Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2015 Apr 1.
Published in final edited form as: Ann Epidemiol. 2014 Jan 24;24(4):304–311. doi: 10.1016/j.annepidem.2014.01.008

Do metropolitan HIV epidemic histories and programs for people who inject drugs and men who have sex with men predict AIDS incidence and mortality among heterosexuals?

Samuel R Friedman 1,2, Brooke S West 1, Barbara Tempalski 1, Cory M Morton 1, Charles M Cleland 3, Don C Des Jarlais 4, H Irene Hall 5, Hannah LF Cooper 6
PMCID: PMC3954755  NIHMSID: NIHMS560371  PMID: 24529517

Abstract

Purpose

We focus on a little-researched issue—how HIV epidemics and programs in key populations in metropolitan areas affect epidemics in other key populations. We consider: 1) How are earlier epidemics among people who inject drugs (PWID) and men who have sex with men (MSM) related to later AIDS incidence and mortality among heterosexuals?; 2) Were prevention programs targeting PWID or MSM associated with lower AIDS incidence and mortality among heterosexuals?; and 3) Was the size of the potential bridge population of non-injecting drug users (NIDUs) in a metropolitan area associated with later AIDS incidence and mortality among heterosexuals?

Methods

Using data for 96 large US metropolitan areas, Poisson regression assessed associations of population prevalences of HIV-infected PWID and MSM (1992); NIDU population prevalence (1992–1994); drug use treatment coverage for PWID (1993); HIV counseling and testing coverage for MSM and for PWID (1992); and syringe exchange presence (2000) with CDC data on AIDS incidence and mortality among heterosexuals in 2006 – 2008, with appropriate socioeconomic controls.

Results

Population density of HIV+ PWID and of NIDUs were positively related, and prevention programs for PWID negatively related, to later AIDS incidence among heterosexuals and later mortality among heterosexuals living with AIDS. HIV+ MSM population density and prevention programs for MSM were not associated with these outcomes.

Conclusions

Efforts to reduce HIV transmission among PWID and NIDUs may reduce AIDS and AIDS-related mortality among heterosexuals. More research is needed at metropolitan area, network and individual levels into HIV bridging across key populations and how interventions in one key population affect HIV epidemics in other key populations.

Keywords: Urban health, metropolitan areas, HIV, AIDS, epidemiology, prevention, bridging, heterosexuals, people who inject drugs, IDU, men who have sex with men, MSM, epidemics

INTRODUCTION

Community network studies, research on group sex events (1, 2) and other data show that sexual relationships among people who inject drugs (PWID), among men who have sex with men (MSM), among non-injection drug users (NIDUs) and among other heterosexuals are common (3–11). NIDUs may be a group through which HIV is transmitted from PWID and MSM to heterosexuals (12–19).

Insofar as we know, little research has been conducted on how epidemics and programs in one key population affect those in other key populations, although one phylogenetic study (20), one historical study and some attempts to use mathematical modeling (21, 22) have explored this issue. Previously, we investigated the association between HIV prevalence among MSM and that among PWID in 96 large metropolitan statistical areas in 1992 (23, 24). Here, in the absence of adequate metropolitan-level data on HIV incidence or prevalence among heterosexuals after 1992, we focus on an important subset of research questions on this topic: 1) How are earlier metropolitan HIV epidemics among PWID and MSM related to later AIDS incidence and mortality among heterosexuals?; 2) Were prevention programs that targeted PWID or MSM associated with lower AIDS incidence and mortality among heterosexuals?; and 3) Was the size of the potential bridge population of non-injecting drug users in a metropolitan area associated with later AIDS incidence and mortality among heterosexuals (25–29)? We use the term “heterosexuals” here to mean heterosexuals who do not inject drugs, although we note that an undetermined proportion of those so classified may have injected drugs but not reported it.

METHODS

We studied these questions using longitudinal data from 1992 – 2008 on a cohort of metropolitan statistical areas (MSAs). The US Census Bureau defines MSAs as contiguous counties containing a central city of 50,000 people or more that form a socioeconomic unity (30); we used MSA boundaries as they were defined in 1992. Our studies of HIV epidemics among PWID at the MSA level (23–25, 31–37) have shown that each MSA has its own epidemic history, HIV prevalence rate, history of prevention programming, and socioeconomic contexts.

This paper is part of a study on HIV epidemics among PWID in the 96 MSAs that had populations of 500,000 or more in 1992. The study design is a longitudinal study at the MSA level of analysis. As such, it can be considered an “ecological cohort” study of MSAs as social and epidemiologic units. Given the complex pathways likely to connect HIV epidemics among different key populations, this design has important strengths and limitations that are described in the Discussion section below. Due to missing data on dependent and key independent variables, the number of MSAs in the models presented in Tables 3 and 4 is less than 96 and varies depending on which variables are included in a given model.

Table 3.

Multivariate results for AIDS incidence rates among heterosexuals (per 10,000 adult population) in large US metropolitan statistical areas, 2006–2008

Model 1: Selected variables from bivariate using IRR (n=85) Model 2: Backward selection of Model 1 using IRR (n=91) Model 3: Model 2 + HIV counseling and testing coverage (n=75) Model 4: Backward selection of Model 1 using AIC (n=85) Model 5: Model 4 + HIV counseling and testing coverage (n=71)
IRR (95% CI) IRR (95% CI) IRR (95% CI) IRR (95% CI) IRR (95% CI)
HIV-positive PWID per 10,000 adult population 1992 1.23*** (1.09, 1.39) 1.32**** (1.20, 1.45) 1.22**** (1.12, 1.33) 1.31**** (1.17, 1.46) 1.27**** (1.13, 1.43)
HIV-positive MSM per 10,000 adult population 1992 1.09 (0.94, 1.26) -- -- -- --
Non-injection drug users per 10,000 adult population 1992–1994 1.11** (1.01, 1.22) 1.30**** (1.22, 1.39) 1.47**** (1.26, 1.72) 1.18**** (1.12, 1.26) 1.24*** (1.05, 1.46)
Gini 1989 1.04 (0.89, 1.23) -- -- -- --
Black/White Dissimilarity Index 1990 1.11 (0.97, 1.26) -- -- 1.11* (0.98, 1.27) 1.15 (0.96, 1.37)
Congregations per 10,000 adult population 1990 0.89 (0.77, 1.04) -- -- -- --
Syringe Exchange Program 2000 0.92 (0.69, 1.23 0.82 (0.60, 1.11) 0.86 (0.62, 1.19) -- --
IDU Drug Use Treatment Coverage 1993 0.64**** (0.52, 0.78) -- -- 0.66***** (0.54, 0.80) 0.65**** (0.53, 0.79)
CTS Coverage PWID 1992 -- -- 0.73**** (0.62, 0.87) -- 0.78** (0.62, 0.98)
CTS Coverage MSM 1992 -- -- -- -- 1.22 (0.91, 1.64)
AIC VALUE 147.77 151.07 125.11 140.78 119.77
Open in a new tab
*

p<0.10

**

p<0.05

***

p<0.01

****

p<0.001

IRRs and 95% CIs for IRRs presented. All p-values and confidence intervals are pseudo-p’s and pseudo CIs, as is discussed in the Methods section of the text; and some variables have substantial measurement error.

+

Estimates of HIV prevalence per 10,000 heterosexual population were omitted from the statistical analysis because they were highly correlated with HIV prevalence estimates for PWID (r = 0.88) and because they were for the hard-to-define “high-risk” heterosexual population rather than the entire heterosexual population, which is the denominator for the outcomes.

