Skip To Content
Click for DHHS Home Page
Click for the SAMHSA Home Page
Click for the OAS Drug Abuse Statistics Home Page
Click for What's New
Click for Recent Reports and HighlightsClick for Information by Topic Click for OAS Data Systems and more Pubs Click for Data on Specific Drugs of Use Click for Short Reports and Facts Click for Frequently Asked Questions Click for Publications Click to send OAS Comments, Questions and Requests Click for OAS Home Page Click for Substance Abuse and Mental Health Services Administration Home Page Click to Search Our Site

Substance Abuse by Older Adults:  Estimates of Future Impact on the Treatment System

Table Of Contents

Chapter 5. The Aging Baby Boom Cohort and Future Prevalence of Substance Abuse

Joseph C. Gfroerer,* B.A.
Michael A. Penne, M.P.H.
Michael R. Pemberton, Ph.D.
Ralph E. Folsom, Jr., Ph.D.

Abstract: Because of the size of the baby boom cohort and the relatively higher rate of substance use relative to earlier cohorts, there is concern that as this cohort ages, there will be a substantial increase in the number of older adults with substance abuse problems. To address this concern, projections of future substance abuse prevalence were developed using data from the 1999 National Household Survey on Drug Abuse. Regression models were developed to predict problematic substance abuse among the older adult population, defined here as those aged 50 or older. The regression parameters from these models were then applied to the projected 2020 population to obtain estimates of the number of older adults with substance abuse problems in 2020. The number of older adults with substance abuse problems is estimated to increase from 2.5 million in 1999 to 5.0 million in 2020. The aging baby boom cohort will place increasing demands on the substance abuse treatment system in the next two decades and will require a shift in focus to address the special needs of an older population of substance abusers. There is also a need to develop improved tools for measuring substance use and abuse among older adults.

Historically, alcohol and illicit drug abuse have been associated with young populations. Rates of problematic use have been shown by many studies to decline with increasing age, starting in the mid- to late 20s (National Institute on Alcohol Abuse and Alcoholism [NIAAA], 2000; Office of Applied Studies [OAS], 2000a). This is primarily due to reduced use of both alcohol and illicit drugs by people as they age. When people are in their 20s and 30s, the reduced use is related to significant shifts in responsibilities, such as a having a regular job, marriage, and parenthood (Bachman, Wadsworth, O'Malley, Johnston, & Schulenberg, 1997; Gotham, Sher, & Wood, 1997). The continued reductions in prevalence rates at later ages could be related to "maturing out" (Winick, 1962) or to elevated mortality rates among substance abusers (Moos, Brennan, & Mertens, 1994).

Birth cohorts that experience high rates of illicit drug use in youth have subsequently shown higher rates of use and associated problems as they age, relative to other cohorts (OAS, 2000a). Illicit drug use was rare in cohorts immediately preceding the baby boom cohort, defined as those born between 1946 and 1964. The rate peaked in 1979, when the baby boom cohort was aged 15 to 33. During that peak year, approximately 10 percent of the estimated 25 million current illicit drug users were aged 35 or older. In 1995, when the baby boom cohort was aged 31 to 49, the percentage of current illicit drug users who were over the age of 35 had increased to 27 percent. In 1995, 49 percent of the baby boom cohort had ever used illicit drugs in their lifetime compared with only 11 percent of adults aged 50 or older (OAS, 1996). In 1996, the baby boom cohort began to reach age 50. In addition to being more likely to have used illicit drugs than previous cohorts, the baby boom cohort is larger than earlier cohorts. Rates of heavy alcohol use have also been shown to be higher among baby boomers than in earlier cohorts (NIAAA, 2000).

Taken together, these data suggest that the prevalence of problematic substance use among older adults may increase as the baby boom cohort ages. In 1998, only 7 percent of admissions to publicly funded substance abuse treatment programs involved patients aged 50 or older (OAS, 2000b). The higher rates of problematic substance abuse among the baby boom cohort will likely lead to an increase in this number. This will require a shift in focus for treatment programs, which have dealt primarily with young populations, in order to address the special needs of an older population of substance abusers.

