Injection Drug Users in the United States, 1979-2002: An Aging Population

Background  Injection drug use (IDU) is important in the epidemiology of blood-borne pathogens. Herein, trends in IDU from 1979 to 2002 are analyzed.

Methods  The National Household Survey on Drug Abuse is an ongoing survey of drug use among the US population 12 years and older. Participants were chosen using a multistage sampling design and interviewed by written questionnaire (1979-1998) or audio computer-assisted self-interviewing (1999-2002). Herein, we examine the prevalence of a history of IDU at any time in the past (IDU-ever) or within the past year.

Results  In the 2000-2002 surveys, 1.5% (95% confidence interval [CI], 1.4%-1.6%) reported IDU-ever (weighted estimate, 3.4 million persons). Prevalence was highest in persons aged 35 to 49 years (3.1%; 95% CI, 2.8%-3.4%), was higher in men (2.0%; 95% CI, 1.8%-2.2%) than women (1.0%; 95% CI, 0.9%-1.1%), and was higher in whites (1.7%; 95% CI, 1.5%-1.8%) than blacks (0.8%; 95% CI, 0.7%-1.1%) or Hispanics (1.1%; 95% CI, 0.8%-1.4%). Prevalence decreased with increasing annual income and educational level. Of all participants, 0.19% (95% CI, 0.16%-0.23%) reported IDU within the past year (weighted estimate, 440 000 persons). Ten years earlier (1990-1992), 1.6% (95% CI, 1.5%-1.8%) reported IDU-ever; prevalence did not differ by race. From 1979 through 2002, the mean age of participants with IDU within the past year increased from 21 to 36 years; the age of participants with IDU-ever increased from 26 to 42 years. From 2000 to 2002, 59.4% of all persons with IDU-ever were aged 35 to 49 years.

Conclusions  The mean age of injection drug users has increased substantially. Persons born between the late 1940s and early 1960s have the highest prevalence of IDU-ever. Self-reported IDU rates are now lower among young blacks than young whites.

Injection drug use (IDU) has an immediate and deleterious effect on individuals and communities and provides an efficient mechanism for transmitting blood-borne viruses, including human immunodeficiency virus, hepatitis C virus (HCV), and hepatitis B virus. All 3 can cause chronic asymptomatic infections that can eventually lead to severe illnesses years after the initial infection.^1-3 Thus, former injection drug users remain at risk for the consequences of IDU long after they have quit using drugs.

This is particularly true with HCV infection, which is the most common chronic blood-borne viral infection in the United States.^4,5 A recent survey⁵ estimated that 4.1 million US residents had antibodies to HCV and that 3.2 million had a chronic infection. Approximately half of HCV-infected survey participants acknowledged having used injection drugs. Most of these persons (83%) were former injection drug users, and most (83%) were at least 35 years old.⁵ These older former drug users may not recognize that their behavior years earlier had put them at risk for HCV infection.

Therefore, from a public health perspective, the epidemiology of current IDU and the epidemiology of past IDU are important. This study uses a series of large national surveys to examine trends in IDU over a 23-year period.

The National Household Survey on Drug Abuse (now called the National Survey on Drug Use and Health) has been conducted periodically since 1971 and annually since 1990.^6,7 This analysis uses the public–use data sets from 1979 through 2002 inclusively. Before 1991, the survey included the civilian noninstitutionalized population of the contiguous 48 states. In 1991, the sampling frame was expanded to include Alaska and Hawaii, residents of noninstitutional group quarters (eg, college dormitories, group homes, or civilians dwelling on military installations), and homeless persons living in shelters or single-room occupancy hotels. Persons excluded from the survey, including members of the military and institutionalized persons (eg, prisoners and patients in nursing homes), represent less than 2% of the US population.⁷

Survey participants were selected according to a stratified multistage sampling design. Final response rates varied from 69% to 89%. To account for oversampling of certain demographic groups and for other factors, such as nonresponse, participants were assigned weights equal to the inverse of their probability of selection.

Participants were interviewed at home by trained interviewers using methods to ensure privacy and confidentiality and to promote disclosure. For all sensitive questions, including questions about drug use, participants were interviewed in a private area and asked to record, on an answer sheet, responses to questions read aloud by the interviewer. Participants did not reveal their responses to the interviewers, and interviewers were not allowed to view the completed answer sheets. From 1999 through 2002, responses to sensitive questions were gathered by audio computer-assisted self-interviewing, in which participants listened to questions through a headphone and recorded their answers on a touch screen.⁷

In 1979 and 1982, respondents were asked if they had ever used heroin with a needle, but were not asked about any other IDU. Questions were subsequently added about lifetime use of cocaine (1985) or amphetamines (1988) by injection. Beginning in 1990, participants were asked specifically about lifetime history of IDU with questions such as, “Have you ever, even once, used a needle to inject a drug that was not prescribed for you, or that you took only for the experience or feeling it caused?” For this analysis, persons were considered to have ever used injection drugs (IDU-ever) if they responded yes to any of these questions.