Table 4.

Multivariate results for death rates among heterosexuals living with AIDS (per 10,000 adult population) in large US metropolitan statistical areas, 2006–2008

Model 1: Selected variables from bivariate using IRR (n=84) Model 2: Backward selection of Model 1 using IRR (n=91) Model 3: Model 2 + HIV counseling and testing coverage (n=75) Model 4: Backward selection of Model 1 using AIC (n=85) Model 5: Model 4 + HIV counseling and testing coverage (n=72)
IRR (95% CI) IRR (95% CI) IRR (95% CI) IRR (95% CI) IRR (95% CI)
HIV-positive PWID per 10,000 adult population 1992 1.23*** (1.06, 1.42)-- 1.24**** (1.10, 1.40) 1.23**** (1.09, 1.40) 1.43**** (1.24, 1.64) 1.35**** (1.16, 1.57)
HIV-positive MSM per 10,000 adult population 1992 1.07 (0.89, 1.28) -- -- --
Non-injection drug users per 10,000 adult population 1992–1994 1.10* (0.99, 1.23) -- -- 1.30**** (1.21, 1.39) 1.50**** (1.19, 1.88)
Gini 1989 1.20** (1.01, 1.42) 1.31**** (1.15, 1.50) 1.26**** (1.10, 1.43) -- --
Black/White Dissimilarity Index 1990 1.09 (0.89, 1.34) -- -- -- --
Congregations per 10,000 adult population 1990 0.80** (0.65, 0.97) 0.72**** (0.60, 0.86) 0.71*** (0.57, 0.89) -- --
Syringe Exchange Program 2000 0.69** (0.51, 0.96) 0.71** (0.54, 0.93) 0.77* (0.56, 1.05) -- --
Hard Drug Arrests per 10,000 adult population 1993 1.04 (0.84, 1,28) -- -- -- --
IDU Drug Use Treatment Coverage 1993 0.53**** (0.42, 0.67) 0.51**** (0.42, 0.61) 0.51**** (0.40, 0.67) 0.55**** (0.42, 0.71) 0.58**** (0.42, 0.80)
CTS Coverage PWID 1992 -- -- 0.81** (0.67, 1.00) -- 0.79* (0.61, 1.01)
AIC VALUE 90.03 86.78 72.99 82.23 68.68
Open in a new tab
*

p<0.10

**

p<0.05

***

p<0.01

****

p<0.001

IRRs and 95% CIs for IRRs presented. All p-values and confidence intervals are pseudo-p’s and pseudo CIs, as is discussed in the Methods section of the text; and some variables have substantial measurement error.

+

Estimates of HIV prevalence per 10,000 heterosexual population were omitted from the statistical analysis because they were highly correlated with HIV prevalence estimates for PWID (r = 0.88) and because they were for the hard-to-define “high-risk” heterosexual population rather than the entire heterosexual population, which is the denominator for the outcomes.

Data

Data on both outcomes (AIDS incidence; and mortality among heterosexuals aged 15 – 64 years living with AIDS) for 2006 – 2008 in each MSA were obtained by special request from the US Centers for Disease Control and Prevention. We divided these by the number of adults (aged 15–64) living in each MSA to calculate population-based rates.

Other data sources are described in Table 1. Estimates of numbers of HIV-positive MSM and PWID and on population numbers of MSM and PWID in 1992 were taken from Holmberg (38). Although reported in Table 1, data on HIV-positive high-risk heterosexuals per 10,000 adult population were not included in the statistical analyses because they were highly correlated with HIV prevalence estimates for PWID (r = 0.88) and because they were for the hard-to-define “high-risk” heterosexual population rather than the entire heterosexual population, which is who this paper is studying. These were calculated using methods such as back-calculation that have become more complicated after the development and spread of highly active anti-retroviral therapy (HAART) (39). Although these techniques have been used to create national-level incidence estimates (40), they would be very time-consuming and less accurate at the metropolitan area level. Although estimates for numbers of PWID for these metropolitan areas are available (see Table 1), estimates of numbers of MSM are available only for 1992 and estimates of HIV incidence among heterosexuals for metropolitan areas are not available after 1992.

Table 1.

Descriptive statistics on independent variables for 96 large US Metropolitan Statistical Areas

N Median (Range) Mean (SD) 1st Quartile 3rd Quartile Data Source
Epidemiologic Factors
HIV-positive high-risk heterosexuals per 10,000 adult population (1992)* 96 1.35 (0.20–19.14) 2.63 (3.48) 0.83 2.60 Holmberg (38)
HIV-positive PWID per 10,000 adult population (1992)* 96 4.06 (0.74–80.63) 9.20 (13.88) 2.28 9.97 Holmberg (38)
HIV-positive MSM per 10,000 adult population (1992)* 96 11.25 (2.64–134.07) 15.63 (15.29) 7.97 17.62 Holmberg (38)
Non-injection drug users per 10,000 adult population (1992–1994) 92 456.88 (90.30–3371.61) 598.52 (484.34) 319.50 651.48 TEDS (88) and Tempalski et al (29)
Economic Conditions
Household Gini coefficient (1989) 95 0.43 (0.38–0.51) 0.43 (0.02) 0.41 0.44 Harper (89)
Percent Living Below Poverty Level (1989) 96 11.15 (4.23–52.87) 11.92 (5.61) 9.47 13.15 US Census (90, 91)
Racial/ethnic Residential Segregation
Black/White Dissimilarity Index (1990)** 90 64.31 (37.52–89.95) 64.49 (11.86) 56.26 73.21 Mumford Center (92)
Social Cohesion
Religious Membership per 10,000 adult population (1990) 95 1314.22 (294.65–4945.15) 1669.51 (1036.00) 859.28 2148.30 ARDA (93)
Congregations per 10,000 adult population ARDA (93)
Congregations per 10,000 adult population (1990) 95 6.16 (3.19–15.91) 7.01 (2.88) 5.08 8.07 ARDA (93)
Interventions
Syringe Exchange Program (2000) 96 0.00 (0–1) 0.44 (0.50) - - Beth Israel Medical Center (94)
Hard Drug Arrests per 10,000 adult population (1993)*** 94 11.42 (0.53–71.87) 15.06 (14.18) 4.47 20.41 FBI (95)
Drug Use Treatment Coverage for PWID (% among total PWID pop in 1993) 90 5.60 (0.80–16.40) 6.76 (3.71) 4.20 9.40 Tempalski et al. 2010 (35)
HIV Counseling and Testing Coverage for PWID (% among PWID at risk in 1992) 79 7.70 (0.14–31.51) 9.16 (6.42) 4.78 12.52 CTS (96) and Holmberg (38)
HIV Counseling and Testing Coverage for MSM (% among MSM at risk in 1992) 81 7.09 (0.03–31.38) 7.78 (5.00) 5.00 8.74 CTS (96) and Holmberg (38)
Open in a new tab

Reminder: Some variables have substantial measurement error.

*

Holmberg (1996) has already reported on these variables.

**

Values on the Index of Dissimilarity are independent of the number of Black and White residents in each MSA.

***

These arrests are an intervention to remove drug users from the street; deter drug use; and perhaps provide medical and social services to those arrested. See Friedman et al. (2006, 2011 2013) for data on their relationship to key parameters for people who inject drugs (26, 27, 34).

