Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2012 Nov 1.
Published in final edited form as: Drug Alcohol Depend. 2011 Apr 29;118(2-3):237–243. doi: 10.1016/j.drugalcdep.2011.03.030

Understanding Subtypes of Inner-City Drug Users with a Latent Class Approach

SJ Kuramoto 1, ASB Bohnert 2,3, CA Latkin 4
PMCID: PMC3153580  NIHMSID: NIHMS287391  PMID: 21530105

Abstract

Aims

We empirically identified subtypes of inner-city users of heroin and cocaine based on type of drug used and route of administration.

Method

The sample was recruited from the communities in Baltimore, MD (SHIELD study) and consisted of 1,061 participants who used heroin and or cocaine in the past 6 months on a weekly basis or more. Latent class analysis (LCA) was used to identify subtypes of drug users based on type of drug and route of administration. Logistic regression was used to compare the subtypes on depressive symptoms, injection risk and drug network compositions.

Findings

Inner-city drug users were classified into five subtypes: three subtypes of injection drug users (IDUs) [heroin injecting (n=134; 13%), polydrug and polyroute (n=88, 8%), and heroin and cocaine injecting (n=404, 38%)], and two subtypes with low proportions of IDUs (LIDUs) [heroin snorting (n=275, 26%) and crack smoking (n=160; 14%)]. The polydrug and polyroute subtype had the highest depressive symptoms risk among all subtypes. Injection risk was lowest in the heroin injecting subtype and significantly differed from heroin and cocaine injecting subtype. The IDU subtypes also varied in the drug network compositions. The LIDU subtypes had similar depressive symptoms risk but vastly differed in the drug network compositions.

Conclusions

Subgroups of inner-city cocaine and heroin users based on type and route of administration differed in their depressive symptoms, injection risk and drug network compositions. Future studies should longitudinally examine factors associated with transitioning across these subtypes to better inform prevention and treatment efforts.

Keywords: latent class analysis, heroin, cocaine, polydrug use, injection

1. Introduction

1.1. Drug abuse in inner-city communities

Drug abuse among individuals residing in inner-city communities has been a long-term public health concern and contributes to significant health consequences (Cornish and O’Brien, 1996; Johnson et al., 1990). Research suggests that polydrug use is also relatively common among those who abuse drugs (Gossop et al., 2002; Latkin et al., 1996b; Leri et al., 2003), and understanding variations in co-use of drugs has important public health implications among inner-city drug users as well. Co-use of heroin and cocaine has been associated with poorer heroin treatment outcomes (Williamson et al., 2006), and individuals who use both may benefit from different therapies (Dobler-Mikola et al., 2005). Additionally, individuals with polydrug use histories and disorders have been shown to be at high risk for a number of outcomes, including psychopathology (Kandel et al., 2001), HIV risk behaviors (Klee et al., 1990) and drug overdose (Darke and Hall, 2003).

1.2. Heterogeneity of drug users in inner-city communities

Previous studies have examined variations in drug users residing in these inner-city communities by either type of drug used or route of drug administration. Both environmental context (Caprioli et al., 2009) and individual characteristics (Hopwood et al., 2008) may drive what type of drug an individual uses. Understanding differences between individuals who use heroin, cocaine, or both is important, given the differing pharmacological actions of these two substances, their associated risks and the specialized treatments that have been developed for each. The route of administration is also relevant to those who use heroin or cocaine. Injecting differs from snorting or smoking in the risk for HIV transmission. Use of multiple routes may also indicate an increased risk; for example, crack smokers who also inject have been noted to engage in greater HIV risk than those who inject but do not smoke crack (Booth et al., 1993). Studies that compared injectors (IDUs) to non-injectors (NIDUs) also suggested differences between these groups on factors such as age, drug use frequency and social networks (Carpenter et al., 1998; Des Jarlais et al., 2007; Neaigus et al., 2006). Previous studies have noted the importance of social networks on drug initiation, persistence and cessation (Galea et al., 2004; Schroeder et al., 2001), as well as for injection risk behavior (De et al., 2007). Drug use and social networks can also be bidirectional (Bohnert et al., 2009), for individuals can shape their drug social network, such as by reducing drug-using social networks (Buchanan and Latkin, 2008) and affiliating with those who use similar types of drugs (Latkin et al., 2001). As drug use is a social behavior, examination of drug network composition as related to drug use pattern would further sharpen our understanding of the context surrounding polydrug use.

1.3. Using latent class analysis to understand pattern of drug use

Latent class analysis (LCA) has been a particularly helpful strategy for empirically identifying subtypes and distinct patterns based on a set of characteristics (McCutcheon, 1987). This technique has been used to examine drug use patterns in national epidemiologic samples of drug users (Agrawal et al., 2007; Carlson et al., 2005; Lynskey et al., 2006). Although these studies have the ability to generalize to a broader population of drug users, they have had a limited ability to identify and describe types of heroin and cocaine users (aside from an omnibus category of heroin and/or cocaine users), largely because the sampling of users of these substances was low. Additionally, these studies identified patterns of polydrug use based solely on type of drug used or dependence on these drugs.

To date, at least one study has used LCA to distinguish subtypes of drug users beyond pharmacological properties, based on the route and type of drugs. This study was conducted in Canada and found three classes: one class that consisted of high use of Tylenol 3 and benzodiazepine with a high rate of depression and pain, a second included non-injectors and crack smokers, and a third consisted of injection drug users (IDUs) with a high rate of HIV and Hepatitis C (Monga et al., 2007). An extension of this study using a more recent sample found eight classes (Patra et al., 2009). However, both studies were restricted to opioid users, which may only capture a portion of inner-city drug users. Although use of cocaine is relatively common among opioid users (Leri et al., 2003), sampling of opioid users alone may miss a subset of crack cocaine users who do not use opioids (Latkin et al., 2001). The current study attempts to refine our understanding of drug users in an inner-city setting by examining heterogeneity of heroin and/or cocaine drug users, and polydrug users in particular, who used any of these drugs on a weekly basis in the prior six months. We further examined if these resulting subtypes differed in depressive symptoms, injection risk and drug network compositions.