The purpose of this chapter is to estimate the number of persons with substance abuse problems in the year 2020. By that year, the 50 or older age group will include all of the surviving baby boomers as well as a post-baby boom cohort (born between 1965 and 1970) that experienced a high rate of illicit drug use during their youth (OAS, 1996). The methodology used is similar to that used in a previous study (Gfroerer & Epstein, 1999) and is based on data from the National Household Survey on Drug Abuse (NHSDA). Employing a narrow definition of illicit drug treatment need, the previous study estimated that the number of persons aged 50 or older who would need treatment for illicit drugs would increase from 147,000 in 1995 to 911,000 in 2020. The current study employs some methodological improvements and focuses on a broader population of substance abusers, including heavy users of alcohol as well as illicit drug users.

 

Methods

The estimation of problem substance use among older adults involved two steps. First, a series of regression models was run predicting substance abuse among the older adult population in 1999. The purpose of these models was to determine parameter estimates that characterized the relationships between a set of independent variables and problem substance use among older adults. Second, the parameters estimated from these models were applied to the projected 2020 older adult population, whose values for the independent variables were determined from 1999 data, to generate estimates of substance abuse prevalence in 2020. These two steps are described in more detail below, following descriptions of the data source and definitions used.

 

Data Source

Data from the 1999 NHSDA were used in this study. Regression models were based on the data for respondents age 50 or older (n = 5,292), and the projected older adult population in 2020 was constructed from the 1999 NHSDA respondents age 29 or older (n = 16,744). The 1999 NHSDA was a nationally representative survey of the civilian, noninstitutional population aged 12 or older in the United States. The survey obtained data on substance use from 66,706 respondents interviewed anonymously in their homes using audio computer-assisted self-interviewing (ACASI) for all substance use questions. The household screening response rate (weighted) was 89.6 percent. The interview response rates (weighted) were 68.0 percent for those aged 30 to 49 and 64.6 percent for those aged 50 or older (OAS, 2000a).

 

Definition of Problem Substance Use

For this analysis, a broad definition of problem substance use was employed. To be classified as a problem substance user, at least one of the following had to be present:

  1. Dependence. Meets the criteria for alcohol or drug dependence in the past year as defined in the 4th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV, American Psychiatric Association [APA], 1994).
  2. Heavy Drug or Alcohol Use. As indicated by any of the following: (a) used marijuana daily in the past year; (b) used an illicit drug other than marijuana at least 52 times in past year (drugs include inhalants, hallucinogens, cocaine, or nonmedical use of prescription-type stimulants, pain relievers, sedatives, or tranquilizers); (c) used heroin in any form or injected cocaine or stimulants at least once in the past year; and (d) had five or more drinks on five or more separate occasions in the past month.
  3. Treatment. Received any type of treatment for a substance use problem in the past year.

Based on this definition, 196 of the 5,292 respondents aged 50 or older were classified as problem substance users. This corresponds to an estimate of 3.5 percent (weighted) of the population, which in turn translates to an estimated 2.5 million older adults being problem substance users. Of these, 7.9 percent were problem illicit drug users only, 86.6 percent were problem alcohol users only, and 5.5 percent were both.

