Faculty Bibliography

PURPOSE/OBJECTIVE:After years of stable or declining HIV prevalence and declining incidence among people who inject drugs (PWID) in the United States, some rapidly emerging outbreaks have recently occurred in new areas (e.g., Scott County, Indiana). However, to our knowledge, trends over time in HIV prevalence among PWID in US metropolitan statistical areas (MSAs) across all major regions of the country have not been systematically estimated beyond 2002, and the extent to which HIV prevalence may be increasing in other areas is largely unknown. This article estimates HIV prevalence among PWID in 89 of the most populated US MSAs, both overall and by geographic region, using more recent surveillance and HIV testing data. METHODS:We computed MSA-specific annual estimates of HIV prevalence (both diagnosed and undiagnosed infections) among PWID for these 89 MSAs, for 1992-2013, using several data series from the Centers for Disease Control and Prevention's (CDC) National HIV Surveillance System and National HIV Prevention Monitoring and Evaluation data; Holmberg's (1997) estimates of 1992 PWID population size and of HIV prevalence and incidence among PWID; and research estimates from published literature using 1992-2013 data. A mixed effects model, with time nested within MSAs, was used to regress the literature review estimates on all of the other data series. Multiple imputation was used to address missing data. Resulting estimates were validated using previous 1992-2002 estimates of HIV prevalence and data on antiretroviral (ARV) prescription volumes and examined for patterns based on geographic region, numbers of people tested for HIV, and baseline HIV prevalence. RESULTS:Mean (across all MSAs) trends over time suggested decreases through 2002 (from approximately 11.4% in 1992 to 9.2% in 2002), followed by a period of stability, and steep increases after 2010 (to 10.6% in 2013). Validation analyses found a moderate positive correlation between our estimates and ARV prescription volumes (r = 0.45), and a very strong positive correlation (r = 0.94) between our estimates and previous estimates by Tempalski et al. (2009) for 1992-2002 (which used different methods). Analysis by region and baseline prevalence suggested that mean increases in later years were largely driven by MSAs in the Western United States and by MSAs in the Midwest that had low baseline prevalence. Our estimates suggest that prevalence decreased across all years in the Eastern United States. These trends were particularly clear when MSAs with very low numbers of people tested for HIV were removed from analyses to reduce unexplained variability in mean trajectories. CONCLUSIONS:Our estimates suggest a fairly large degree of variation in 1992-2013 trajectories of PWID HIV prevalence among 89 US MSAs, particularly by geographic region. They suggest that public health responses in many MSAs (particularly those with larger HIV prevalence among PWID in the early 1990s) were sufficient to decrease or maintain HIV prevalence over time. However, future research should investigate potential factors driving the estimated increase in prevalence after 2002 MSAs in the West and Midwest. These findings have potentially important implications for program and/or policy decisions, but estimates for MSAs with low HIV testing denominators should be interpreted with caution and verified locally before planning action.

Assessment of the long-term population-level effects of HIV interventions is an ongoing public health challenge. Following the implementation of a Transmission Reduction Intervention Project (TRIP) in Odessa, Ukraine, in 2013-2016, we obtained HIV pol gene sequences and used phylogenetics to identify HIV transmission clusters. We further applied the birth-death skyline model to the sequences from Odessa (n = 275) and Kyiv (n = 92) in order to estimate changes in the epidemic's effective reproductive number (R_e) and rate of becoming uninfectious (δ). We identified 12 transmission clusters in Odessa; phylogenetic clustering was correlated with younger age and higher average viral load at the time of sampling. Estimated R_e were similar in Odessa and Kyiv before the initiation of TRIP; R_e started to decline in 2013 and is now below R_e = 1 in Odessa (R_e = 0.4, 95%HPD 0.06-0.75), but not in Kyiv (R_e = 2.3, 95%HPD 0.2-5.4). Similarly, estimates of δ increased in Odessa after the initiation of TRIP. Given that both cities shared the same HIV prevention programs in 2013-2019, apart from TRIP, the observed changes in transmission parameters are likely attributable to the TRIP intervention. We propose that molecular epidemiology analysis can be used as a post-intervention effectiveness assessment tool.