2. Methods

2.1 Study Description

The present study used data from the baseline assessment of the Self-Help In Eliminating Life-threatening Diseases (SHIELD) study, a network-based HIV prevention intervention study conducted in Baltimore, MD (Latkin et al., 2003b). Participants were recruited from communities with high rates of drug use by trained recruiters from 1997 to 1999 for the baseline assessment. Participants were eligible for the SHIELD study if they had at least weekly contact with drug users, were 18 years of age or older, were willing to become peer educators to educate their network members on HIV risk reduction techniques, were willing to bring in a risk network member for assessment, and were not recently enrolled in other HIV behavioral interventions. Among the 1,637 participants who completed the baseline visit, 1,183 individuals reported using illegal drugs in the past 6 months. This study further focused on a final sample of 1,061 individuals who used heroin and/or cocaine once a week or more in the past six months and had complete data on the sociodemographic covariates considered in this study. Study procedures were reviewed and approved by the Committee on Human Research at the Johns Hopkins School of Public Health. All participants provided an informed consent and were paid $15 at the completion of the interview.

2.2 Measures

2.2.1 Alcohol and Drug use

Interviewers asked how often the participants used the following drugs in the past 6 months: drank alcohol, smoked marijuana, took stimulants (other than cocaine), took sedatives, used poppers, injected heroin, snorted heroin, injected speedball (cocaine mixed with heroin), injected methadone, snorted cocaine, injected cocaine, or smoked crack. The available responses ranged from ‘none in the past 6 months’, ‘less than once a week’, ‘1–2 times a week’, ‘3–4 times a week’, ‘almost everyday’, ‘everyday’, ‘2–5 times a day’ and ‘more than 5 times a day’.

2.2.2 Sociodemographic

Participants were asked a series of demographic questions. The demographic characteristics considered in this study were gender, race, education, age, unemployment, homelessness, self-reported HIV status, and ever been in a drug treatment or detoxification program.

2.2.3 Depressive Symptoms

Depressive symptoms were assessed using the Center for Epidemiologic Studies for Depression scale (CES-D) (Radloff, 1977). This scale has been shown to have high reliability and validity among non-clinical drug users (Golub et al., 2004). Participants were asked 20 items regarding their psychological well-being in the past two weeks. The possible answers of ‘none’, ‘some of the time’, ‘most of the time’ or ‘all the time,’ were given the score of 0, 1, 2, or 3, respectively. These 20 items were summed, with a total possible score of 60. The cutoff point of ≥ 16 was used to define moderate depressive symptoms (Mandell et al., 1999) and ≥ 23 to denote severe depressive symptoms (Husaini et al., 1980).

2.2.4. HIV Injection Risk

Two questions were used to assess if the respondent had engaged in injection risk. Injection risk was defined as using unclean needles or cookers in the past 30 days.

2.2.4. Drug Network Composition

The composition of drug-using networks was examined using the Personal Network Inventory, which is a modified version of the Arizona Social Support Inventory (Barrera and Gottlieb, 1981). This inventory has been shown to have good concurrent and predictive validity and internal consistency (Latkin et al., 1996a). This inventory delineated up to twenty individuals (network members) who comprised the respondent’s support and risk network. This inventory also asked participants to list the type of drugs that the participant’s network members used. A drug network member was defined as a network member who used drugs, such as heroin, crack or cocaine in the past 6 months. Since many of these variables were not normally distributed, drug network compositions were explored by examining the presence of network members who have used each of the following drugs: injected heroin, injected cocaine, injected speedball, snorted heroin, snorted cocaine or smoked crack.

2.3. Statistical Analyses

2.3.1 Latent class analysis

Latent class analysis (LCA) (McCutcheon, 1987) was used to empirically define and describe subtypes of the drug users in terms of the types of drugs used and route of administration in the past 6 months. The following eight items were used to define the subtypes: 1) drank alcohol, 2) smoked marijuana 3) injected heroin, 4) injected speedball, 5) injected cocaine, 6) snorted heroin, 7) snorted cocaine, and 8) smoked crack. The use of drugs such as sedative, stimulants, poppers and methadone was rare in this sample (weekly use of sedative <5%, <2% for stimulants and <1% for poppers and methadone) and was not considered in the LCA. The frequencies of drug use were not normally distributed; therefore, eight items were dichotomized as ‘using on a weekly basis or more’ versus ‘less than once a week or never.’ A dichotomous response was preferred over the categorical frequency responses to better meet the assumption of conditional independence required for LCA.

The LCA was conducted using MPlus version 5.0 (Muthén and Muthén, 1998–2007). Missing data on these items were minimal; twenty-one individuals were missing on one drug item, one individual was missing responses on two drug items and another individual was missing responses on three drug items, but were still considered in the LCA. The two parameters of interest, a conditional probability of endorsing the item given the latent class and a latent class prevalence, were obtained from the model.

We first performed the analysis iteratively for two- through seven-class models without adjusting for demographic covariates and compared the models using several fit indices (Akaike’s Information Criteria (AIC), Bayesian Information Criteria (BIC), Entropy, Lo-Mendel Rubin Test (LMRT) and Bootstrap Likelihood Tests (BLT)) and the interpretability of the results. We then selected a model with the best fit indices and re-ran the latent class analysis adjusting for sociodemographic covariates in Table 3 (gender, age, race, education, unemployment in the past 6 months, homelessness in the past 6 months, HIV self-report positive and ever been in drug treatment), as recommended by Muthen et al. (2004) to better classify individuals into their respective class membership (Muthén, 2004).

Table 3.