 

Step One: Regression Models

The 50 or older sample was split into three mutually exclusive groups with low, medium, or high risk of having substance abuse problems as older adults. Persons who had not used alcohol before age 30 were defined as low risk. This group accounted for 23.7 percent of the population of older adults in 1999. Those who had used alcohol before age 30 but not marijuana were defined as medium risk (67.6 percent of the population), and those who had used both alcohol and marijuana before age 30 were defined as high risk (8.8 percent of the population). The rates of problem substance use were 0.8 percent in the low risk group, 3.7 percent in the medium risk group, and 9.6 percent in the high risk group. Splitting the sample into these three groups made it possible to include age at first use of alcohol (a continuous variable) as a predictor in the models for the medium and high risk groups because all members of these groups had used alcohol in their lifetime and therefore had a legitimate value for this predictor. Similarly, age at first use of marijuana could be included in the model for the high risk group because all members of this group had used marijuana in their lifetime. Age at first use of alcohol and marijuana are known to be important predictors of later problem substance use (Anthony & Petronis, 1995; Gfroerer & Epstein, 1999; Grant & Dawson, 1997; OAS, 2000a).

Logistic regression models were run on each of these three groups. The dependent variable in all models was problem substance use, defined above. Independent variables are listed below. Only those predictors obtained by the NHSDA that could be assumed to remain unchanged as people age beyond age 29 were considered for the models. Because initiation of cigarettes, alcohol, or marijuana rarely occurs after age 29 (Chen & Kandel, 1995), the age at first use variables (as well as the definition of low, medium, and high risk) essentially correspond to "ever use" of these substances. Nevertheless, the models did take into account initiation after age 29 because a few cases with initiation at age 30 or older would be included in the sample and classified as "no use before age 30." In the following list, the reference group for categorical variables is the first category listed after the variable.

Logistic regressions were run using analysis weights and SUrvey DAta ANalysis (SUDAAN) software to account for the complex sample design of the NHSDA in the calculation of parameter estimates and estimates of standard errors (Shah, Barnwell, & Bieler, 1998). An alpha level of .05 was used in determining statistical significance of regression parameters for the discussion of results.

To determine the adequacy of the fit of each model, the Hosmer-Lemeshow Lack-of-Fit statistic (Hosmer & Lemeshow, 1989) was utilized. Following the methodology proposed by Nagelkerke (1991), we also utilized a maximum rescaled R-square to determine the absolute percentage of variation explained by each model. These statistics are not available in SUDAAN, so model diagnostics were run using SAS V8.1.

 

Step Two: Projection to 2020 Population

A sample representing the total population aged 50 or older in 2020 was constructed from the 1999 NHSDA sample of respondents aged 29 or older. This was done by taking each respondent's age in 1999 and increasing it by 21 years. The gender, race/ethnicity, and substance use characteristics of each respondent in 1999 were assumed to be the same in 2020. Thus, all of the independent variables included in the regression models of the 50 or older population in 1999 were also known for the 29 or older population who will be aged 50 or older in 2020.

Two adjustments were made to the analysis weights of these sample cases to ensure that the sample appropriately represented the 2020 population. First, the age-gender-race distribution of the sample was forced through statistical adjustment to match population projections for the year 2020 developed by the U.S. Bureau of the Census (2000). This was done within the following groups:

Second, we adjusted the weights of the sample cases to account for an expected higher death rate among substance abusers than among nonsubstance abusers. Based on their 1999 data, persons dependent on alcohol in 1999 were assumed to have a 1.7 times higher risk of death after 21 years, and persons dependent on illicit drugs in 1999 were assumed to have a 2.8 times higher risk of death (Neumark, Van Etten, & Anthony, 2000a, 2000b). Persons who were dependent on both alcohol and illicit drugs in 1999 were assumed to have a 1.7 × 2.8 = 4.8 times higher risk of death. Thus, weights for respondents who were substance dependent in 1999 were reduced, while weights for nondependent cases were adjusted upward to result in a total (dependent plus nondependent) sum of weights in each age-race-gender group that matched the Census projection for that age-race-gender group.

The parameters estimated from the three regression models were then applied to the constructed 2020 population. A predicted probability of being a problem substance user was assigned to each sample case, based on the substance use category (low, medium, or high risk) the case fell into and the logistic regression model parameters associated with that category. Weighted sums of the predicted probabilities were then tabulated, representing the estimated prevalence in 2020.