Non-injection drug users (NIDUs) were defined as people who use heroin, cocaine, non-prescription methadone or amphetamines but do not inject drugs. The population prevalence of NIDUs in 1992–1994 was estimated by multiplying PWID population prevalence by an adjustment factor: the ratio of events (reported by TEDS, as cited in Table 1) in which NIDUs entered drug use treatment divided by the sum of the numbers of events in which PWID or NIDUs entered treatment.

CDC data on numbers of HIV testing and counseling sessions (CTS) in 1992 for MSM and PWID were used to estimate “coverage rates” for CDC-funded CTS for these populations. These coverage rates are defined as the number of testing events performed per member of the population in question (and expressed as a percent). Data on syringe exchange program presence (SEPs) come from the Beth Israel Syringe Exchange Survey for 2000; many of them were established previous to that (36, 37). Other intervention data are described in Table 1.

The time-ordering of variables is important. Our focus here is on associations between the size of the HIV epidemics among MSM and among PWID in 1992 and AIDS-related outcomes in 2006 – 2008. We also want to determine how prevention programs affected these outcomes. As discussed above, HIV counseling and testing coverage for MSM could only be calculated for 1992, so we used PWID coverage data for this year as well. Since the number of US syringe exchanges increased rapidly in the 1990s (and since SEPs are important not only as way to prevent HIV transmission, but also as major referral sources for both NIDUs and PWID to drug use treatment programs and to HIV care that would slow disease progression (41–43)), we used 2000 as our year for measuring this variable even though there would be few if any instances in which infections prevented among PWID in 2000 would transmit HIV to a heterosexual who would develop AIDS by 2006.

Our previous research on HIV and AIDS among PWID in these MSAs (23, 24, 26, 27, 31, 34) has indicated that a number of social and economic variables (see Table 1) should be controlled for. Some of these were available only for Census years 1990 and 2000. In these cases, their values for the two years were correlated with r > 0.90, so we entered the earlier value into the statistical model building process. Similarly, for treatment coverage for PWID, the values for 1993 and 2000 had r = .76; we used the 1993 value which had a higher correlation with the dependent variables.

Statistical approach

Since this is a study of all US MSAs with populations of 500,000 or more in 1992 that had data available on our key variables, our sample is a fully enumerated universe. This means there is no sampling error. Nonetheless, we report statistical significance and confidence intervals as heuristic guides to the importance of a variable in an equation (27, 44). We compute them as if we had a random sample of MSAs, but interpret results as “pseudo-p-values” and “pseudo-CIs” to guide our interpretation (as in previous articles: (23, 31, 45)). In addition, a number of our variables have substantial measurement error (as is true in much of epidemiology and related fields)—which standard statistics do not take account of—and this may affect the accuracy of our results.

Poisson regression with a log link function, used to model count data (46), was used for bivariate and multivariate analyses of AIDS incidence and mortality rates (per 10,000 adult population) in 2006–2008. As suggested by Stokes and colleagues (47) as well as Cameron and Trivedi (48), a modified Poisson approach, using the Repeated statement in SAS PROC GENMOD, was used to estimate models with robust standard errors. Since we did find evidence of underdispersion, our use of robust standard errors was an appropriate correction that resulted in smaller standard errors and narrower confidence intervals(47, 48).

Continuous independent variables were rescaled before inclusion in bivariate or multivariate analysis. The standard deviation of the predictor was chosen as the scaling distance, which standardized continuous predictors such that a one-unit change represented a change of one standard deviation (49). Rescaling predictors allowed for more meaningful interpretation of regression coefficients (49). Regression coefficients were exponentiated to obtain incidence-rate ratios (IRR) for independent variables to ease interpretation of results.

Bivariate relationships between each independent variable and outcome measure were assessed using Poisson regression (46). For each dependent variable, predictors with an IRR<=0.83 or an IRR>=1.20 were entered into multivariate models unless collinearity issues were present (45). These IRR values were chosen to select variables with moderate associations for further exploratory analysis. For variables with data at more than one time point, earlier years were generally selected for use in multivariate models. In one case, SEPs, we used a later year (2000) to reflect the large growth in SEPs during the 1990s (37, 50, 51). These results constitute Model 1 in each of Tables 2 and 3.

Table 2.

Bivariate results for AIDS incidence and death rates among heterosexuals (per 10,000 adult population) in large US metropolitan statistical areas, 2006–2008

AIDS Incidence Rates AIDS Mortality Rates
IRR (95% CI) IRR (95% CI)
Epidemiologic Factors
HIV-positive PWID per 10,000 adult population 1992 1.37**** (1.25, 1.50) 1.44**** (1.27, 1.63)
HIV-positive MSM per 10,000 adult population 1992 1.21* (0.99, 1.46) 1.24** (1.00, 1.55)
Non-injection drug users per 10,000 adult population 1992–1994 1.36**** (1.26, 1.47) 1.51**** (1.40, 1.64)
Economic Conditions
Gini 1989 1.47**** (1.27, 1.71) 1.77**** (1.44, 2.18)
Percent Living Below Poverty 1989 1.14 (0.84, 1.55) 1.26 (0.82, 1.91)
Racial Disparity
Black/White Dissimilarity Index 1990 1.32**** (1.12, 1.56) 1.41*** (1.13, 1.77)
Social Cohesion
Religious Membership per 10,000 adult population 1990 0.96 (0.81, 1.14) 0.90 (0.71, 1.14)
Congregations per 10,000 adult population 1990 0.78** (0.62, 0.98) 0.64** (0.43, 0.95)
Interventions
Syringe Exchange Program 2000 0.73* (0.49, 1.07) 0.50*** (0.31, 0.83)
Hard Drug Arrests per 10,000 adult population 1993 1.18* (1.00, 1.41) 1.26** (1.01, 1.58)
IDU Drug Use Treatment Coverage 1993 0.61**** (0.49, 0.77) 0.46**** (0.33, 0.65)
CTS Coverage PWID 1992 0.62**** (0.50, 0.77) 0.52**** (0.37, 0.73)
CTS Coverage MSM 1992 0.86 (0.66, 1.13) 0.77* (0.59, 1.02)
Open in a new tab
*

p<0.10

**

p<0.05

***

p<0.01

****

p<0.001

IRRs and 95% CIs for IRRs presented. All p-values and confidence intervals are pseudo-p’s and pseudo CIs, as is discussed in the Methods section of the text; and some variables have substantial measurement error.

+

Estimates of HIV prevalence per 10,000 heterosexual population were omitted from the statistical analysis because they were highly correlated with HIV prevalence estimates for PWID (r = 0.88) and because they were for the hard-to-define “high-risk” heterosexual population rather than the entire heterosexual population, which is the denominator for the outcomes.

Since existing theory and research has relatively little to say about the cross-population processes being studied, we used exploratory analytic techniques to study these relationships by conducting backwards model selection based on incidence-rate ratios. Using the same IRR cutoffs as in the bivariate analysis, predictors were eliminated one at time until all variables in the model had an IRR<=0.83 or an IRR>=1.20. The minimal equation based on these specifications is presented as Model 2. Since data on HIV counseling and testing coverage were not available for over fifteen of our MSAs (see Table 1), these variables were added to Model 2 to form a separate model (Model 3) based on this smaller data set.