Comparison of Sociodemographic, Depression Symptoms and Drug Network Characteristics within the 5-Class Model, SHIELD study (n=1,061)

Sociodemographic Characteristics Overall Injection Drug users (IDUs)
Low proportion of IDUs (LIDUs)
Comparisons across subtypes
Class 1:Heroin injecting Class2: Polydrug and polyroute injecting Class 3: Heroin and cocaine injecting Class 4: Heroin snorting Class 5:Crack smoking Overall Among IDUs Among LIDUs
n (%) n (%) n (%) n (%) n (%) n (%) p-value p-value p-value
Female gender 397 (37) 57 (43) 41 (47) 103 (26) 140 (51) 56 (35) <0.001 <0.001 0.001
African-American 997 (94) 118 (88) 86 (98) 368 (91) 271 (99) 154 (96) <0.001 0.04 0.12
At least HS/GED education 547 (52) 64 (48) 51 (58) 186 (46) 146 (53) 100 (63) 0.005 0.13 0.06
Age (Mean±SD) 39±7.0 38±6.8 36±5.5 42±6.5 36±7.1 38±6.5 <0.001 <0.001 0.004
Unemployed in the past 6 months 925 (87) 114 (85) 81 (92) 367 (91) 235 (85) 128 (80) 0.004 0.12 0.14
Homeless in the past 6 months 265 (25) 12 (9) 32 (36) 113 (28) 49 (18) 59 (37) <0.001 <0.001 <0.001
HIV self-report positive 163 (15) 15 (11) 15 (17) 103 (26) 12 (4) 18 (11) <0.001 0.001 0.006
Ever been in drug treatment 727 (69) 94 (70) 59 (67) 320 (79) 172 (63) 82 (51) <0.001 0.01 0.02
Depressive Symptoms
Moderate (CES-D >=16 but <23) 252 (24) 39 (29) 30 (34) 89 (22) 56 (20) 38 (24) 0.001 0.002 0.61
Severe (CES-D>=23) 436 (41) 44 (33) 43 (49) 183 (45) 109 (40) 57 (36)
HIV Injection Risk 374 (36) 58 (44) 44 (51) 239 (60) 18 (7) 15 (10) <0.001 0.003 0.25
Drug Network Composition
Presence of Injectors
 Heroin Injectors 332 (32) 67 (51) 32 (36) 144 (36) 68 (25) 21 (13) <0.001 0.008 0.004
 Speedball Injectors 406 (39) 55 (42) 43 (49) 235 (58) 38 (14) 35 (22) <0.001 0.002 0.03
 Cocaine Injectors 86 (8) 9 (7) 8 (9) 48 (12) 7 (3) 14 (9) 0.001 0.22 0.004
Presence of Non-injectors
 Heroin Snorters 437 (42) 49 (37) 42 (48) 106 (26) 189 (69) 51 (32) <0.001 <0.001 <0.001
 Cocaine Snorters 65 (6) 4 (3) 6 (7) 20 (5) 19 (7) 16 (10) 0.096 0.43 0.25
 Crack Smokers 402 (38) 31 (24) 38 (43) 92 (23) 113 (42) 128 (81) <0.001 <0.001 <0.001
Open in a new tab

2.3.2 Comparison across the subtypes

To examine the differences across these classes on the outcomes of interest, we then used the most likely class membership obtained from the adjusted LCA to classify individuals into a mutually exclusive subtype. These subtypes were then compared across sociodemographics, depressive symptoms, injection risk and drug network composition using the Chi-square tests for categorical variables and Kruskal-Wallis test for continuous variable (age). Sub-comparisons among the injecting (IDU) classes and low proportion of injecting (LIDU) classes were also conducted. We also ran unadjusted and adjusted logistic regressions to examine the association between these subtypes on depressive symptoms, injection risk and drug network compositions. These analyses were conducted using Stata version 10 (StataCorp, 2007).

3. Results

3.1. Description of the sample

The study included 1,061 respondents of active drug users who reported using heroin or cocaine on a weekly basis or more in the past 6 months. Thirty-seven percent of the sample were female (n=397) and predominantly African-American (94%). The mean age of the respondents was 39 (SD=7.0). Weekly use of drugs was high, except for cocaine injecting and marijuana use, which was reported in 12% and 20% of the entire sample. Table 2 and Table 3 further describe the sample.

Table 2.

Conditional Probability and Drug Use Characteristics from the Five-Class Most Likely Class Membership, SHIELD study (n=1,061)

Overall Class 1: Heroin injecting Class2: Polydrug and polyroute Class 3: Heroin and cocaine injecting Class 4: Heroin snorting Class 5: Crack smoking
n (%) n (%) n (%) n (%) n (%) n (%)
Class Probabilities 134 (13) 88 (8) 404 (38) 275 (26) 160 (14)
Item Probabilities (weekly vs. less than weekly)
Alcohol use 566 (53) 46 (34) 59 (67) 200 (50) 139 (51) 122 (77)
Marijuana use 210 (20) 15 (11) 38 (43) 48 (12) 57 (21) 52 (33)
Heroin injecting 620 (59) 129 (96) 81 (95) 370 (92) 40 (15) 0 (0)
Heroin snorting 398 (38) 11 (8) 73 (84) 10 (2) 275 (100) 29 (18)
Speedball injecting 495 (47) 48 (36) 77 (88) 370 (92) 0 (0) 0 (0)
Cocaine injecting 127 (12) 0 (0) 76 (87) 381 (95) 0 (0) 18 (11)
Cocaine snorting 475 (45) 3 (2) 40 (46) 5 (1) 38 (14) 41 (26)
Crack smoking 378 (36) 18 (13) 49 (56) 42 (10) 119 (43) 150 (94)
Routes of Drug Administration
 Snorting alone 377 (36) 0 (0) 0 (0) 0 (0) 235 (85) 142 (89)
 Injecting alone 450 (42) 103 (77) 0 (0) 347 (86) 0 (0) 0 (0)
 Both injecting and snorting 234 (22) 31 (23) 88 (100) 57 (14) 40 (15) 18 (11)
Number of single-drug users1 309 (29) 72 (54) 0 (0) 18 (4) 116 (42) 103 (64)
Open in a new tab
1

Does not include marijuana or alcohol use

3.2. Number of classes

The fit statistics of the models for 2 through 7 classes are described in Table 1. The five-class model was selected based on the lowest BIC and highest entropy. The average probability of most likely class membership ranged from 86% to 94%, with most ambiguity to differentiate between Class 4 (heroin snorting subtype) and Class 5 (crack smoking subtype).