In the analysis of trends over time, we divided data into 6 periods of 2 to 3 surveys each. To be as consistent as possible across the various periods, we defined IDU-ever in this analysis as ever having used heroin, cocaine, or stimulants by injection.

Persons were considered to have had recent IDU if they met 1 of 2 criteria: (1) the respondent had used heroin, cocaine, or stimulants by injection during the previous year (1988-2002) or (2) the respondent had ever used heroin (1979-1985) or cocaine (1985) by injection and had used that drug during the previous year.

For this analysis, participants were classified as non-Hispanic black, non-Hispanic white, Hispanic, or other based on their self-reported race and ethnicity. To increase the precision of our estimates, we combined the most recent 3 surveys, those from 2000 through 2002, when analyzing characteristics of injection drug users. We also chose an earlier 3-year period, 1990 through 1992, for comparison.

All analyses were performed using computer software (SUDAAN; SAS Institute Inc, Cary, NC) to account for the sample design and weights. When surveys from multiple years (eg, 2000-2002) were combined, it was assumed that primary sampling units had been selected independently each year. This was not strictly true from 1999 to 2002. However, because of the many primary sampling units during these years, any effect of this assumption on the estimated uncertainty would be small. Prevalence estimates within a given combination of surveys were compared by the χ² test. Prevalence estimates from 2 separate combinations of years were compared by a t test. Confidence intervals (CIs) were assumed to be symmetrical on a logit scale.

A multivariate logistic regression model was used to assess the independence of risk factors for IDU-ever in the 2000 to 2002 surveys. The model only included data from non-Hispanic black, non-Hispanic white, and Hispanic participants because of limited data on other racial/ethnic groups. Interactions among variables were examined for statistical significance and epidemiologic plausibility.

“Lifetime probability of IDU” in a given birth cohort was modeled based on IDU-ever by a separate logistic regression model. This model included data from all surveys and included dummy variables for age (category, 12-13, 14-15, 16-17, 18-19, 20-23, 24-27, 28-33, and ≥34 years) and birth cohort (category, before 1945, 1945-1949, 1950-1955, and so on). Because of statistically significant interactions between birth cohort, race, and sex, separate variables were included for birth cohorts in 6 demographic groups: male and female Hispanics, male and female non-Hispanic whites, and male and female non-Hispanic blacks. Because there were no statistically significant interactions with age category, a single set of dummy variables was used to model age in all 6 demographic groups.

The 17 surveys included 452 567 participants. The number of participants varied from 5624 in 1982 to 58 680 in 2000.

Of the 168 320 survey participants, 1841 had ever used injection drugs and 363 had used injection drugs within the previous year. The most common drugs used by injection were heroin (46.0%), cocaine (59.1%), and stimulants (46.1%). Of all participants who had ever used injection drugs, 87.2% had used at least 1 of these 3 drugs. Among the remaining 12.8%, 32.8% had injected other opiates, 25.4% had injected anabolic corticosteroids, and 41.9% either injected other drugs or did not specify the drugs they had injected (percentages do not total 100 because of rounding).

The weighted prevalence of IDU-ever was 1.5%, implying that 3.4 million persons had used injection drugs during their lifetimes. The prevalence of IDU-ever was higher among males than females (P<.001) and was highest among persons aged 35 to 49 years (3.1%; 95% CI, 2.8%-3.4%; P<.001 for differences by age) (Table 1). Among participants younger than 50 years, the prevalence of IDU-ever was higher among non-Hispanic whites (henceforth referred to as “whites”) than non-Hispanic blacks (“blacks”) or Hispanics (P<.001 for both). Among participants 50 years and older, there were no statistically significant differences between the 3 race/ethnic categories (P>.80 for all comparisons).

On univariate analysis, IDU-ever was also associated with birth in the United States, lower educational level, lower annual income, and lack of full- or part-time employment (Table 2). Prevalence was higher among persons who were divorced, separated, or never married and was lower among widowers. Military service was associated with IDU-ever only among men aged 20 to 49 years. Among persons aged 20 to 64 years who were employed full- or part-time, neither random nor nonrandom drug testing in the workplace was associated with IDU-ever.

In a multivariate logistic regression model that included an interaction term for age and race, IDU-ever was independently associated with several factors (Table 3). Adjusted odds ratios were statistically significantly (P<.05) lower among blacks than whites in all age groups except the oldest, in which the adjusted odds ratio was higher among blacks.