As a further exploratory step, we estimated an additional Model 4 using Akaike information criteria (AIC) for model selection (52). AIC combines estimation and model selection and is particularly useful when comparing multiple models (53, 54). Assessing AIC involves comparing AIC values to a minimum AIC or “best” model using the formula: Δi = AICi − AICmin; models with Δi ≤ 2 demonstrate substantial support, models with 4 ≤ Δi ≤ 7 have less support, and models where Δi > 10 have essentially no support (54). AIC was used since: 1) it let us compare multiple models in an exploratory way; 2) it let us compare models that were not nested; 3) it helped us identify and avoid over-fitted models; while 4) avoiding choice of models that overly-restrict the number of variables included because they use model fit statistics like the Bayes Information Criterion (55) that more heavily penalize extra variables. Starting with Model 1, one variable at a time was removed and AIC values were compared. The model with the lowest AIC value was selected before moving to the next round of variable removal. Model 4 in Tables 2 and 3 represents the minimal equation with the lowest AIC value. HIV counseling and testing coverage was added to the final model produced through AIC selection to produce Model 5.

RESULTS

There are wide variations across MSAs in HIV population prevalence rates among MSM, PWID and heterosexuals in 1992 (Table 1). In bivariate analyses (Table 2), a wide range of epidemiologic, economic, racial disparity, social cohesion, and intervention indicators were associated with subsequent AIDS incidence rates among heterosexuals and death rates among heterosexuals living with AIDS. Many of these variables were intercorrelated, so additional exploratory analyses were conducted. (See Table 3 for AIDS incidence results and Table 4 for mortality results).

Using the IRR criteria (IRR >= 1.20 or IRR <= 0.83), Model 2 shows that AIDS incidence rates among heterosexuals were higher where there was a higher population prevalence of HIV-positive PWID in 1992, where there was a higher population prevalence of NIDUs in 1992 – 1994, and where there were no syringe exchange programs in 2000 (Table 3).

In Model 3, we added data on CDC-funded HIV counseling and testing program coverage for PWID (and for MSM, data not shown). HIV counseling and testing coverage for PWID was associated with lower AIDS incidence among heterosexuals, though coverage for MSM was not. The reduced sample size in Model 3 (N = 75 in models that include HIV counseling and testing coverage) increased the incidence-rate ratio for syringe exchange programs to slightly over the criterion value of 0.83.

The alternative approach to model reduction, using AIC rather than IRR as the criterion for backward selection, yielded a somewhat different model (Model 4). As in Model 2, AIDS incidence rates among heterosexuals were higher where there was a higher population prevalence of HIV-positive PWID in 1992 and where there was a higher population prevalence of NIDUs in 1992 – 1994 (although the IRR on this variable (1.18) was < 1.20). Drug use treatment coverage rates for PWID in 1993 were associated with less AIDS incidence; and Black/White Residential Dissimilarity in 1990 was slightly (1.11) associated with higher AIDS incidence. When HIV counseling and testing coverage for PWID and for MSM were added to the model, counseling and testing for PWID was protective (IRR = 0.78); but CTS for MSM was not protective. (Its apparent association with higher rates of heterosexual AIDS incidence is probably an artifact of high correlations with CTS for PWID [r = 0.50], PWID drug use treatment coverage [r = .39] and Black/White dissimilarity index [r = −0.36]).

In Model 2, AIDS mortality rates among heterosexuals in 2006–2008 were higher where there was a higher population prevalence of HIV-positive PWID in 1992 or a higher Gini coefficient of income inequality; and lower where there was one or more syringe exchange programs in 2000, higher drug use treatment coverage for PWID in 1993, and more religious congregations per capita in 1990 (Table 4). In Model 3, with reduced N, HIV counseling and testing coverage for PWID in 1992 was associated with lower mortality rates in 2006 – 2008.

In the alternative Model 4, using the AIC criterion for model selection, a higher population prevalence of HIV-positive PWID in 1992 and lower drug use treatment rates among PWID remained associated with higher mortality. The only other predictor in this equation, however, was a higher population prevalence of NIDUs in 1992 – 1994. In Model 5, these variables remained in the equation and higher coverage rates of HIV counseling and testing among PWID in 1992 was associated with lower mortality rates.

DISCUSSION

These analyses suggest the possibility that AIDS burdens among heterosexuals in 2006 – 2008 may have been shaped by bridging from HIV-infected PWID to heterosexuals and possibly, though to a lesser degree, from HIV-positive MSM to heterosexuals. Such bridging may have taken the form of sexual transmission from PWID (or MSM) to NIDUs and then, perhaps, from NIDUs to (other) heterosexuals. Network studies have shown that many PWID and many MSM have sex with NIDUs; and that NIDUs have sex with people who do not use injectable drugs (8, 11).

Programs to reduce HIV transmission to and from PWID are statistically associated in this study with smaller later AIDS epidemics among heterosexuals. In our exploratory analyses using different methods of model selection, there were some differences in which programs entered into which models. Overall, however, the data suggest that syringe exchange, drug use treatment coverage for PWID, and HIV counseling and testing coverage for PWID all were associated with lower rates of AIDS incidence among heterosexuals and with lower mortality rates among heterosexuals living with AIDS in 2006 – 2008, which may be an example of combination prevention (56–60) across population groups. Although our ecological cohort design cannot establish causality, these findings are consistent with the possibility that these relationships were causal of MSA-specific rates (but clearly not of individual infection or disease progression); we suggest that further research be conducted to clarify possible causal patterns. Possible pathways might include these programs leading to lower HIV incidence among PWID or to their facilitating access to successful antiretroviral therapy and thus to lower viral load among HIV-positive PWID, making the population of PWID less likely to transmit HIV to heterosexuals. SEPs also reduce sexual risk behaviors among PWID (61, 62), as does knowing that one is HIV-positive, thus making HIV-positive PWID less likely to transmit HIV to heterosexual NIDUs or other heterosexuals. These programs may possibly have also helped produce a programmatic and social context where infected heterosexual NIDUs or other heterosexuals were more able to get and to adhere to antiretroviral and other treatment, which could rapidly reduce probabilities of developing AIDS or of dying from it thereafter (63, 64).

Counseling and testing for MSM, however, did not seem to be associated with reduced rates of either AIDS incidence or mortality among heterosexuals. Given the continued high rates of HIV transmission among MSM (as opposed to among PWID, where transmission has fallen, in part due to the high effectiveness of syringe exchange programs where they have been implemented (43, 65–67)), this could imply that counseling and testing among MSM has had limited effect upon transmission behaviors by infected MSM—perhaps because it has not been reaching the highest-risk MSM (11, 68, 69). It also is possible that there is a correlation between high rates of counseling and testing services among MSM and better prevention or medical services for those heterosexuals more likely to have sex with MSM. Further research is clearly needed to understand these results.

To our knowledge this is one of the first empirical analyses to discover a relationship between racial/ethnic residential segregation, a form of institutional racism (70) at the supra-individual level, and rates of HIV or other STIs. Biello et al have recently found that MSAs with higher levels of two dimensions of segregation (isolation and unevenness) had higher rates of gonorrhea among black residents (71). Segregation may generate vulnerability to HIV and other STIs among black adults through several mechanisms, including by affecting the structure of sexual networks (72–78). Also, segregation may limit access to healthcare (74), and thus reduce access to HIV testing and counseling and to therapies that slow the progression of HIV infection.