Table 1.

Fit Statistics of the LCA Models, SHIELD study (n=1,061)

Likelihood AIC BIC Entropy LMT p-value Bootstrap LRT p-value
2 class −4430.649 8895.298 8979.737 0.875 <0.001 <0.001
3 class −4348.077 8748.155 8877.296 0.830 <0.001 <0.001
4 class −4291.037 8652.073 8825.917 0.827 <0.001 <0.001
5 class −4233.989 8555.978 8774.525 0.857 0.001 <0.001
6 class −4216.909 8539.817 8803.067 0.827 0.0509 <0.001
7 class −4197.690 8519.380 8827.332 0.905 0.0078 <0.001
Open in a new tab

3.2 Characteristics of the Five-Class Model

Table 2 shows the conditional probability of endorsing each item (route and type of drugs) among participants classified into their most likely class membership for the five-class model. Three classes emerged as the injecting subtypes (referred to as IDUs subtypes; Classes 1, 2, and 3) and two classes as the minimally injecting subtypes (referred to as Low proportion of IDUs (LIDUs) subtypes; Classes 4 and 5). Thirteen percent of the drug users (n=134) was classified as heroin injecting subtype (Class 1). Ninety-six percent of individuals assigned to this subtype had injected heroin, and the majority used one drug that was not marijuana or alcohol. Class 2 was labeled as polydrug and polyroute subtype (n=88; 8%). All individuals estimated to be in this class used injecting and non-injecting drugs (which included crack). Class 3 was referred to as the heroin and cocaine injecting subtype, since majority of respondents classified into this subtype used multiple injecting drugs. Twenty-six percent of the overall sample was classified into Class 4 (heroin snorting subtype). All individuals assigned to this class snorted heroin, and forty-three percent of these individuals also smoked crack. Fourteen percent (n=160) were classified into Class 5 (crack smoking subtype); 94% of these individuals were crack smokers.

3.4. Comparison of Sociodemographics across Subtypes

There were statistically significant differences across these subtypes on all sociodemographic characteristics examined (Table 3). The majority of these differences were driven by differences between IDU and LIDU subtypes. In general, individuals classified into IDU subtypes were more likely to have a history of drug treatment and be HIV-positive than those classified into LIDU subtypes.

To further examine the sources of these differences, sub-comparisons were made within the IDU subtypes and within the LIDU subtypes. Only 26% classified into the heroin and cocaine injecting subtype was female, whereas this was over 40% for the other two IDU subtypes. A greater proportion of individuals classified into heroin and cocaine injecting subtype had HIV-positive status and history of drug treatment. Among the three IDU subtypes, the proportion of homelessness was lowest in the heroin injecting subtype. Among the LIDU subtypes, the heroin snorting subtype had a greater proportion of females and history of drug treatment but lower proportion of homeless and HIV-positive individuals than the crack smoking subtype.

3.5. Comparison of Subtypes on Depressive Symptoms, Injection Risk and Drug Network Compositions

Table 4 examines the association between these subtypes with depressive symptoms, injection risk and drug network compositions, after adjusting for sociodemographic characteristics examined in Table 3. The LIDU subtypes were less likely to have severe depressive symptoms than the IDU subtypes [AOR=0.74 95%CI=0.56–1.00]. Among the IDU subtypes, the polydrug and polyroute subtype had the highest risk for both moderate and severe depressive symptoms, as compared to the heroin and cocaine injecting subtype [AOR=3.50; 95%CI=1.80–6.80 and AOR=2.63; 95%CI=1.40–4.94, respectively], and as compared to the heroin injecting subtype [AOR=3.32; 95%CI=1.79–6.48 and AOR=2.92; 95%CI=1.55–5.53, respectively]. No differences in depressive symptoms between the two LIDU subtypes were noted.

Table 4.

Adjusted Association of Most Likely Class Membership with Depressive Symptoms, Injection Risk and Drug Network Composition, SHIELD study (n=1,061)1,2

Low proportion of IDUs (LIDUs) (ref: IDUs) Among IDUs Among LIDUs
Heroin injecting (ref: Heroin and cocaine injecting) Polydrug and polyroute injecting (ref: Heroin and cocaine injecting) Polydrug and polyroute injecting (ref: Heroin injecting) Crack smoking (ref: Heroin snorting)
AOR (95%CI) AOR (95%CI) AOR (95%CI) AOR (95%CI) AOR (95%CI)
Depressive Symptoms
 Moderate vs. None 0.74 (0.53–1.03) 1.33 (0.84–2.13) 3.50 (1.80–6.80)*** 3.32 (1.70–6.48)*** 0.95 (0.61–1.50)
 Severe vs. None 0.74 (0.56–1.00)* 0.83 (0.53–1.30) 2.63 (1.40–4.94)** 2.92 (1.55–5.53)** 0.85 (0.57–1.29)
Injection Risk - 0.54 (0.35–0.83)** 0.79 (0.47–1.32) 1.46 (0.82–2.60) -
Drug Network Composition
Presence of Injectors
 Heroin Injectors 0.42 (0.32–0.56)*** 1.86 (1.22–2.85)** 0.95 (0.56–1.60) 0.51 (0.29–0.91)* 0.53 (0.31–0.91)*
 Speedball Injectors 0.23 (0.17–0.30)*** 0.50 (0.33–0.77)** 0.74 (0.45–1.23) 1.48 (0.83–2.64) 1.41 (0.86–2.31)
 Cocaine Injectors 0.58 (0.35–0.96)* 0.62 (0.29–1.36) 0.86 (0.37–2.02) 1.38 (0.50–3.86) 3.38 (1.42–8.03)**
Presence of Non-Injectors
 Heroin Snorters 2.12 (1.63–2.75)*** 1.53 (0.98–2.39) 2.41 (1.43–4.06)** 1.58 (0.88–2.83) 0.32 (0.21–0.48)***
 Cocaine Snorters 1.92 (1.14–3.22)* 0.62 (0.20–1.93) 1.37 (0.48–3.95) 2.22 (0.56–8.83) 1.11 (0.55–2.22)
 Crack Smokers 3.50 (2.67–4.58)*** 0.93 (0.56–1.52) 2.52 (1.48–4.29)** 2.72 (1.46–5.04)** 5.18 (3.24–8.29)***
Open in a new tab