The weighted prevalence of recent IDU was 0.19% (95% CI, 0.16%-0.23%), corresponding to 440 000 (95% CI, 360 000-530 000) persons. Recent IDU was more prevalent among males than females (P=.01), but was not significantly different by race (Table 2). Prevalence varied less markedly by age: 0.15% (95% CI, 0.12%-0.20%) among persons aged 12 to 17 years, 0.28% (95% CI, 0.23%-0.34%) among persons aged 18 to 49 years, and 0.06% (95% CI, 0.02%-0.16%) among persons 50 years and older.

Recent IDU prevalence varied with other demographic factors in a manner similar to that for IDU-ever, except for a lack of association with military service among 20- to 49-year-old men (Table 2). Also, recent IDU was less prevalent among full- or part-time employees aged 20 to 64 years who were subjected to nonrandom workplace drug testing compared with those not subjected to drug testing (Table 2).

On multivariate analysis, recent IDU was independently associated with male sex, residence in a metropolitan statistical area, birthplace in the United States, marital status other than married or widowed, lower educational level, lower annual family income, and unemployment. Recent IDU was not associated with a history of military service. Adjusted odds ratios were lower among blacks than whites aged 20 to 34 years, but were not statistically significantly different among participants in older age groups (Table 3).

Of the 70 685 participants in the 1990-1992 surveys, the weighted prevalence of IDU-ever was 1.6% (95% CI, 1.5%-1.8%), corresponding to 3.3 million (95% CI, 3.0-3.6 million) persons. The weighted prevalence of recent IDU was 0.32% (95% CI, 0.26%-0.40%), corresponding to 650 000 (95% CI, 530 000-810 000) persons.

The prevalence of IDU-ever was not statistically significantly different between whites (1.6%; 95% CI, 1.4%-1.8%), blacks (2.0%; 95% CI, 1.6%-2.5%), or Hispanics (1.8%; 95% CI, 1.4%-2.4%). Among participants aged 12 to 29 years, IDU-ever was more prevalent in whites than blacks (2.2% vs 1.5%; P=.02), but there was no difference for recent IDU (0.6% vs 0.5%; P=.70). Among participants 30 years and older, IDU-ever was more prevalent among blacks than among whites (2.3% vs 1.3%; P=.02), as was recent IDU (0.5% vs 0.2%; P=.04).

Among participants younger than 35 years, the prevalence of IDU-ever generally decreased throughout the 1990s (Figure 1A and B). The decline in prevalence was particularly striking among young blacks (Figure 1B), in whom the prevalence was less than 0.3% in the final period (2001-2002) in all 3 age groups younger than 35 years. Among white participants 35 years and older (Figure 1C), the prevalence of IDU-ever increased markedly. This increase began in the 35- to 49-year age group around 1980 and in the 50- to 64-year age group approximately 15 years later, when birth cohorts born in 1945 and later first became part of these groups. Trends in the prevalence of IDU-ever in non-Hispanic black participants 35 years and older varied by age group (Figure 1D).

A trend of decreasing recent IDU among participants younger than 35 years (Figure 2A and B) was particularly striking among black participants (Figure 2B). Among older participants, almost all recent IDU was in the 35- to 49-year age group (Figure 2C and D).

From 1979 to 2002, the mean age of participants with recent IDU increased from 21 to 36 years (Figure 3). During the same period, the mean age of participants with IDU-ever increased from 26 to 42 years.

Lifetime probability of engaging in IDU, estimated from a logistic regression model that included age, sex, race, and birth cohort as dependent variables, showed strong cohort effects (Figure 4) and peaked approximately 5 to 10 years earlier among blacks compared with whites. From cohorts born before 1955, trends among Hispanics were generally similar to those among whites. For later cohorts, the lifetime probability of IDU was lower among Hispanics. Among black cohorts, the probability of IDU was higher than among whites for cohorts born before 1955, but was substantially lower for cohorts born after 1955. The probability of IDU was almost always higher among men than among women, although differences between men and women progressively narrowed over time.

The demographic characteristics of US injection drug users changed substantially during the 23-year period of this analysis. The marked increase in the mean age of injection drug users—an increase of 16 years for those who have ever used injection drugs and 15 years for those who are actively using drugs—is consistent with the epidemic of IDU that started in the 1960s and peaked in the 1970s and 1980s. The increasing mean age and the fact that middle-aged adults are more likely than young adults to have ever used injection drugs would be consistent with a large decline in the initiation of IDU in the early 1990s, which could at least partly account for the large decrease in the incidence of acute hepatitis C observed in the United States at that time.⁸ This analysis does not address why IDU has declined among younger cohorts, but other data suggest the decline is not solely attributable to switching from injection use to noninjection use. According to surveys⁹ of adolescents and young adults in the United States, illicit drug use among youth has declined since the late 1970s.