 

Estimation of Standard Errors and Confidence Intervals

Standard errors for 1999 estimates were computed using SUDAAN software that accounts for the complex sample design of the NHSDA. Standard errors for the 2020 projections were calculated by applying a jackknife procedure in which the entire estimation process was repeated 38 times (19 superstrata by 2 replicates for each superstrata) with different random subsamples that each generated a different 2020 estimate. This methodology helps account for bias and variance associated with the modeling and prediction. Variances were calculated as follows:

This equation is used for estimating variance. It states that the variance is equal to the sum of the superstrata times the sum of the two replicates selected for each superstrata. The product of these are multiplied by the square of the product of theta sub hj minus theta divided by 2.

where

Because the estimated prevalence rates were small and necessarily between zero and one, asymmetric 95 percent confidence intervals were computed using a logit transformation. Standard errors and confidence intervals for the estimated numbers of persons were computed by multiplying the standard error and confidence intervals for corresponding rates by the population estimates.

 

Results

 

Model Diagnostics

The Hosmer-Lemeshow tests suggest that there was adequate fit in both the low risk (p = 0.3861) and medium risk (p = 0.7323) models. This test for the high risk model suggested that the model did not adequately fit the data (p = 0.0003). The maximum rescaled R-squares were r2 = .03 for the low risk model, r2 = .09 for the medium risk model, and r2 = .21 for the high risk model. Thus, the model diagnostics results were mixed, suggesting that other predictors may need to be identified to improve these projections. It is possible that splitting the sample into high, medium, and low risk groups was more important in the prediction of problem substance use than were the specific independent variables in the three models.

 

Regression Models

None of the independent variables in the low risk model were significant predictors of problem substance use (Table  1). Because of the low prevalence of problem substance use and the small proportion of the population it represents, this model has a small impact on the overall estimates for 2020.

For the medium risk model, age, gender, and age at first alcohol use were all significant predictors of problem substance use. Probability of problem substance use declined with increasing age and with increasing age at first alcohol use. Males were more likely than females to have problem substance use (odds ratio [OR] = 3.1).

Among the high risk population, age, gender, and age at first use of marijuana were significant predictors of problem substance use. In this high risk population, the probability of problem substance abuse declined with increasing age, males were significantly more likely to have problem substance use (OR = 6.1) than females, and early use of marijuana was significantly associated with problem substance use. The OR for age at first use of marijuana (0.86) indicates that for each year marijuana initiation is delayed during youth, there is a 14 percent reduction in the risk of problem substance use after reaching age 50.

Although some predictors were not significant, there was consistency across the three models in the directions for these predictors. Daily smoking was not significant in any of the models, but the direction of the ORs suggested that daily smokers in all three risk groups were more likely to have a substance abuse problem. ORs for Hispanic and black-not Hispanic were greater than 1 in all models, suggesting that these groups were more likely to have a substance abuse problem than white/other-not Hispanics.

 

 

Table 1 Logistic Regression Modeling Results

Low Risk Model

Covariates

beta

SE of beta

p-Value

Odds Ratio

95 Percent Confidence Interval

Intercept

-5.80

2.87

0.044

--

(--

-

--)

Age (continuous)

0.01

0.04

0.857

1.01

(0.93

-

1.09)

Males vs. females

0.56

0.95

0.554

1.75

(0.27

-

11.27)

Hispanic vs. white/other-not Hispanic

0.69

0.90

0.446

1.99

(0.34

-

11.74)

Black-not Hispanic vs. white/other-not Hispanic

1.20

1.12

0.283

3.33

(0.37

-

29.91)

Smoked daily before age 30 vs. not smoked daily before age 30

0.35

1.00

0.727

1.42

(0.20

-

10.06)

Medium Risk Model

Intercept

-0.80

1.11

0.470

--

(--

-

--)

Age (continuous)