- indicates models not estimated

***

p<0.001;

**

p<0.01;

*

p<0.05

1

Adjusted for gender, race, age, education, unemployment in the past 6 months, homeless in the past 6 months, HIV self-report status and lifetime drug treatment

2

IDUs refers to Classes 1–3 and LIDU refers Classes 4 and 5, which consists of low proportion of IDUs

Individuals classified into heroin injecting subtype were less likely to share needles or cookers in the past 30 days than individuals classified into heroin and cocaine injecting subtype [AOR=0.54; 95%CI=0.35–0.83]. However, injection risk was not significantly different between the polydrug and polyroute subtype from the other two IDU subtypes.

Table 4 also demonstrates significant differences in the composition of drug network members across these subtypes. The LIDU subtypes were less likely to have any injecting networks and were more likely to have snorting and crack smoking networks, as compared to the IDU subtypes. When the heroin injecting subtype was compared to the heroin and cocaine injecting subtype, significant differences were noted in the presence of networks who inject heroin [AOR=1.86; 95%CI=1.22–2.85] or speedball [AOR=0.50; 95%CI=0.33–0.77], but these subtypes did not differ in the presence of non-injecting networks. The presence of injecting networks was similar between the polydrug and polyroute subtype and the heroin and cocaine injecting subtype. However, the two groups differed in the presence of a heroin snorting network [AOR=2.41; 95%CI=1.43–4.06] and crack smoking network [AOR=2.52; 95%CI=1.48–4.29]. The polydrug and polyroute subtype was also significantly different from the heroin injecting subtype in the presence of a heroin injecting network [AOR=0.51; 95%CI=0.29–0.91] and crack smoking network [AOR=2.72; 95%CI=1.46–5.04]. The crack smoking subtype differed from the heroin snorting subtype in the presence of both injecting and non-injecting networks. The crack smoking subtype was less likely to have a heroin injecting network and was more likely to have a cocaine injecting network, as compared to the heroin snorting subtype [AOR=0.53, 95%CI=0.31–0.91 and AOR=3.38; 95%CI=1.42–8.03, respectively]. These subtypes also differed in the presence of a heroin snorting network and presence of a crack smoking network [AOR=0.32; 95%CI=0.21–0.48 and AOR=5.18; 95%CI=3.24–8.29, respectively].

4. Discussion

This study empirically supported the heterogeneity of inner-city drug users recruited from community settings in Baltimore, MD. The LCA identified five classes of inner-city drug users among individuals who used heroin or cocaine on a weekly basis or more in the past 6 months; these classes were distinguished by both route and type of drugs. The five subtypes found in this study were broadly categorized as IDU subtypes and LIDU subtypes. This study further found heterogeneity within the IDU and LIDU subtypes. Within the IDU subtypes, the polydrug and polyroute subtype used both non-injecting and injecting drugs, whereas the other two IDU subtypes were more likely to be injecting alone (heroin injecting and heroin and cocaine injecting subtypes). These subtypes differed in depressive symptoms, injection risk and drug network compositions. The polydrug and polyroute subtype was particularly distinct in drug network compositions compared to the other IDU subtypes. The two LIDU subtypes (crack smoking and heroin snorting) were indistinguishable in depressive symptoms but varied in their drug network compositions.

Consistent with previous studies that used LCA (Monga et al., 2007; Patra et al., 2009), we found a considerable heterogeneity among inner-city drug users. However, we found more discrete classes of drug users than Monga and colleagues (2007) and fewer than Patra and colleagues (2009). The differences may be attributable to the distinction between crack cocaine and powder cocaine use, more sampling of heroin users than prescription opioid users in our study, and more broadly, to regional variations in drug use preferences and drug availability. This study also partially supported Latkin and colleagues’ (2001) study, which used a non-LCA approach to examine differences in five predominant groups of past year drug users. Four of these five classes were replicated in our study using LCA; instead of the heroin injecting subtype, their study found individuals who snorted heroin and smoked crack as another predominant group. Differences may be due to the different statistical approach taken to classify drug users, as well as our focus on more active drug users who used drugs on a weekly basis or more in the past six months.

Despite some differences across studies, this study empirically solidified the existing literature on heterogenic drug use patterns of inner-city drug users, which has implications for understanding health risk transmission and developing tailored interventions. IDUs are generally considered to be at a more severe end of the spectrum of drug use, with more dependence (Gossop et al., 1992; Gossop et al., 1994; Strang et al., 1998), higher rates of HIV infection and greater history of treatment (Carpenter et al., 1998; Strang et al., 1992). This study motivates further understanding of polydrug and polyroute subtype in particular, as this subtype had more depressive symptoms and was more enmeshed in polydrug networks. The high prevalence of depressive symptoms across all subtypes was concerning and suggests the need for more attention to mental health. Depressive symptoms have been found to complicate treatment outcomes among drug dependent individuals (Compton III et al., 2003). A high risk of depressive symptoms also increases concern for suicide-related risks, as polydrug use has been associated with subsequent suicide attempt (Darke et al., 2007).

The heroin injecting subtype had the lowest injection risk and differed significantly from the heroin and cocaine injecting subtype. Interestingly, the polydrug and polyroute subtype did not have the highest injection risk among the IDU subtypes, despite the high depressive symptoms and use of multiple drugs, which are both noted as risk factors for injection risks (Lloyd-Smith et al., 2009; Mandell et al., 1999; Tyndall et al., 2003). The heroin injecting subtype also had a much lower proportion of homelessness; such factor could reflect more resources in individuals classified into this subtype, as compared to the other IDUs, which could potentially serve as a protective factor against sharing injection risk (Galea and Vlahov, 2002).

