Deaths 24.6 People with diabetes Hospital admissions 382.5 Nursing home admissions 13.50 Deaths 60.22

% change No Baseline diabetes

% change


% change


% change


Women Aged 65–74 years Men Women Aged 45–64 years Men

Table 3—Attributable risks of diabetes for all people and diabetic subjects, per 1,000 people, at year 10 and by age and sex

RESULTS — Table 1 shows risk factor means for all NHANES III adults and people with diabetes weighted to represent noninstitutionalized adults in 1988 – 1994. Eight percent of men and women aged 45– 64 years, 11.2% of men aged 65–74 years, and 12.6% of women aged 65–74 years had a diagnosis of diabetes. People with diabetes were more likely to report having had heart attack, heart failure, or stroke. The effects of diabetes on hospital admissions, nursing home admissions, and deaths are shown in Table 2 (1st year after baseline) and Table 3 (10th year after baseline) for all people and people with diabetes. Percentage reductions from baseline show the attributable risk for each outcome. For all people aged 45–74 years, diabetes accounted for 8.6% of hospital admissions per 1,000 people in the 1st year, 12.3% of nursing home admissions, and 10.3% of all-cause mortality. Percentages were larger for people aged 45– 64 years than for the elderly and larger for women than men. For example, diabetes accounted for 8.0% of hospital admissions in middle-aged men and 11.5% in middle-aged women compared with 4.8 and 9.1% for elderly men and women, respectively. The impact on the elderly in numbers of events was larger, however, because their rates of admission and death were higher than those of people aged 45– 64 years. The population impact reflects a much larger impact on the group directly affected: people with diabetes. For them, diabetes accounts for nearly half of all three outcomes: 43% of hospital admissions per 1,000, 52% of nursing home admissions, and 47% of deaths. Although the percentages are larger for middle-aged than elderly people, they are large in all age-sex groups. Among elderly men with diabetes, diabetes accounted for 29.4% of hospital admissions; baseline admissions per 1,000 were 280 compared with 197

Total

Differences between baseline and scenario 1 represent admissions and deaths attributable to diabetes. Differences between baseline and scenario 2 show admissions and deaths attributable to diabetes plus related cardiovascular risk factors.


% change

rates of these risk factors in people with diabetes.

⫺5.7% ⫺7.6% ⫺5.7%




Table 4—Attributable risks of diabetes and diabetes plus cardiovascular conditions at year 1 All people

Hospital admissions per 1,000 people Baseline (1) Diabetes ⫽ 0 (2) Diabetes ⫽ 0, cardiovascular conditions equalized* Nursing home admissions per 1,000 people Baseline (1) Diabetes ⫽ 0 (2) Diabetes ⫽ 0, cardiovascular conditions equalized* Deaths per 1,000 people Baseline (1) Diabetes ⫽ 0 (2) Diabetes ⫽ 0, cardiovascular conditions equalized*

People with diabetes

Rate

% change

Rate

% change

130.1 119.0 116.9

— ⫺8.6% ⫺10.2%

284.0 160.8 138.0

— ⫺43.4% ⫺51.4%

1.62 1.42 1.40

— ⫺12.3% ⫺13.6%

4.24 2.03 1.82

— ⫺52.1% ⫺57.1%

4.56 4.09 4.00

— ⫺10.3% ⫺12.3%

10.89 5.77 4.70

— ⫺47.0% ⫺56.8%

*In addition to setting diabetes equal to zero for all people with diabetes, blood pressure, cholesterol, and prevalence of heart attack, heart failure, and stroke are adjusted so that their means equal those for people without diabetes.