-0.03

0.01

0.035

0.97

(0.94

-

1.00)

Males vs. females

1.14

0.32

0.000

3.12

(1.67

-

5.82)

Hispanic vs. white-not Hispanic

0.38

0.54

0.486

1.46

(0.50

-

4.23)

Black-not Hispanic vs. white/other-not Hispanic

0.61

0.50

0.223

1.84

(0.69

-

4.88)

Smoked daily before age 30 vs. not smoked daily before age 30

0.44

0.32

0.159

1.56

(0.84

-

2.89)

Age at first alcohol use (continuous)

-0.10

0.03

0.000

0.91

(0.87

-

0.95)

High Risk Model

Intercept

6.02

3.10

0.052

--

(--

-

--)

Age (continuous)

-0.13

0.05

0.009

0.88

(0.80

-

0.97)

Males vs. females

1.81

0.68

0.008

6.10

(1.59

-

23.37)

Hispanic vs. white/other-not Hispanic

0.13

0.93

0.888

1.14

(0.18

-

7.13)

Black-not Hispanic vs. white/other-not Hispanic

0.09

0.62

0.882

1.10

(0.33

-

3.69)

Smoked daily before age 30 vs. not smoked daily before age 30

0.83

0.58

0.150

2.29

(0.74

-

7.10)

Age at first alcohol use (continuous)

-0.03

0.06

0.650

0.97

(0.87

-

1.09)

Age at first marijuana use (continuous)

-0.16

0.07

0.021

0.86

(0.75

-

0.98)

 

2020 Projections

Applying these regression results to the projected population in 2020 resulted in a doubling of the number of older adult problem substance users—from 2.5 million in 1999 to 5.0 million (95 percent confidence interval: 3.6 million to 6.9 million) in 2020. As shown in Table  2, this is the result of a 55 percent increase in the population size (from 72.4 million to 112.5 million) combined with a 29 percent increase in the rate of problem substance use (from 3.5 to 4.5 percent) in the older adult population. Increases are projected for all gender, racial, and age groups. More than half of the projected 2020 population of older adult problem substance users are aged 50 to 59, and more than four fifths are male.

 

 

Table 2 Estimated 1999 and Projected 2020 Persons Aged 50 or Older with Problem Substance Use

Domains of Interest

1999 Estimates

 

2020 Projections

Population (1,000s)

Number of Persons with Problem Substance Use (1,000s)

Percentage of Population with Problem Substance Use (SE)

 

Population (1,000s)

Number of Persons with Problem Substance Use (1,000s)

Percentage of Population with Problem Substance Use

Percentage of Population Needing Treatment (SE)1

Total

72,460

2,548

3.5 (0.4)

 

112,476

5,037

4.5

4.5 (0.7)

Gender

               

    Male

32,865

2,028

6.2 (0.7)

 

52,424

4,060

7.7

7.7 (1.3)

    Female

39,595

519

1.3 (0.3)

 

60,052

977

1.6

1.6 (0.5)

Race/Ethnicity

               

    Hispanic

4,717

202

4.3 (1.7)

 

12,298

552

4.5

4.5 (1.7)

    Black-not Hispanic

6,590

321

4.9 (1.5)

 

12,147

593

4.9

4.9 (1.8)

    White/other-not Hispanic

61,152

2,025

3.3 (0.4)

 

88,031

3,892

4.4

4.4 (0.8)

Age Group in Years

               

    50-59

29,943

1,501

5.0 (0.7)

 

40,935

3,131

7.6

7.6 (1.5)

    60-69

19,706

607

3.1 (0.6)

 

37,927

1,296

3.4

3.4 (0.8)

    70-79

16,112

336

2.1 (0.6)

 

22,759

436

1.9

1.9 (0.4)

    80-89

5,965

105

1.8 (0.8)

 

8,636

147

1.7

1.7 (0.6)

    90+

734

0

0.0 (0.0)

 