This study also strongly supported differential association by type and route of drugs, which suggests that drug users affiliate with individuals who use similar drugs (Oetting and Donnermeyer, 1998; Tittle et al., 1986). The polydrug and polyroute subtype was also more likely to have crack smokers in their networks than the other IDU subtypes, suggesting that this subgroup is potentially an important bridge between the IDU and LIDUs subtypes. Although depressive symptoms were similar among the LIDU subtypes, differences in their drug network composition, in both injecting and non-injecting networks, suggest that these two LIDU subtypes may be relatively distinct. The differences in these drug network compositions also suggest potential differences in social norms regarding HIV risk behavior (Latkin et al., 2003a; Latkin et al., 2010), which may also help further understand the variations in injection risk across the IDU subtypes and sexual HIV transmission across the LIDU subtypes.

This is one of the few studies to empirically identify subtypes of inner-city drug users drawn from a community setting using the empirical latent class approach. This study included a large sample of hard to reach inner-city drug users from Baltimore, MD. Exploration of the drug network compositions further solidified the validity of these subtypes. However, several limitations should be noted. The study findings may not be generalizable to drug users in other settings with different types of drug availability, especially with regards to club drugs, nonmedical use of prescription drugs and marijuana users. This study was also cross-sectional and conducted between 1997 and 1999; drug use pattern may have changed since then. We did not have information on timing of drug use among participants who reported polydrug use, although individuals who use multiple drugs concurrently may be different from those who use each drug at different occasion but within a certain time period (Leri et al., 2003). Using the most-likely class membership assignments in the analysis of socio-demographic correlates introduces the possibility of misclassification. However, the average probability for the most likely class ranged between 86% and 94%, indicating that the probability of misclassification was low for most participants.

A sample of inner-city drug users from a community setting can be further distinguished by type of drug used and route of its administration. Many IDUs used multiple drugs; in particular, this study noted that individuals classified into a polydrug and polyroute subtype is potentially highly addicted and may be in a great need for multiple modalities of treatment. Recognizing variations of drug users further provides motivation to better understand the drug use pattern of hard to reach drug users and develop tailored HIV prevention and drug treatment programs. Additional research that examines HIV risk consequences and treatment outcomes associated with these subtypes will also be informative for HIV and drug treatment studies. Further research should also examine these subtypes in a longitudinal setting to sharpen the knowledge on factors that may facilitate transitioning into the lower risk subtypes.

Acknowledgments

Role of funding source

National Institute on Drug Abuse funded the SHIELD study (R01 DA010446 (PI: Latkin)). National Institute on Drug Abuse also provided training support of S. Janet Kuramoto (F31DA0263182). Additional training support was provided by the Department of Veterans Affairs, Health Services Research and Development (CDA-09-204; PI: Bohnert). These funding sources had no further role in the study analysis, interpretation of data and in manuscript preparation and submission for publication.

The authors wish to acknowledge National Institute on Drug Abuse for project support (R01 DA010446 (PI: Latkin)) and training support of S. Janet Kuramoto (F31DA0263182). Additional training support was provided by the Department of Veterans Affairs, Health Services Research and Development (CDA-09-204; PI: Bohnert).

Footnotes

Contributors

Janet Kuramoto designed the research question, undertook the statistical analyses and wrote the first draft of the manuscript. Amy Bohnert advised on statistical analysis. All authors contributed to designing of the study and interpretation of the findings. Carl Latkin provided the data. All authors revised the manuscript drafts. All authors contributed to and have approved the final manuscript.