Importance/UNASSIGNED:Treatment with methadone or buprenorphine is the current standard of care for opioid use disorder. Given the paucity of research identifying which patients will respond best to which medication, both medications should be accessible to all patients so that patients can determine which works best for them. However, given differences in the historical contexts of their initial implementation, access to each of these medications may vary along racial/ethnic lines. Objective/UNASSIGNED:To examine the extent to which capacity to provide methadone and buprenorphine vary with measures of racial/ethnic segregation. Design, Setting, and Participants/UNASSIGNED:This cross-sectional study included all counties and county-equivalent divisions in the US in 2016. Data on racial/ethnic population distribution were derived from the American Community Survey, and data on locations of facilities providing methadone and buprenorphine were obtained from Substance Abuse and Mental Health Services Administration databases. Data were analyzed from August 22, 2018, to September 11, 2019. Exposures/UNASSIGNED:Two county-level measures of racial/ethnic segregation, including dissimilarity (representing the proportion of African American or Hispanic/Latino residents who would need to move census tracts to achieve a uniform spatial distribution of the population by race/ethnicity) and interaction (representing the probability that an African American or Hispanic/Latino resident will interact with a white resident and vice versa, assuming random mixing across census tracts). Main Outcomes and Measures/UNASSIGNED:County-level capacity to provide methadone or buprenorphine, defined as the number of facilities providing a medication per 100 000 population. Results/UNASSIGNED:Among 3142 US counties, there were 1698 facilities providing methadone (0.6 facilities per 100 000 population) and 18 868 facilities providing buprenorphine (5.9 facilities per 100 000 population). Each 1% decrease in probability of interaction of an African American resident with a white resident was associated with 0.6 more facilities providing methadone per 100 000 population. Similarly, each 1% decrease in probability of interaction of a Hispanic/Latino resident with a white resident was associated with 0.3 more facilities providing methadone per 100 000 population. Each 1% decrease in the probability of interaction of a white resident with an African American resident was associated with 8.17 more facilities providing buprenorphine per 100 000 population. Similarly, each 1% decrease in the probability of interaction of a white resident with a Hispanic/Latino resident was associated with 1.61 more facilities providing buprenorphine per 100 000 population. Conclusions and Relevance/UNASSIGNED:These findings suggest that the racial/ethnic composition of a community was associated with which medications residents would likely be able to access when seeking treatment for opioid use disorder. Reforms to existing regulations governing the provisions of these medications are needed to ensure that both medications are equally accessible to all.

SUMMARY:We investigate convolutional neural networks (CNNs) for filtering small genomic variants in short-read DNA sequence data. Errors created during sequencing and library preparation make variant calling a difficult task. Encoding the reference genome and aligned reads covering sites of genetic variation as numeric tensors allows us to leverage CNNs for variant filtration. Convolutions over these tensors learn to detect motifs useful for classifying variants. Variant filtering models are trained to classify variants as artifacts or real variation. Visualizing the learned weights of the CNN confirmed it detects familiar DNA motifs known to correlate with real variation, like homopolymers and short tandem repeats (STR). After confirmation of the biological plausibility of the learned features we compared our model to current state-of-the-art filtration methods like Gaussian Mixture Models, Random Forests and CNNs designed for image classification, like DeepVariant. We demonstrate improvements in both sensitivity and precision. The tensor encoding was carefully tailored for processing genomic data, respecting the qualitative differences in structure between DNA and natural images. Ablation tests quantitatively measured the benefits of our tensor encoding strategy. Bayesian hyper-parameter optimization confirmed our notion that architectures designed with DNA data in mind outperform off-the-shelf image classification models. Our cross-generalization analysis identified idiosyncrasies in truth resources pointing to the need for new methods to construct genomic truth data. Our results show that models trained on heterogenous data types and diverse truth resources generalize well to new datasets, negating the need to train separate models for each data type. AVAILABILITY AND IMPLEMENTATION:This work is available in the Genome Analysis Toolkit (GATK) with the tool name CNNScoreVariants (https://github.com/broadinstitute/gatk). SUPPLEMENTARY INFORMATION:Supplementary data are available at Bioinformatics online.

OBJECTIVE:To examine the impact of Section 1115 waivers on Medicaid coverage and opioid agonist therapy (OAT) utilization among substance use treatment admissions. DATA SOURCE/METHODS:Treatment Episode Data Set-Admissions (TEDS-A) (2001-2012). STUDY DESIGN/METHODS:We examined effects of 1115 waiver implementation on proportions of substance use treatment admissions with Medicaid and receiving OAT, using random intercept linear regression. PRINCIPAL FINDINGS/RESULTS:1115 waiver implementation was associated with an average of a 6 percentage point increase in proportion of all admissions with Medicaid, and 4 percentage point increase among opioid outpatient admissions. Implementation was not associated with change in proportion of opioid outpatient admissions receiving OAT. CONCLUSIONS:1115 waivers influence Medicaid coverage among substance use treatment admissions. The findings improve our understanding of how state policies impact substance use treatment utilization.