2,219

26

1.2

1.2 (0.6)

1 Standard errors for projections are derived through jackknife replication methodology.

 

Discussion

These analyses suggest that the number of adults over the age of 50 with substance abuse problems will increase from approximately 2.5 million in 1999 to approximately 5 million in 2020. Although these analyses did not distinguish the specific kinds of problems that the estimated 5 million older adult substance abusers in 2020 will have, it is apparent that the increasing rate of problem substance use in this population is driven by an increase in problems related to the use of illicit drugs or nonmedical use of prescription drugs. In 1999, only 8.8 percent of persons 50 or older were classified in the high risk group defined in our regression models (used alcohol and marijuana before age 30). However, among those aged 29 to 49, the group projected to be aged 50 to 70 in 2020, 49.9 percent were in the high risk group. In an earlier analysis (Gfroerer & Epstein, 1999) that focused only on illicit drugs, the number of persons aged 50 or older needing treatment for illicit drugs was projected to increase over 500 percent between 1995 and 2020.

The methodology used here incorporated several improvements over the earlier work in making projections for the older adult population. The previous effort made projections for all ages 12 and older, which required the inclusion of estimates of drug initiates in future years. Because this chapter focuses only on older adults, assumptions about future rates of initiation were not critical and could be ignored. Projections in the earlier paper were based on the aging of the 1995 and 1996 NHSDA samples, with application of mortality rates by age, race, and gender. For the current study, we incorporated differential mortality rates for substance users and we adjusted the population to established Census projections for 2020 that take into account immigration.

To assess the impact of our adjustment for differential mortality, we reran the projections assuming no difference in mortality between substance abusers and nonsubstance abusers. This showed that the adjustment we used only reduced the estimated number of problem substance users in 2020 by about 100,000, or 2 percent of the total. It should also be noted that the adjustments we used were based on a 14-year follow-up of cohorts of drug- and alcohol-dependent persons. We had no basis for calculating a differential mortality over a 21-year period, so we used the estimates as they were published. Better data on the association between substance abuse and mortality is needed for future research on the aging of the baby boom cohort. Several available datasets could yield valuable information on the mortality of substance abusers:

Another improvement in the projection for the 50 or older population was to use only the sample of respondents aged 50 or older from the NHSDA to develop the regression models. This was possible because of the larger sample available in the 1999 NHSDA due to the expansion of the survey in that year. The prior study (Gfroerer & Epstein, 1999) was based on the combined 1995 and 1996 NHSDAs, which had a sample of only 200 lifetime marijuana users aged 50 or older. With that small sample and the focus on younger populations, we used a single model based on all respondents aged 35 or older to estimate projections for those aged 50 or older. Nevertheless, the difficulty in developing good models of rare characteristics with a limited set of predictors is evident from the somewhat unsatisfactory model diagnostics we obtained.

To build better models, more questions related to substance use history or other known predictors of future substance abuse could be added to future rounds of the NHSDA. For example, historical information on the quantity and frequency of use and problems associated with use at earlier ages would likely improve the prediction. The impact of the adjustment for differential mortality on projections might also increase with the use of better predictors in the models. Analysis of larger samples, either by combining several years of data or by increasing the older adult sample, would also help in developing better prediction models. In particular, it might be important to have separate regression models for age subgroups within the 50 or older population to account for potential age-related differences in the relationships between predictors and outcomes (e.g., late onset substance abuse disorders).

One important caveat of this study, as well as the prior one (Gfroerer & Epstein, 1999), is the implicit assumption that the models developed for the 50 or older population in 1999 are correct for future cohorts of older adults, particularly the 29 to 49 year olds in 1999. Recent research has suggested that patterns of progression from nonuse to alcohol/tobacco use to marijuana use to the use of hard drugs are different in different cohorts (Golub & Johnson, 2001). However, those results do not necessarily imply that our models of problem substance use are invalid for other cohorts. This is an important issue that can be studied as future waves of the expanded NHSDA become available.