Conflict of Interest

None of the authors report any conflict of interests.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Agrawal A, Lynskey MT, Madden PAF, Bucholz KK, Heath AC. A latent class analysis of illicit drug abuse/dependence: results from the National Epidemiological Survey on Alcohol and Related Conditions. Addiction. 2007;102:94–104. doi: 10.1111/j.1360-0443.2006.01630.x. [DOI] [PubMed] [Google Scholar]
  2. Barrera M, Gottlieb BH. Social support in the adjustment of pregnant adolescents: assessment issues. In: Gottlieb BH, editor. Social Networks and Social Support. Sage; Beverly Hills, CA: 1981. pp. 61–96. [Google Scholar]
  3. Bohnert AS, Bradshaw CP, Latkin CA. A social network perspective on heroin and cocaine use among adults: evidence of bidirectional influences. Addiction. 2009;104:1210–1218. doi: 10.1111/j.1360-0443.2009.02615.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Booth RE, Watters JK, Chitwood DD. HIV risk-related sex behaviors among injection drug users, crack smokers, and injection drug users who smoke crack. Am J Public Health. 1993;83:1144–1148. doi: 10.2105/ajph.83.8.1144. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Buchanan AS, Latkin CA. Drug use in the social networks of heroin and cocaine users before and after drug cessation. Drug Alcohol Depend. 2008;96:286–289. doi: 10.1016/j.drugalcdep.2008.03.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Caprioli D, Celentano M, Dubla A, Lucantonio F, Nencini P, Badiani A. Ambience and drug choice: cocaine- and heroin-taking as a function of environmental context in humans and rats. Biol Psychiatry. 2009;65:893–899. doi: 10.1016/j.biopsych.2008.12.009. [DOI] [PubMed] [Google Scholar]
  7. Carlson RG, Wang J, Falck RS, Siegal HA. Drug use practices among MDMA/ecstasy users in Ohio: a latent class analysis. Drug Alcohol Depend. 2005;79:167–179. doi: 10.1016/j.drugalcdep.2005.01.011. [DOI] [PubMed] [Google Scholar]
  8. Carpenter MJ, Chutuape MA, Stitzer ML. Heroin snorters versus injectors: comparison on drug use and treatment outcome in age-matched samples. Drug Alcohol Depend. 1998;53:11–15. doi: 10.1016/s0376-8716(98)00103-3. [DOI] [PubMed] [Google Scholar]
  9. Compton WM, III, Cottler LB, Jacobs JL, Ben-Abdallah A, Spitznagel EL. The role of psychiatric disorders in predicting drug dependence treatment outcomes. Am J Psychiatry. 2003;160:890–895. doi: 10.1176/appi.ajp.160.5.890. [DOI] [PubMed] [Google Scholar]
  10. Cornish JW, O’Brien CP. Crack cocaine abuse: an epidemic with many public health consequences. Annu Rev Public Health. 1996;17:259–273. doi: 10.1146/annurev.pu.17.050196.001355. [DOI] [PubMed] [Google Scholar]
  11. Darke S, Hall W. Heroin overdose: research and evidence-based intervention. J Urban Health. 2003;80:189–200. doi: 10.1093/jurban/jtg022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Darke S, Ross J, Williamson A, Mills KL, Havard A, Teesson M. Patterns and correlates of attempted suicide by heroin users over a 3-year period: findings from the Australian treatment outcome study. Drug Alcohol Depend. 2007;87:146–152. doi: 10.1016/j.drugalcdep.2006.08.010. [DOI] [PubMed] [Google Scholar]
  13. De P, Cox J, Boivin JF, Platt RW, Jolly AM. The importance of social networks in their association to drug equipment sharing among injection drug users: a review. Addiction. 2007;102:1730–1739. doi: 10.1111/j.1360-0443.2007.01936.x. [DOI] [PubMed] [Google Scholar]
  14. Des Jarlais DC, Arasteh K, Perlis T, Hagan H, Heckathorn DD, McKnight C, Bramson H, Friedman SR. The transition from injection to non-injection drug use: long-term outcomes among heroin and cocaine users in New York City. Addiction. 2007;102:778–785. doi: 10.1111/j.1360-0443.2007.01764.x. [DOI] [PubMed] [Google Scholar]
  15. Dobler-Mikola A, Hättenschwiler J, Meili D, Beck T, Böni i, Modestin J. Patterns of heroin, cocaine, and alcohol abuse during long-term methadone maintenance treatment. J Subst Abuse Treat. 2005;29:259–265. doi: 10.1016/j.jsat.2005.08.002. [DOI] [PubMed] [Google Scholar]
  16. Galea S, Nandi A, Vlahov D. The social epidemiology of substance use. Epidemiol Rev. 2004;26:36–52. doi: 10.1093/epirev/mxh007. [DOI] [PubMed] [Google Scholar]
  17. Galea S, Vlahov D. Social determinants and the health of drug users: socioeconomic status, homelessness, and incarceration. Public Health Rep. 2002;117(Suppl 1):S135–145. [PMC free article] [PubMed] [Google Scholar]
  18. Golub ET, Latka M, Hagan H, Havens JR, Hudson SM, Kapadia F, Campbell JV, Garfein RS, Thomas DL, Strathdee SA, Project S. Screening for depressive symptoms among HCV-infected injection drug users: examination of the utility of the CES-D and the Beck Depression Inventory. J Urban Health. 2004;81:278–290. doi: 10.1093/jurban/jth114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Gossop M, Griffiths P, Powis B, Strang J. Severity of dependence and route of administration of heroin, cocaine and amphetamines. Br J Addict. 1992;87:1527–1536. doi: 10.1111/j.1360-0443.1992.tb02660.x. [DOI] [PubMed] [Google Scholar]
  20. Gossop M, Griffiths P, Powis B, Strang J. Cocaine: patterns of use, route of administration, and severity of dependence. Br J Psychiatry. 1994;164:660–664. doi: 10.1192/bjp.164.5.660. [DOI] [PubMed] [Google Scholar]
  21. Gossop M, Marsden J, Stewart D. Dual dependence: assessment of dependence upon alcohol and illicit drugs, and the relationship of alcohol dependence among drug misusers to patterns of drinking, illicit drug use and health problems. Addiction. 2002;97:169–178. doi: 10.1046/j.1360-0443.2002.00028.x. [DOI] [PubMed] [Google Scholar]
  22. Hopwood CJ, Baker KL, Morey LC. Personality and drugs of choice. Pers Individ Dif. 2008;44:1413–1421. [Google Scholar]
  23. Husaini BA, Neff JA, Harrington JB, Hughes MD, Stone RH. Depression in rural communities: validating the CES-D scale. J Community Psychol. 1980;8:20–27. [Google Scholar]
  24. Johnson BD, Williams T, Dei KA, Sanabria H. Drug abuse in the inner city: impact on hard-drug users and the community. Crime Justice. 1990;13:9–67. [Google Scholar]
  25. Kandel DB, Huang FY, Davies M. Comorbidity between patterns of substance use dependence and psychiatric syndromes. Drug Alcohol Depend. 2001;64:233–241. doi: 10.1016/s0376-8716(01)00126-0. [DOI] [PubMed] [Google Scholar]