per 1,000 for scenario 1. Among elderly women with diabetes, diabetes accounted for 37% of hospital admissions (374 vs. 235 per 1,000). Diabetes also accounted for 56% of nursing home admissions for elderly men and 33% for elderly women and ⬎40% of deaths in both groups. At 10 years, the population risks attributable to diabetes were somewhat less, reflecting higher mortality rates among people with diabetes, especially the elderly (Table 3). Even with diabetes eliminated, Table 2 shows that deaths per 1,000 for diabetic elderly men were 10.8 in year 1, compared with 10 for all elderly men, and 6.8 per 1,000 for diabetic elderly women, compared with 5.2 for all elderly women, because of their higher rates of other serious chronic conditions. Thus, by year 10, diabetic individuals were a smaller share of the cohort. The impact on people with diabetes themselves, however, remained large at year 10 (Table 3). Percentage reductions were close to those in year 1 for diabetic people aged 45– 64 years and somewhat lower for people aged 65–74 years. Fortythree percent of hospital admissions in men aged 45– 64 years and 52% in women aged 45– 64 years were attributable to diabetes. Diabetes accounted for 21 and 31% of hospitalizations in elderly men and women, respectively. It accounted for about two-thirds of nursing home admissions in middle-aged people, DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005

51% in elderly men, and 26% in elderly women. Approximately 45% of deaths in diabetic people aged 45– 64 years were attributable to diabetes and a third of deaths in those aged 65–74 years. Table 4 shows the effects of equalizing cardiovascular risk factors related to diabetes so that people with diabetes had the same means as people without diabetes. In addition to eliminating diabetes (scenario 1), scenario 2 set the means of systolic blood pressure, total cholesterol, and rates of heart attack, heart failure, and stroke in diabetic people equal to those in people without diabetes. Since diabetes clusters with these risk factors, scenario 2 may more closely approximate the true attributable risks of diabetes. In people with diabetes, percentages of hospitalizations and deaths attributable to diabetes increased by 8 to 10 percentage points to 51.4 and 56.8%, respectively, after cardiovascular conditions were equalized. The attributable percentage of nursing home admissions increased 5 percentage points. CONCLUSIONS — We have presented new estimates of the risks of hospitalization, nursing home admission, and all-cause mortality attributable to diabetes based on the NHANES I Epidemiologic Follow-up Study and NHANES III. For the population aged 45–74 years as a whole, diagnosed diabetes accounted for

8.6% of hospitalizations in 1988 –1994, 12.3% of nursing home admissions, and 10.3% of deaths. The percentages were lower for older people than middle-aged people, but absolute reductions in event rates were similar or greater because of their higher rates. Nursing home admissions, disproportionately an experience of old age, showed the greatest effect of diabetes. For people with diabetes, the percentages were naturally much larger: 43% of hospital admissions, 52% of nursing home admissions, and 47% of deaths due to diabetes. The percentages remained large in the 10th year, and absolute reductions were larger for all age-sex groups and all outcomes than in year 1. The attenuated effects in the overall population at year 10 are explained by higher mortality among diabetic people due to their higher rates of other chronic conditions. Adjusting for factors that cluster with diabetes and are part of the metabolic syndrome (higher blood pressure and cholesterol, higher rates of heart attack, heart failure, and stroke) increased the risks attributable to diabetes by 8 to 10 percentage points for hospitalizations and deaths but less for nursing home admissions. These adjustments corrected for conditions associated with diabetes that were diagnosed at baseline in NHANES III adults. Besides showing substantial health benefits from controlling or preventing diabetes, these estimates represent medical resources that would not be used if diabetes were prevented. Studies (19 –24) show that people with diabetes have higher medical costs than nondiabetic people. Most analyzed data for enrollees in a single managed care plan and measured services as a step toward estimating costs. The hospitalization and nursing home estimates that underlie the cost estimates of the ADA are closest to those presented here (25). The ADA used an attributable-risk methodology, for the U.S. population, and defined diabetes based on a doctor’s diagnosis. The methodology used in this report has several advantages. It used two linked nationally representative datasets rather than multiple datasets as the ADA did to estimate effects of diabetes on outcomes. In estimating diabetes’ effects, the methodology controlled not only for age, sex, and race but for numerous patient-level characteristics, including other risk fac1615