There were also important differences in temporal trends by race, particularly among men: the prevalence of IDU may have peaked in earlier cohorts of black men, but, in every cohort born since the late 1950s, the estimated lifetime probability of IDU was higher among whites than blacks. Furthermore, the 2000-2002 surveys show much lower rates of IDU among blacks than whites younger than 50 years. Other data sources also suggest that IDU by young adults is now lower among blacks than whites. Among young adults in another recent US survey,⁵ self-reported IDU rates were lower among blacks than whites and the prevalence of antibodies to HCV was not significantly different. In a recent study¹⁰ among new inmates in Maryland correctional facilities, the prevalence of antibodies to HCV among blacks was less than half that among whites.

The National Household Survey on Drug Abuse/National Survey on Drug Use and Health may provide the most representative assessment of IDU in the United States, but the survey data are subject to limitations. Most important, all drug use data are self-reported; some participants may be reluctant to disclose drug use despite efforts to encourage disclosure. In one longitudinal study, for example, a small proportion of participants who had disclosed prior drug use on one survey denied the same drug use when asked on a subsequent survey. This rate of recanting was 5.1% for marijuana and 7.4% for cocaine.¹¹ Another limitation is the scope of the survey, which excludes certain important groups, such as prisoners¹² and the homeless not living in shelters, both of whom have high rates of illicit drug use. For this reason, the total number of Americans with current or past IDU is certainly higher than the number estimated by this survey. Among the roughly 2 million US prisoners, for example, the prevalence of antibodies to HCV is approximately 15% to 40%.¹² Assuming, based on this, that 30% of prisoners have ever engaged in IDU, then the National Household Survey on Drug Abuse/National Survey on Drug Use and Health would have undercounted injection drug users by approximately 600 000. There are few representative data on which to base an estimate of the number of homeless injection drug users.

Furthermore, the trends highlighted in this assessment may be subject to certain biases. The survey itself changed during the 23-year period analyzed herein. Most questions on the survey have been constant over time, but several have changed since 1979 and there have been some important changes in survey design, such as the expansion of the sampling frame in the early 1990s and the implementation of audio computer-assisted self-interviewing in 1999. In addition, the lifetime probability of IDU model implicitly assumed that mortality rates among persons who have ever used injection drugs have been similar to those who have not used injection drugs. Mortality among active injection drug users, a minority of this group, is as high as 1% to 3% per year, largely because of narcotic overdose, trauma, and AIDS.^13-15 Whether mortality rates are also elevated among those who have quit using injection drugs is unclear. Higher mortality rates among injection drug users would reduce the modeled lifetime probability of IDU among earlier cohorts.

The National Household Survey on Drug Abuse/National Survey on Drug Use and Health suggests that the number of persons actively engaging in IDU has decreased since the 1980s. The number of users is arguably still not low enough and there is no guarantee against a future increase. Recent trends, including the increase in use of amphetamines and prescription opioids, are concerning in this regard.

An important implication of this analysis is that “Baby Boomers,” specifically those born between the late 1940s and early 1960s, are at considerably higher risk of having ever used injection drugs than are older or younger Americans. This finding is consistent with 2 national surveys^4,5 showing this cohort to have a higher prevalence of HCV infection, for which the main risk factor is IDU. This cohort is 40 to 60 years old. In 2015, they will be 50 to 70 years old. Almost one third have an annual family income of at least $50 000. These observations—including that a 55-year-old person may be more likely to have used injection drugs than a 25-year-old person—may challenge conventional stereotypes of injection drug users.

Prevention of the long-term consequences of blood-borne virus transmission is a major public health priority. Clinicians should routinely ask patients, regardless of their demographic characteristics, about drug use, past and present.¹⁶ Clinicians should also remember that heroin is not the only recreational drug used by injection. Current drug users should be encouraged to stop using drugs. Anyone who has ever used injection drugs, no matter how infrequently or how remotely in the past, should be appropriately counseled and offered testing for the human immunodeficiency virus, the hepatitis B virus, and HCV. Clinicians should understand that injection drug users are a heterogeneous group and that, on average, they are no longer young.

Correspondence: Gregory L. Armstrong, MD, Centers for Disease Control and Prevention, Mail Stop E-03, 1600 Clifton Rd NE, Atlanta, GA 30333 (GArmstrong@cdc.gov).

Accepted for Publication: September 30, 2006.

Financial Disclosure: None reported.

Previous Presentation: This study was presented in part as a poster at the 42nd Annual Meeting of the Infectious Diseases Society of America; October 2, 2004; Boston, Mass.

Acknowledgment: We thank Allison Greenspan, MPH, for her assistance in editing the manuscript; and Beth Bell, MD, MPH, for her support in this project.