  26. Klee H, Faugier J, Hayes C, Boulton T, Morris J. AIDS-related risk behaviour, polydrug use and temazepam. Br J Addict. 1990;85:1125–1132. doi: 10.1111/j.1360-0443.1990.tb03437.x. [DOI] [PubMed] [Google Scholar]
  27. Latkin C, Mandell W, Vlahov D, Oziemkowska M, Celentano D. People and places: behavioral settings and personal network characteristics as correlates of needle sharing. J Acquir Immune Defic Syndr. 1996a;13:273–280. doi: 10.1097/00042560-199611010-00010. [DOI] [PubMed] [Google Scholar]
  28. Latkin CA, Forman VL, Knowlton A, Sherman S. Norms, social networks, and HIV-related risk behaviors among urban disadvantaged drug users. Soc Sci Med. 2003a;56:465–476. doi: 10.1016/s0277-9536(02)00047-3. [DOI] [PubMed] [Google Scholar]
  29. Latkin CA, Knowlton AR, Sherman S. Routes of drug administration, differential affiliation, and lifestyle stability among cocaine and opiate users: implications to HIV prevention. J Subst Abuse. 2001;13:89–102. doi: 10.1016/s0899-3289(01)00070-0. [DOI] [PubMed] [Google Scholar]
  30. Latkin CA, Kuramoto S, Davey-Rothwell M, Tobin K. Social norms, social networks, and HIV risk behavior among injection drug users. AIDS Behav. 2010;14:1159–68. doi: 10.1007/s10461-009-9576-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Latkin CA, Mandell W, Vlahov D. The relationship between risk networks’ patterns of crack cocaine and alcohol consumption and HIV-related sexual behaviors among adult injection drug users: a prospective study. Drug Alcohol Depend. 1996b;42:175–181. doi: 10.1016/s0376-8716(96)01279-3. [DOI] [PubMed] [Google Scholar]
  32. Latkin CA, Sherman S, Knowlton A. HIV prevention among drug users: outcome of a network-oriented peer outreach intervention. Health Psychol. 2003b;22:332–339. doi: 10.1037/0278-6133.22.4.332. [DOI] [PubMed] [Google Scholar]
  33. Leri F, Bruneau J, Stewart J. Understanding polydrug use: review of heroin and cocaine co-use. Addiction. 2003;98:7–22. doi: 10.1046/j.1360-0443.2003.00236.x. [DOI] [PubMed] [Google Scholar]
  34. Lloyd-Smith E, Wood E, Li K, Montaner JS, Kerr T. Incidence and determinants of initiation into cocaine injection and correlates of frequent cocaine injectors. Drug Alcohol Depend. 2009;99:176–182. doi: 10.1016/j.drugalcdep.2008.07.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Lynskey M, Agrawal A, Bucholz K, Nelson EC, Madden PAF, Todorov AA, Grant JD, Martin NG, Heath AC. Subtype of illicit drug users: a latent class analysis of data from an Australian twin sample. Twin Res Hum Genet. 2006;9:523–530. doi: 10.1375/183242706778024964. [DOI] [PubMed] [Google Scholar]
  36. Mandell W, Kim J, Latkin C, Suh T. Depressive symptoms, drug network, and their synergistic effect on needle-sharing behavior among street injection drug users. Am J Drug Alcohol Abuse. 1999;25:117–127. doi: 10.1081/ada-100101849. [DOI] [PubMed] [Google Scholar]
  37. McCutcheon AL. Latent Class Analysis. Sage Publications, Inc; Newbury Park, CA: 1987. [Google Scholar]
  38. Monga N, Rehm Jr, Fischer B, Brissette S, Bruneau J, El-Guebaly N, NoÎl L, Tyndall M, Wild C, Leri F, Fallu JS, Bahl S. Using latent class analysis (LCA) to analyze patterns of drug use in a population of illegal opioid users. Drug Alcohol Depend. 2007;88:1–8. doi: 10.1016/j.drugalcdep.2006.08.029. [DOI] [PubMed] [Google Scholar]
  39. Muthén B. Latent variable analysis: growth mixture modeling and related techniques for longitudinal data. In: Kaplan D, editor. Handbook of Quantitative Methodology for the Social Sciences. Sage Publications; Newbury Park, CA: 2004. pp. 345–368. [Google Scholar]
  40. Muthén LK, Muthén BO. Mplus User’s Guide. Muthén and Muthén; Los Angeles, CA: 1998–2007. [Google Scholar]
  41. Neaigus A, Gyarmathy VA, Miller M, Frajzyngier VM, Friedman SR, Des Jarlais DC. Transitions to injecting drug use among noninjecting heroin users: social network influence and individual susceptibility. J Acquir Immune Defic Syndr. 2006;41:493–503. doi: 10.1097/01.qai.0000186391.49205.3b. [DOI] [PubMed] [Google Scholar]
  42. Oetting ER, Donnermeyer JF. Primary socialization theory: the etiology of drug use and deviance. Subst Use Misuse. 1998;33:995–1026. doi: 10.3109/10826089809056252. [DOI] [PubMed] [Google Scholar]
  43. Patra J, Fischer B, Maksimowska S, Rehm J. Profiling poly-substance use typologies in a multi-site cohort of illicit opioid and other drug users in Canada – a latent class analysis. Addict Res Theory. 2009;17:168–185. [Google Scholar]
  44. Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas. 1977;1:385–401. [Google Scholar]
  45. Schroeder JR, Latkin CA, Hoover DR, Curry AD, Knowlton AR, Celentano DD. Illicit drug use in one’s social network and in one’s neighborhood predicts individual heroin and cocaine use. Ann Epidemiol. 2001;11:389–394. doi: 10.1016/s1047-2797(01)00225-3. [DOI] [PubMed] [Google Scholar]
  46. StataCorp. Stata Statistical Software: Release 10.0. Stata Corporation; College Station, TX: 2007. [Google Scholar]
  47. Strang J, Bearn J, Farrell M, Finch E, Gossop M, Griffiths P, Marsden J, Wolff K. Route of drug use and its implications for drug effect, risk of dependence and health consequences. Drug Alcohol Rev. 1998;17:197–211. doi: 10.1080/09595239800187001. [DOI] [PubMed] [Google Scholar]
  48. Strang J, Des Jarlais DC, Griffiths P, Gossop M. The study of transitions in the route of drug use: the route from one route to another. Br J Addict. 1992;87:473–483. doi: 10.1111/j.1360-0443.1992.tb01948.x. [DOI] [PubMed] [Google Scholar]
  49. Tittle CR, Burke MJ, Elton FJ. Modeling sutherland’s theory of differential association: toward an empirical clarification. Soc Forces. 1986;65:405–432. [Google Scholar]
  50. Tyndall MW, Currie S, Spittal P, Li K, Wood E, O’Shaughnessy MV, Schechter MT. Intensive injection cocaine use as the primary risk factor in the Vancouver HIV-1 epidemic. AIDS. 2003;17:887–893. doi: 10.1097/00002030-200304110-00014. [DOI] [PubMed] [Google Scholar]
  51. Williamson A, Darke S, Ross J, Teesson M. The effect of persistence of cocaine use on 12-month outcomes for the treatment of heroin dependence. Drug Alcohol Depend. 2006;81:293–300. doi: 10.1016/j.drugalcdep.2005.08.010. [DOI] [PubMed] [Google Scholar]

RESOURCES