tors for cardiovascular disease. The data used for the projection equations were prospective and traced outcomes over 2 decades, providing a stronger basis for correctly measuring causal relationships than cross-sectional data. The estimates presented here agree in magnitude with the ADA estimates, lending strength to their findings and suggesting that diabetes itself is driving crosssectional differences in costs. The ADA reported that hospital days attributable to diabetes accounted for 9% of all hospital days and nursing home days for 15% of all nursing home days, similar to the 8.6 and 12.3% of admissions estimated here for diabetes alone and the 10.2 and 13.6% estimated after adjustment for related cardiovascular conditions. Overall, the ADA reported that diabetic people had more than double the hospital inpatient and nursing home costs of nondiabetic people, which implies that diabetes accounts for ⬎50% of the costs of diabetic people in these settings. Our estimates are for admissions, not costs, but are in general agreement. In particular, our estimates for diabetes plus related cardiovascular conditions are ⬎50 –51.4% of hospitalizations and 57.1% of nursing home admissions. Our estimates have potential limitations. First, since our study measures the impact of all diabetes, without differentiating between types 1 and 2, it overstates the impact of eliminating type 2; however, the overstatement is modest since type 2 accounts for 90 –95% of all diabetes. Second, current treatment is more effective than treatment during the 1970s and 1980s; hence, our baseline estimates may overstate the current impact of diabetes. Finally, national trends in hospital admissions during NHEFS follow-up suggested that our hospitalization projections might be high. Hospitalization rates measured by the National Hospital Discharge Survey (NHDS) rose from 1970 to the early 1980s then fell until the early 1990s; they continued to fall through 2000 for people aged 45– 64 years but rose again for people aged ⱖ65 years (26). Rates for NHANES III adults should be lower than NHDS rates because they were drawn from the noninstitutionalized population; since they were able to travel to an examination site, they may have been even healthier than the general noninstitutionalized population. To check that our projected admission rates were 1616

reasonable, we compared year 1 projections with 1990 –2000 NHDS rates for the same age-sex groups. As they should be, our projections were uniformly lower, 60 –90% of the NHDS rates. Our results support the extensive work documenting diabetes’ adverse effects by estimating its impact on hospitalizations, nursing home admissions, and mortality in a representative sample of older U.S. adults. They suggest that efforts to generalize the Diabetes Prevention Program, for example through the “Small Steps, Big Rewards” campaign sponsored by the National Institutes of Health and the Centers for Disease Control and Prevention, could have a substantial impact on diabetes-related outcomes. If diabetes could be prevented completely, we estimate that rates of hospitalization, nursing home admission, and death would fall by about half for people spared diabetes, and the beneficial effects would be sustained over 10 years. Our study offers another insight into the value of efforts to prevent and control diabetes. Acknowledgments — The work reported in this paper was supported in part by grants from the Agency for Healthcare Research and Quality. The data on which this work is based were collected by the National Center for Health Statistics (NCHS) and were made available, in the form of public use tapes, by NCHS and the Inter-university Consortium for Political and Social Research (ICPSR). Neither NCHS nor ICPSR is responsible for the analyses, interpretation, or conclusions presented here.

References 1. American Diabetes Association: Screening for type 2 diabetes (Position Statement). Diabetes Care 27 (Suppl. 1):S11–S14, 2004 2. Haffner SM, Lehto S, Ronnemaa T, Pyorala K, Laakso M: Mortality from coronary heart disease in subjects with type 2 diabetes and in nondiabetic subjects with and without prior myocardial infarction. N Engl J Med 339:229 –234, 1998 3. UK Prospective Diabetes Study (UKPDS) Group: Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet 352:837– 853, 1998 4. UK Prospective Diabetes Study (UKPDS) Group: Tight blood pressure control and risk of macrovascular and microvascular complications in type 2 diabetes (UKPDS