There is some evidence in these data that income inequality may be associated with greater mortality among heterosexuals living with AIDS and that social cohesion (as indicated by religious congregation population density) may be protective. Such findings are parallel to other studies of social causation of mortality (79–84).

Our findings are subject to several limitations. Causal mechanisms are hard to study at a single level of analysis: Both higher-level variables (such as economic changes or changes in health systems at the global, national or regional level) and lower-level variables (such as locally-generated behavioral fads) may affect observed relationships. It is important both to avoid interpretations that make the ecological fallacy and to consider interpretations that are valid at a single level of analysis. We have only limited ability to study how our independent variables come to be associated with outcomes. As with many studies, including some on social determinants of health (85–87), we cannot specify mechanisms through which differences in MSA characteristics like NIDU prevalence or racial residential segregation might affect AIDS incidence or mortality among heterosexuals. Similarly, we cannot specify the time lags between cause and effect for any pathways that might be causal, which could mean that the effects of slower-acting causal variables would be missed in these analyses. Also, all of our variables are subject to measurement error, which may limit the accuracy of our analyses, and our estimates of numbers of NIDUs have not been subjected to rigorous validation. In constructing the dependent variables, for example, we divided by the total adult population to create rate estimates because no data on MSM populations in 2006 – 2008 were available. This may have overestimated the dependent variables for MSAs with particularly large MSM populations. Since data on deaths from AIDS were not available, we used mortality among people living with AIDS as our denominator—which will erroneously include some deaths unrelated to the disease. If missing data was correlated with variables we studied, this might have affected observed relationships. Further research using data for all years during this period for all variables might also strengthen our analytic ability to study these issues—but this would require having estimates for MSM HIV prevalence and population prevalence that would be difficult to estimate for the MSAs.

This is a study of the universe of large US MSAs with available data rather than of a sample. Samples let researchers make statistical inferences to indicate the uncertainty about whether their findings accurately describe relationships within the universe from which the sample was taken. Our findings are descriptive of the relationships of the measured variables in this universe. They do not, however, imply that these findings can necessarily be extended to smaller MSAs, non-metropolitan localities, other time periods or other countries

Associations between the independent variables and the outcome variables might come about through any of the following processes: 1. They might be spurious—that is, both the values of the independent variables and those of the dependent variables might result from some prior unobserved causal variable; 2. The independent variables might have affected rates of HIV incidence among heterosexuals; or 3. The independent variables might have affected rates of access to and adherence to antiretroviral therapy or in other ways affected rates of disease progression to AIDS or death among heterosexuals infected with HIV.

Our ability to study sexual bridging from MSM and PWID through heterosexual (and bisexual) NIDUs and then to other heterosexuals is greatly limited by the lack of data about whether people diagnosed with AIDS are non-injecting drug users. (This leads to inability to determine whether heterosexuals with AIDS who die were NIDUs.) This is because the US HIV surveillance system does not collect data on whether people with HIV are NIDUs.

This exploratory study at the MSA level of analysis showed that efforts to reduce HIV transmission among PWID and NIDUs were associated with reduced AIDS and AIDS-related mortality among heterosexuals. Research on bridging mechanisms and pathways at the community, network and individual levels, including investigation of possible alternative MSA-level predictors that might produce these associations, is clearly needed. Research into whether interventions in one key population affect HIV epidemics in other key populations will be of high policy relevance and should be a priority.

Acknowledgments

This work was supported by National Institute of Drug Abuse grants # R01 DA013336 (Community Vulnerability and Responses to Drug-User-Related HIV/AIDS), R01 DA 003574 (Risk Factors for AIDS among Intravenous Drug Users), and 5T32 DA007233 (Behavioral Sciences Training in Drug Abuse Research program sponsored by Public Health Solutions and National Development and Research Institutes). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

LIST OF ABBREVIATIONS

MSA

metropolitan statistical area

PWID

people who inject drugs

MSM

men who have sex with men

NIDU

non-injecting drug user

HAART

highy active anti-retroviral therapy

CTS

HIV counseling and testing

SEP

syringe exchange program

IRR

incidence-rate ratios

AIC

Akaike information criteria

Footnotes

Disclaimers: The findings and conclusions in this study are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention or the National Institute on Drug Abuse.