Two additional caveats are of importance in interpreting these data. First, the population covered by the NHSDA excludes institutionalized and homeless persons and therefore will underestimate the number of problem substance users. Secondly, the definition of problem substance use employed in this study may not be appropriate for elderly populations. DSM-IV criteria were developed and validated in young and middle-aged samples, so they may not be appropriate for elderly populations (Patterson & Jeste, 1999). Also, the thresholds for heavy drug or alcohol use in our definition may be higher than is appropriate for elderly populations.

Future research on substance abuse among older adults should look at alternative measures of substance abuse in the older adult population and distinguish between different categories of substance abuse. For example, specific groups of interest might include (a) persons in recovery, (b) persons abusing prescription drugs, (c) persons abusing primarily illicit drugs, and (d) persons with a DSM-IV diagnosis of substance dependence or abuse. The prevention and treatment approaches may need to be quite different for each of these groups. To adequately assess these specific characteristics using the NHSDA, it may be necessary to include new survey questions addressing these topics and designed specifically for administration to an elderly population. The current NHSDA questionnaire, for example, does not assess inadvertent misuse of prescription drugs, such as taking the wrong amounts or mix of drugs prescribed.

 

Conclusions

These data support the growing consensus that the aging of the baby boom cohort, with its size and high rate of substance use, will place increasing demands on the substance abuse treatment system in this country in the next two decades. The estimates show a doubling of the number of problem substance users aged 50 or older during the next two decades—from 2.5 million in 1999 to 5.0 million in 2020.

In anticipation of this growing problem, it is essential that improved tools for measuring substance abuse among older adults be developed. Better data are needed for predicting the future trends and also for measuring current problems as they continue to emerge. Some of these data could be obtained from the NHSDA, with modifications to the questionnaire and sample design and size. It may also be necessary to develop new data systems tailored to the unique and unprecedented information needs related to substance abuse among the aging baby boom population.

 

References

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

Anthony, J. C., & Petronis, K. R. (1995). Early-onset drug use and risk of later drug problems. Drug and Alcohol Dependence, 40, 9-15.

Bachman, J. G., Wadsworth, K. N., O'Malley, P. M., Johnston, L. D., & Schulenberg, J. E. (1997). Smoking, drinking, and drug use in young adulthood: The impacts of new freedoms and new responsibilities. Mahwah, NJ: Lawrence Erlbaum Associates.

Chen, K., & Kandel, D. B. (1995). The natural history of drug use from adolescence to the mid-thirties in a general population sample. American Journal of Public Health, 85, 41-47.

Fendrich, M., & Vaughn, C. M. (1994). Diminished lifetime substance use over time: An inquiry into differential underreporting. Public Opinion Quarterly, 58(1), 96-123.

Gfroerer, J. C., & Epstein, J. F. (1999). Marijuana initiates and their impact on future drug abuse treatment need. Drug and Alcohol Dependence, 54, 229-237.

Golub, A., & Johnson, B. D. (2001). Variation in youthful risks of progression from alcohol and tobacco to marijuana and to hard drugs across generations. American Journal of Public Health, 91, 225-232.

Gotham, H. J., Sher, K. J., & Wood, P. K. (1997). Predicting stability and change in frequency of intoxication from the college years to beyond: Individual-difference and role transition variables. Journal of Abnormal Psychology, 106, 619-629.

Grant, B. F., & Dawson, D. A. (1997). Age at onset of alcohol use and its association with DSM-IV alcohol abuse and dependence: Results from the National Longitudinal Alcohol Epidemiologic Survey. Journal of Substance Abuse, 9, 103-110.

Hosmer, D., & Lemeshow, S. (1989). Applied logistic regression. New York, NY: Wiley.