38). BMJ 317:703–713, 1998 5. Heart Outcomes Prevention Evaluation Study Investigators: Effects of ramipril on cardiovascular and microvascular outcomes in people with diabetes mellitus: results of the HOPE study and MICROHOPE substudy. Lancet 355:253–259, 2000 6. The LIFE Study Group: Cardiovascular morbidity and mortality in patients with diabetes in the Losartan Intervention for Endpoint reduction in hypertension study (LIFE): a randomised trial against atenolol. Lancet 359:1004 –1010, 2002 7. Gaede P, Vedel P, Larsen N, Jensen GV, Parving HH, Pedersen O: Multifactorial intervention and cardiovascular disease in patients with type 2 diabetes. N Engl J Med 348:383–393, 2003 8. Collins R, Armitage J, Parish S, Sleigh P, Peto R, the Heart Protection Study Collaborative Group: MRC/BHF Heart Protection Study of cholesterol-lowering with simvastatin in 5963 people with diabetes: a randomised placebo-controlled trial. Lancet 361:2005–2016, 2003 9. Diabetes Control and Complications Trial Research Group: The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 329:977–986, 1993 10. Hu FB, Manson JE, Stampfer MJ, Colditz G, Liu S, Solomon CG, Willett WC: Diet, lifestyle, and risk of type 2 diabetes mellitus in women. N Engl J Med 345:790 – 797, 2001 11. Diabetes Prevention Program Research Group: Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 346:393– 403, 2002 12. Tuomilehto J, Lindstrom J, Eriksson JG, Valle TT, Hamalainen H, Ilanne-Parikka P, Keinanen-Kiukaanniemi S, Laakso M, Louheranta A, Rastas M, Salminen V, Uusitupa M, the Finnish Diabetes Prevention Study Group: Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med 344:1343–1350, 2001 13. Gordis L: Epidemiology. Philadelphia, WB Saunders Company, 2000 14. Lilienfeld DE, Stolley PD: Foundations of Epidemiology, 3rd ed. New York, Oxford, 1994 15. Expert Committee on the Diagnosis and Classification of Diabetes Mellitus: Report of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Diabetes Care 20:1183–1197, 1997 16. Russell LB, Carson JL, Taylor WC, Milan E, Dey A, Jagannathan R: Modeling allcause mortality: estimates of the impact of smoking cessation based on the NHEFS. Am J Public Health 88:630 – 636, 1998



17. Russell LB, Teutsch SM, Kumar R, Dey A, Milan E: Preventable smoking and exercise-related hospital admissions: a model based on the NHEFS. Am J Prev Med 20: 26 –34, 2001 18. Russell LB, Valiyeva E, Carson JL: Effects of prehypertension on admissions and deaths: a simulation. Arch Intern Med 164: 2119 –2124, 2004 19. Brandle M, Zhou H, Smith BR, Marriott D, Burke R, Tabaei BP, Brown MB, Herman WH: The direct medical cost of type 2 diabetes. Diabetes Care 26:2300 –2304, 2003 20. Selby JV, Ray GT, Zhang D, Colby CJ: Ex-


cess costs of medical care for patients with diabetes in a managed care population. Diabetes Care 20:1396 –1402, 1997 21. Brown JB, Nichols GA, Glauber HS, Bakst AW: Type 2 diabetes: incremental medical care costs during the first 8 years after diagnosis. Diabetes Care 22:1116 –1124, 1999 22. O’Brien JA, Shomphe LA, Kavanagh PL, Raggio G, Caro JJ: Direct medical costs of complications resulting from type 2 diabetes in the US. Diabetes Care 21:1122– 1128, 1998 23. Brown JB, Pedula KL, Bakst AW: The progressive cost of complications in type 2 diabetes mellitus. Arch Intern Med 159:

1873–1880, 1999 24. Ramsey SD, Newton K, Blough D, McCulloch DK, Sandhu N, Wagner EH: Patient-level estimates of the cost of complications in diabetes in a managed care population. Pharmacoecon 16:285–295, 1999 25. American Diabetes Association: Economic costs of diabetes in the U.S. in 2002. Diabetes Care 26:917–932, 2003 26. Hall MJ, Owings MF: 2000 National Hospital Discharge Survey. In Advance Data From Vital and Health Statistics, no. 329. Hyattsville, MD, National Center for Health Statistics, 2002

1617