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.Friedman SR, Bolyard M, Khan M, Maslow C, Sandoval M, Mateu-Gelabert P, et al. Group Sex Events and HIV/STI Risk in an Urban Network. J Acq Immun Syn 2008. 2008;49:440–6. doi: 10.1097/qai.0b013e3181893f31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Friedman SR, Mateu-Gelabert P, Sandoval M. Group sex events amongst non-gay drug users: An understudied risk environment. Int J Drug Policy. 2011 Jan;22(1):1–8. doi: 10.1016/j.drugpo.2010.06.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Adimora AA, Schoenbach VJ, Doherty IA. HIV and African Americans in the Southern United States: Sexual Networks and Social Context. Sexually Transmitted Diseases. 2006;33(7):S39–S45. doi: 10.1097/01.olq.0000228298.07826.68. [DOI] [PubMed] [Google Scholar]
  • 4.Des Jarlais DC, Arasteh K, McKnight C, Perlman D, Hagan H, Semaan S, et al. Gender and age patterns in HSV-2 and HIV infection among non-injecting drug users in New York City. Sex Transm Dis. 2010 Oct;37(10):637–43. doi: 10.1097/OLQ.0b013e3181e1a64a. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Des Jarlais DC, Chamberland ME, Yancovitz SR, Weinberg P, Friedman SR. Heterosexual partners: A large risk group for AIDS. Lancet. 1984;841:1346–7. doi: 10.1016/s0140-6736(84)90861-4. [DOI] [PubMed] [Google Scholar]
  • 6.Friedman S, Bolyard M, Mateu-Gelabert P, Goltzman P, Pawlowicz M, Zunino Singh D, et al. Some data-driven reflections on priorities in AIDS network research. AIDS and Behavior 2007. 2007;11:641–51. doi: 10.1007/s10461-006-9166-7. [DOI] [PubMed] [Google Scholar]
  • 7.Friedman SR, Des Jarlais DC, Ward TP, Jose B, Neaigus A, Goldstein M. Drug Injectors and Heterosexual AIDS. In: Sherr L, editor. AIDS and the Heterosexual Population. Harwood Academic Publishers; 1994. pp. 41–65. [Google Scholar]
  • 8.Millett GA, Peterson JL, Flores SA, Hart TA, Jeffries WL, Wilson PA, et al. Comparisons of disparities and risks of HIV infection in black and other men who have sex with men in Canada, UK, and USA: a meta-analysis. Lancet. 2012 Jul 28;380(9839):341–8. doi: 10.1016/S0140-6736(12)60899-X. [DOI] [PubMed] [Google Scholar]
  • 9.Neaigus A, Miller M, Gyarmathy VA, Friedman SR. HIV Heterosexual Sexual Risk From Injecting Drug Users Among HIV-Seronegative Noninjecting Heroin Users. Subst Use Misuse. 2011;46(2–3):208–17. doi: 10.3109/10826084.2011.521473. [DOI] [PubMed] [Google Scholar]
  • 10.Polk S, Ellen JM, Fichtenberg C, Huettner S, Jennings JM. HIV Prevalence Overall and among High-HIV-Risk Behaviorally Defined Subgroups among Heterosexuals at Community-Based Venues in a Mid-Atlantic, US City. J Urban Health. 2012 Nov 8;:1–11. doi: 10.1007/s11524-012-9776-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Shoptaw S, Reback CJ. Methamphetamine use and infectious disease-related behaviors in men who have sex with men: implications for interventions. Addiction. 2007 Apr;102(Suppl 1):130–5. doi: 10.1111/j.1360-0443.2006.01775.x. [DOI] [PubMed] [Google Scholar]
  • 12.Mills H, Colijn C, Vickerman P, Heimer R. HIV transmission from drug injectors to partners who do not inject, and beyond: Modelling the potential for a generalized heterosexual epidemic in St. Petersburg, Russia. Drug Alcohol Depend. 2013 doi: 10.1016/j.drugalcdep.2013.04.028. pii: S0376-8716(13)00170-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Strathdee SA, Stockman JK. Epidemiology of HIV Among Injecting and Non-injecting Drug Users: Current Trends and Implications for Interventions. Curr HIV/AIDS Rep. 2010;7(2):99–106. doi: 10.1007/s11904-010-0043-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Uusküla A, McMahon J, Kals M, Talu A, Abel-Ollo K, Rüütel K, et al. Risk for Heterosexual HIV Transmission Among Non-Injecting Female Partners of Injection Drug Users in Estonia. AIDS Behav. 2013;17(3):879–88. doi: 10.1007/s10461-011-0078-9. [DOI] [PubMed] [Google Scholar]
  • 15.Volz E, Frost SD, Rothenberg R, Meyers L. Epidemiological bridging by injection drug use drives an early HIV epidemic. Epidemics. 2010;2(3):155–64. doi: 10.1016/j.epidem.2010.06.003. [DOI] [PubMed] [Google Scholar]
  • 16.Rawstorne P, Digiusto E, Worth H, Zablotska I. Associations between crystal methamphetamine use and potentially unsafe sexual activity among gay men in Australia. Arch Sex Behav. 2007;36:646–54. doi: 10.1007/s10508-007-9206-z. [DOI] [PubMed] [Google Scholar]
  • 17.Hacker MA, Leite I, Friedman SR, Carrijo RG, Bastos FI. Poverty, bridging between injecting drug users and the general population, and “interiorization” may explain the spread of HIV in southern Brazil. Health Place. 2009 Jun;15(2):514–9. doi: 10.1016/j.healthplace.2008.09.011. [DOI] [PubMed] [Google Scholar]
  • 18.Mausbach B, Semple S, Strathdee S, Zians J, Patterson T. Efficacy of a behavioral intervention for increasing safer sex behaviors in HIV-negative, heterosexual methamphetamine users: results from the Fast-Lane Study. Ann Behav Med. 2007;34:263–74. doi: 10.1007/BF02874551. [DOI] [PubMed] [Google Scholar]
  • 19.Des Jarlais DC, Freelemyer JP, Modi SN, Arasteh K, Mathers BM, et al. Transitions from Injection-Drug-Use-Concentrated to Self-Sustaining Heterosexual HIV Epidemics: Patterns in the International Data. Plos One. 2012;7:e31227. doi: 10.1371/journal.pone.0031227. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Graw F, Leitner T, Ribeiro R. Agent-based and phylogenetic analyses reveal how HIV-1 moves between risk groups: Injecting drug users sustain the heterosexual epidemic in Latvia. Epidemics. 2012;4:104–16. doi: 10.1016/j.epidem.2012.04.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Grassly NC, Lowndes CM, Rhodes T, Judd A, Renton A, Garnett GP. Modelling emerging HIV epidemics: the role of injecting drug use and sexual transmission in the Russian Federation, China and India. Int J Drug Policy. 2003;14(1):25–43. [Google Scholar]
  • 22.Saidel TJ, Des Jarlais DC, Peerapatanapokin W, Dorabjee J, Singh S, Brown T. Potential impact of HIV among IDUs on heterosexual transmission in Asian settings: scenarios from the Asia Epidemic Model. Int J Drug Policy. 2003;14(1):63–74. [Google Scholar]
  • 23.Friedman SR, Perlis T, Des Jarlais DC. Laws prohibiting over-the-counter syringe sales to injection drug users: Relations to population density, HIV prevalence, and HIV incidence. American Journal of Public Health 2001. 2001a;91(5):791–3. doi: 10.2105/ajph.91.5.791. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Friedman SR, Perlis TE, Lynch J, Des Jarlais DC. Economic inequality, poverty, and laws against syringe access as predictors of metropolitan area rates of drug injection and HIV infection. Global Research Network Meeting on HIV Prevention in Drug-Using Populations. 2001:147–9. [Google Scholar]
  • 25.Chatterjee S, Tempalski B, Pouget ER, Cooper HL, Cleland CM, Friedman SR. Changes in the prevalence of injection drug use among adolescents and young adults in large U.S. metropolitan areas. AIDS Behav. 2011 Oct;15(7):1570–8. doi: 10.1007/s10461-011-9992-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Friedman SR, Pouget ER, Chatterjee S, Cleland CM, Tempalski B, Brady JE, et al. Drug arrests and injection drug deterrence. Am J Public Health. 2011 Feb;101(2):344–9. doi: 10.2105/AJPH.2010.191759. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Friedman SR, West BS, Pouget ER, Hall HI, Cantrell J, Tempalski B, et al. Metropolitan Social Environments and Pre-HAART/HAART Era Changes in Mortality Rates (per 10,000 Adult Residents) among Injection Drug Users Living with AIDS. Plos One. 2013;8(2):e57201. doi: 10.1371/journal.pone.0057201. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Pouget ER, Friedman SR, Cleland CM, Tempalski B, Cooper HLF. Estimates of the population prevalence of injection drug users among Hispanic residents of large U.S. metropolitan areas. J Urban Health. 2012;89(3):527–64. doi: 10.1007/s11524-012-9670-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Tempalski B, Pouget E, Cleland C, Brady J, Cooper H, Hall H, et al. Trends in the population prevalence of people who inject drugs in US Metropolitan Areas 1992–2007. Plos One. 2013;8:e64789. doi: 10.1371/journal.pone.0064789. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Office of Management and Budget. Standards for defining metropolitan and micropolitan statistical areas. Federal Register. 2000;65:8228–82238. [Google Scholar]