Kandel, D. B., & Davies, M. (1991). Cocaine use in a national sample of U.S. youth (NLSY): Ethnic patterns, progression, and predictors. In S. Schober, & C. Schade (Eds.), The epidemiology of cocaine use and abuse (DHHS Publication No. ADM 91-1787, pp. 151-188). Rockville, MD: National Institute on Drug Abuse.

Keer, D. W., Colliver, J. D., Kopstein, A. N., Hughes, A. L., Gfroerer, J. C., Rice, S. C. Jr., & Schoenborn, C. A. (1994, April 8). Restricted activity days and other problems associated with use of marijuana or cocaine among persons 18-44 years of age: United States, 1991 [Advance Data]. Vital and Health Statistics, (246), 1-12.

Moos, R. H., Brennan, P. L., & Mertens, J. R. (1994). Mortality rates and predictors of mortality among late-middle-aged and older substance abuse patients. Alcoholism: Clinical and Experimental Research, 18(1), 187-195.

Nagelkerke, N. J. D. (1991). Miscellanea: A note on a general definition of the coefficient of determination. Biometrika, 78, 691-692.

National Institute on Alcohol Abuse and Alcoholism. (2000, June). 10th special report to the U.S. Congress on alcohol and health (NIH Publication No. 00-1583). Retrieved October 31, 2001, from http://www.niaaa.nih.gov/publications/10report/intro.pdf & http://www.niaaa.nih.gov/publications/publications.htm

Neumark, Y. D., Van Etten, M. L., & Anthony, J. C. (2000a). "Alcohol dependence" and death: Survival analysis of the Baltimore ECA sample from 1981 to 1995. Substance Use & Misuse, 35, 533-549.

Neumark, Y. D., Van Etten, M. L., & Anthony, J. C. (2000b). "Drug dependence" and death: Survival analysis of the Baltimore ECA sample from 1981 to 1995. Substance Use & Misuse, 35, 313-327.

Office of Applied Studies. (1996). Preliminary estimates from the 1995 National Household Survey on Drug Abuse (DHHS Publication No. SMA 96-3107; Advance Report No. 18). Rockville, MD: Substance Abuse and Mental Health Services Administration.

Office of Applied Studies. (2000a). Summary of findings from the 1999 National Household Survey on Drug Abuse (DHHS Publication No. SMA 00-3466, NHSDA Series H-12; available at /nsduh.htm). Rockville, MD: Substance Abuse and Mental Health Services Administration.

Office of Applied Studies. (2000b). Treatment Episode Data Set (TEDS): 1993-1998: National admissions to substance abuse treatment services (DHHS Publication No. SMA 00-3465, Drug and Alcohol Services Information System Series S-11; available at /dasis.htm#teds2). Rockville, MD: Substance Abuse and Mental Health Services Administration.

Patterson, T. L., & Jeste, D. V. (1999). The potential impact of the baby-boom generation on substance abuse among elderly persons. Psychiatric Services, 50, 1184-1188.

Shah, B. V., Barnwell, B. G., & Bieler, G. S. (1998). SUDAAN user's manual: Version 7.5. Research Triangle Park, NC: Research Triangle Institute.

U.S. Bureau of the Census. (2000) National population projections: II. Detailed files. Total population by age, sex, race, Hispanic origin, and nativity: (NP-D1-A) Annual projections of the resident population by age, sex, race, and Hispanic origin: Lowest, middle, highest series and zero international migration series, 1999 to 2100. Retrieved June 11 & October 31, 2001, from http://www.census.gov/population/www/projections/natdet-D1A.html

Winick, C. (1962). Maturing out of narcotic addiction. Bulletin on Narcotics, 14, 1-7.

________

* To whom correspondence should be sent at the Office of Applied Studies, SAMHSA, Parklawn Building, Room 16-105, 5600 Fishers Lane, Rockville, MD 20857. Telephone: 301-443-7977. E-mail: JGfroere@samhsa.gov.

Top Of PageTable Of Contents

This page was last updated on June 16, 2008.