  • 31.Cooper H, Friedman SR, Tempalski B, Friedman R. Residential segregation and the prevalence of injection drug use among Black adult residents of US metropolitan areas. American Journal of Public Health. 2007;96(2):344–52. doi: 10.2105/AJPH.2005.074542. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Cooper HLF, Brady JE, Friedman SR, Tempalski B, Gostnell K, Flom PL. Estimating the prevalence of injection drug use among Black and White adults in large US metropolitan areas over time (1992–2002): Estimation methods and prevalence trends. Journal of Urban Health. 2008;85:826–56. doi: 10.1007/s11524-008-9304-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Cooper HLF, Friedman S, Tempalski B, Friedman R, Keem M. Racial/ethnic disparities in injection drug use in 94 U.S. metropolitan statistical areas in 1998. Annals of Epidemiology. 2005;15(5):326–34. doi: 10.1016/j.annepidem.2004.10.008. [DOI] [PubMed] [Google Scholar]
  • 34.Friedman SR, Cooper HLF, Tempalski B, Keem M, Friedman R, Flom PL, et al. Relationships of deterrence and law enforcement to drug-related harms among drug injectors in US metropolitan areas. Aids. 2006;20:93–9. doi: 10.1097/01.aids.0000196176.65551.a3. [DOI] [PubMed] [Google Scholar]
  • 35.Tempalski B, Cleland CM, Pouget ER, Chatterjee S, Friedman SR. Persistence of low drug treatment coverage for injection drug users in large US metropolitan areas. Subst Abuse Treat Prev Policy. 2010;5:23. doi: 10.1186/1747-597X-5-23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Tempalski B, Cooper H, Friedman SR, Des Jarlais DC, Brady J. Correlates of syringe coverage for heroin injection in 35 large metropolitan areas in the US in which heroin is the dominant injected drug. International Journal of Drug Policy 2008. 2008;19S:S47–S58. doi: 10.1016/j.drugpo.2007.11.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Tempalski B, Flom PL, Friedman SR, Des Jarlais DC, Friedman JJ, McKnight C, et al. Social and political factors predicting the presence of syringe exchange programs in 96 Metropolitan areas in the United States. American Journal of Public Health 2007. 2007;97(3):437–47. doi: 10.2105/AJPH.2005.065961. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Holmberg SD. The estimated prevalence and incidence of HIV in 96 large US metropolitan areas. American Journal of Public Health. 1996;86(5):642–54. doi: 10.2105/ajph.86.5.642. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Hall HI, Song R, Rhodes P, Prejean J, An Q, Lee LM, et al. Estimation of HIV Incidence in the United States. JAMA 2008. 2008;300:520–9. doi: 10.1001/jama.300.5.520. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Hall HI, Song R, Rhodes P, Prejean J, An O, Lee LM, et al. Estimation of HIV Incidence in the United States. JAMA. 2008;300(5):520–9. doi: 10.1001/jama.300.5.520. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Heinzerling KG, Kral AH, Flynn NM, Anderson R, Scott A, Gilbert ML, et al. Unmet need for recommended preventive health services among clients of California syringe exchange programs: implications for quality improvement. Drug and alcohol dependence. 2006;81(2):167–78. doi: 10.1016/j.drugalcdep.2005.06.008. [DOI] [PubMed] [Google Scholar]
  • 42.Tempalski B, Des Jarlais DC, Friedman SR, Burris S. Spatial variations of legality, funding and services among US SEPs. 14th International Conference on the Reduction of Drug Related Harm; April 6–10, 2003; Chiang Mai, Thailand. [Google Scholar]
  • 43.MacDonald M, Law M, Kaldor J, Hales J, Dore G. Effectiveness of needle and syringe programmes for preventing HIV transmission. International Journal of Drug Policy 2003. 2003;14(5–6):351–2. [Google Scholar]
  • 44.Bluthenthal RN. Syringe exchange as a social movement: a case study of harm reduction in Oakland, California. Subst Use Misuse. 1998 Apr;33(5):1147–71. doi: 10.3109/10826089809062212. [DOI] [PubMed] [Google Scholar]
  • 45.Friedman SR, Tempalski B, Brady J, Friedman JJ, Cooper H, Flom PL, et al. Predictors of the degree of drug treatment coverage for injection drug users in 94 metropolitan areas in the United States. International Journal of Drug Policy 2007. 2007;18:475–85. doi: 10.1016/j.drugpo.2006.10.004. [DOI] [PubMed] [Google Scholar]
  • 46.Cameron AC, Trivedi PK. Regression Analysis of Count Data. Cambridge: Cambridge University Press; 1998. [Google Scholar]
  • 47.Stokes M, Davis C, Koch G. Categorical Data Analysis Using SAS. Cary, NC: SAS Institute; 2012. [Google Scholar]
  • 48.Cameron A, Trivedi P. Microeconometrics Using Stata. College Station, TX: Stata Press; 2009. [Google Scholar]
  • 49.Babyak M. Rescaling continuous predictors in regression models. Statistical Tips from the Editors of Psychosomatic Medicine: Journal of Biobehavioral Medicine. 2009 Sep 24; 2009 [July 22, 2013]; Available from: http://stattips.blogspot.com/2009/08/rescaling-continuous-predictors-in.html.
  • 50.Des Jarlais DC. Research, politics, and needle exchange. American Journal of Public Health 2000. 2000;90(9):1392–4. doi: 10.2105/ajph.90.9.1392. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Des Jarlais DC, McKnight C, Milliken J. Public funding of U.S. syringe exchange programs. Journal of Urban Health 2004. 2004;18:118–21. doi: 10.1093/jurban/jth093. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Akaike H. A new look at the statistical model identification. IEEE Transactions on Automatic Control. 1974 Dec;19(6):716–23. [Google Scholar]
  • 53.Beal D. SAS Code to Select the Best Multiple Linear Regression Model for Multivariate Data Using Information Criteria. 2005. Contract No.: Paper SA01_05. [Google Scholar]
  • 54.Burnham K, Anderson D. Multimodal Inference: Understanding AIC and BIC in Model Selection. Sociological Methods and Research. 2004;33:261–304. [Google Scholar]
  • 55.Kuha J. AIC and BIC: Comparisons of Assumptions and Performance. Sociological Methods and Research. 2004;33:188–229. [Google Scholar]
  • 56.Fulton County Department of Health. Ryan While Title 1 Grant Application: Atlanta EMA FY 2000. 2000;2000 [Google Scholar]
  • 57.Coates T, Richter L, Caceres C. Behavioural strategies to reduce HIV transmission: how to make them work better. Lancet. 2008;372(9639):669–84. doi: 10.1016/S0140-6736(08)60886-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Kurth A, Celum C, Baeten J, Vermund S, Wasserheit J. Combination HIV prevention: significance, challenges, and opportunties. Curr HIV/AIDS Rep. 2011;8(1):62–72. doi: 10.1007/s11904-010-0063-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Padian N, McCoy S, Karim S, et al. HIV prevention transformed: the new prevention research agenda. Lancet. 2011;378(9787):269–78. doi: 10.1016/S0140-6736(11)60877-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Vermund SH, Hayes R. Combination prevention: new hope for stopping the epidemic. Curr HIV/AIDS Rep. 2013;10(2):169–86. doi: 10.1007/s11904-013-0155-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Huo D, Ouellet L. Needle exchange and sexual risk behaviors among a cohort of injection drug users in Chicago, Illinois. Sex Transm Dis. 2009;36(1):35–40. doi: 10.1097/OLQ.0b013e318186dee3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.National Research Council. Preventing HIV Infection among Injecting Drug Users in High Risk Countries: An Assessment of the Evidence. Washington, DC: The National Academies Press; 2006. [Google Scholar]
  • 63.Sterne JA, Hernan MA, Ledergerber B, Tilling K, Weber R, Sendi P, et al. Long-term effectiveness of potent antiretroviral therapy in preventing AIDS and death: a prospective cohort study. Lancet. 2005;366:378–84. doi: 10.1016/S0140-6736(05)67022-5. [DOI] [PubMed] [Google Scholar]
  • 64.Mocroft A, Ledergerber B, Katlama C, Kirk O, Reiss P, d’Arminio Montforte A, et al. Decline in the AIDS and death rates in the EuroSIDA study: an observational study. Lancet. 2003;362(9377):22–9. doi: 10.1016/s0140-6736(03)13802-0. [DOI] [PubMed] [Google Scholar]
  • 65.Mathers BM, Degenhardt L, Phillips B, Wiessing L, Hickman M, Strathdee SA, et al. Global epidemiology of injecting drug use and HIV among people who inject drugs: a systematic review. Lancet. 2008 Nov 15;372(9651):1733–45. doi: 10.1016/S0140-6736(08)61311-2. [DOI] [PubMed] [Google Scholar]
  • 66.Hurley SF, Jolley DJ, Kaldor JM. Effectiveness of needle-exchange programs for prevention of HIV infection. The Lancet 1997. 1997;21(349):1797–800. doi: 10.1016/S0140-6736(96)11380-5. [DOI] [PubMed] [Google Scholar]
  • 67.Vickerman P, Hickman M, Rhodes T, Watts C. Model projections on the required coverage of syringe distribution to prevent HIV epidemics among injecting drug users. J Acquir Immune Defic Syndr. 2006;42:355–61. doi: 10.1097/01.qai.0000219788.73539.47. [DOI] [PubMed] [Google Scholar]
  • 68.Semple S, Zians J, Strathdee SA, Patterson T. Sexual marathons and methamphetamine use among HIV-positive men who have sex with men. Arch Sex Behav. 2009;38(4):583–90. doi: 10.1007/s10508-007-9292-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 69.Rawstorne P, Fogarty A, Crawford J, Prestage G, Grierson J, Grulich A, et al. Differences between HIV-positive gay men who ‘frequently’, ‘sometimes’ or ‘never’ engage in unprotected anal intercourse with serononconcordant casual partners: positive Health cohort, Australia. AIDS Care. 2007;19(4):514–22. doi: 10.1080/09540120701214961. [DOI] [PubMed] [Google Scholar]
  • 70.Massey D, Denton N. American Apartheid: Segregation and the Making of the Underclass. Cambridge, Massachusetts: Harvard University Press; 1993. [Google Scholar]
  • 71.Biello KB, Kershaw T, Nelson R, Hogben M, Ickovics J, Niccolai LM. Racial Residential Segregation and Rates of Gonorrhea in the United States, 2003–2007. American Journal of Public Health. 2012;102(7):1370–7. doi: 10.2105/AJPH.2011.300516. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 72.Hogben M, Leichliter J. Social determinants and sexually transmitted disease disparities. Sexually Transmitted Disease. 2008;35:S13–S8. doi: 10.1097/OLQ.0b013e31818d3cad. [DOI] [PubMed] [Google Scholar]
  • 73.Poundstone K, Strathdee SA, Celentano D. The Social Epidemiology of Human Immunodeficiency Virus/Acquired Immunodeficiency Syndrome. Epidemiologic Reviews. 2004;26(1):22–35. doi: 10.1093/epirev/mxh005. [DOI] [PubMed] [Google Scholar]
  • 74.Williams DR, Collins C. Racial residential segregation: a fundamental cause of racial disparities in health. Public Health Reports. 2001;116(5):404–16. doi: 10.1093/phr/116.5.404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 75.Adimora AA, Schoenbach VJ. Contextual factors and the black-white disparity in heterosexual HIV transmission. Epidemiology. 2002 Nov;13(6):707–12. doi: 10.1097/00001648-200211000-00016. [DOI] [PubMed] [Google Scholar]
  • 76.Adimora AA, Schoenbach VJ. Social context, social networks, and racial disparities in rates of sexually transmitted infections. Journal of Infectious Diseases. 2005;191(Suppl 1):S115–S22. doi: 10.1086/425280. [DOI] [PubMed] [Google Scholar]
  • 77.Adimora AA, Schoenbach VJ, Doherty IA. HIV and African Americans in the southern United States: sexual networks and social context. Sexually Transmitted Diseases. 2006 Jul;33(7 Suppl):S39–45. doi: 10.1097/01.olq.0000228298.07826.68. [DOI] [PubMed] [Google Scholar]
  • 78.Adimora AA, Schoenbach VJ, Floris-Moore MA. Ending the epidemic of heterosexual HIV transmission among African Americans. American Journal of Preventive Medicine. 2009;37(5):468–71. doi: 10.1016/j.amepre.2009.06.020. [Research Support, N.I.H Extramural] [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 79.Stuckler D, Basu S. The Body Economic: Why Austerity Kills. New York, NY: Basic Books; 2013. [Google Scholar]
  • 80.Berkman LF, Kawachi I, editors. Social Epidemiology. New York, NY: Oxford; 2000. [Google Scholar]
  • 81.Wilkinson R, Pickett K. The Spirit Level: Why Equality is Better for Everyone. London: Penguin Books Ltd; 2010. [Google Scholar]
  • 82.Wilkinson R. Income inequality, social cohesion, and health: clarifying the theory--a reply to Muntaner and Lynch. Int J Health Serv. 1999;29(3):525–43. doi: 10.2190/3QXP-4N6T-N0QG-ECXP. [DOI] [PubMed] [Google Scholar]
  • 83.Chida Y, Steptoe A, Powell LH. Religiosity/spirituality and mortality. A systematic quantitative review. Psychother Psychosom. 2009;72(2):81–90. doi: 10.1159/000190791. [DOI] [PubMed] [Google Scholar]
  • 84.Holt-Lunstad J, Smith T, Layton J. Social Relationships and Mortality Risk: A Meta-analytic Review. PLoS Med. 2010;7(7) doi: 10.1371/journal.pmed.1000316. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 85.Galea S, Tracy M, Hoggatt K, Dimaggio C, Karpati A. Estimated deaths attributable to social factors in the United States. American Journal of Public Health. 2011;101(8):1456–65. doi: 10.2105/AJPH.2010.300086. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 86.Galea S, Ahern J, Rudenstine S, Wallace A, Vlahov D. Urban built environment and depression: a multilevel analysis. J Epidemiol Community Health. 2005 Jan;59(10):822–7. doi: 10.1136/jech.2005.033084. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 87.Chaix B, Leal C, Evans D. Neighborhood-level confounding in epidemiologic studies: unavoidable challenges, uncertain solutions. Epidemiology. 2010 Jan;21(1):124–7. doi: 10.1097/EDE.0b013e3181c04e70. [DOI] [PubMed] [Google Scholar]
  • 88.Substance Abuse and Mental Health Services Administration. Treatment Episode Data Set (TEDS) Series. 1992–2002. [Computer file]. In press. [Google Scholar]
  • 89.Harper S. Estimates of GINI Coefficients based on US Census Bureau Data. 2005. [Google Scholar]
  • 90.US Census Bureau. Population Estimates. Washington, DC: US Census Bureau; 1990. [Google Scholar]
  • 91.US Census Bureau. Population Estimates. Washington, DC: US Census Bureau; 2000. [Google Scholar]
  • 92.Tempalski B, McQuie H. Drugscapes and the role of place and space in injection drug use-related HIV risk environments. Int J Drug Policy. 2009 Jan;20(1):4–13. doi: 10.1016/j.drugpo.2008.02.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 93.Association of Religious Data Archives. Churches and church membership in the United States, 1990 (Counties) 2009 Available from: http://www.thearda.com/Archive/Files/Descriptions/CMS90CNT.asp.
  • 94.Beth Israel Medical Center. New York: Beth Israel Medical Center; 2000. 2000. [Google Scholar]
  • 95.Federal Bureau of Investigations. Crime in the United States. 2009 [November 15, 2010]; Available from: http://www2.fbi.gov/ucr/cius2009/index.html.
  • 96.Centers for Disease Control and Prevention. HIV counseling and testing in publicly funded sites [Computer file] 2008. [Google Scholar]

RESOURCES