AIM/OBJECTIVE:Evidence is emerging that prescription opioid (PO) injection is associated with increased health risks. This mixed-methods study compares the mechanics of PO and heroin injection and examines the demographic and drug-related correlates of lifetime PO injection in a sample of young people who inject drugs (PWID) in New York City (NYC). METHODS:Qualitative analysis of 46 semi-structured interviews with young adult opioid users ages 18-32. Interview segments describing PO injection were analyzed for common themes. Quantitative analysis of structured interviews with 539 young adult opioid users ages 18-29 recruited via respondent-driven sampling (RDS). Analyses are based on the subsample of 353 participants (65%) who reported having ever injected drugs. All variables were assessed via self-report, except hepatitis C virus status, which was established via rapid antibody testing. RESULTS:Participants described injecting POs and reported that preparing abuse-deterrent pills for injection is especially cumbersome, requiring extended manipulation and large amounts of water. Injecting POs, in contrast to injecting heroin, requires repeated injections per injection episode. Among RDS-recruited participants, the majority of injectors reported injecting POs, sporadically (33%) or regularly (26%), but often infrequently (≤ 7 days/month). In separate multivariable analyses controlling for syringe- and cooker-sharing, ever injecting POs was a significant predictor of testing HCV antibody-positive (AOR = 2.97) and lifetime experience of non-fatal overdose (AOR = 2.51). Ever injecting POs was independently associated with lifetime homelessness (AOR = 2.93) and having grown up in a middle-income ($51,000-100,000/year vs. ≤ $50,000/year; AOR = 1.86) or a high-income household (> $100,000/year vs. ≤ $50,000/year; AOR = 2.54). CONCLUSIONS:Even in an urban environment like NYC with widespread heroin access, most young PWID have injected POs, although less frequently than heroin. PO injection involves practices that are known to increase risk for blood-borne viral infection (e.g., repeated injections) and predicted testing HCV-positive, as well as overdose. PO injection may also serve as a marker for a subgroup of PWID at elevated risk for multiple drug use-related comorbidities. Programs that provide prevention services to PWID need to tailor harm reduction measures and messaging to the specific practices and harms associated with the injection of POs.

BACKGROUND:Syringe services programs (SSPs) are effective venues for delivering harm-reduction services to people who inject drugs (PWID). However, SSPs often face significant barriers to implementation, particularly in the absence of known human immunodeficiency virus (HIV) outbreaks. METHODS:Using an agent-based model, we simulated HIV transmission in Scott County, Indiana, a rural county with a 1.7% prevalence of injection drug use. We compared outcomes arising in the absence of an SSP, in the presence of a pre-existing SSP, and with implementation of an SSP after the detection of an HIV outbreak among PWID over 5 years following the introduction of a single infection into the network. RESULTS:In the absence of an SSP, the model predicted an average of 176 infections among PWID over 5 years or an incidence rate of 12.1/100 person-years. Proactive implementation averted 154 infections and decreased incidence by 90.3%. With reactive implementation beginning operations 10 months after the first infection, an SSP would prevent 107 infections and decrease incidence by 60.8%. Reductions in incidence were also observed among people who did not inject drugs. CONCLUSIONS:Based on model predictions, proactive implementation of an SSP in Scott County had the potential to avert more HIV infections than reactive implementation after the detection of an outbreak. The predicted impact of reactive SSP implementation was highly dependent on timely implementation after detecting the earliest infections. Consequently, there is a need for expanded proactive SSP implementation in the context of enhanced monitoring of outbreak vulnerability in Scott County and similar rural contexts.

Importance/UNASSIGNED:Little is known about changes in marijuana use and cannabis use disorder (CUD) after recreational marijuana legalization (RML). Objectives/UNASSIGNED:To examine the associations between RML enactment and changes in marijuana use, frequent use, and CUD in the United States from 2008 to 2016. Design, Setting, and Participants/UNASSIGNED:This survey study used repeated cross-sectional survey data from the National Survey on Drug Use and Health (2008-2016) conducted in the United States among participants in the age groups of 12 to 17, 18 to 25, and 26 years or older. Interventions/UNASSIGNED:Multilevel logistic regression models were fit to obtain estimates of before-vs-after changes in marijuana use among respondents in states enacting RML compared to changes in other states. Main Outcomes and Measures/UNASSIGNED:Self-reported past-month marijuana use, past-month frequent marijuana use, past-month frequent use among past-month users, past-year CUD, and past-year CUD among past-year users. Results/UNASSIGNED:The study included 505 796 respondents consisting of 51.51% females and 77.24% participants 26 years or older. Among the total, 65.43% were white, 11.90% black, 15.36% Hispanic, and 7.31% of other race/ethnicity. Among respondents aged 12 to 17 years, past-year CUD increased from 2.18% to 2.72% after RML enactment, a 25% higher increase than that for the same age group in states that did not enact RML (odds ratio [OR], 1.25; 95% CI, 1.01-1.55). Among past-year marijuana users in this age group, CUD increased from 22.80% to 27.20% (OR, 1.27; 95% CI, 1.01-1.59). Unmeasured confounders would need to be more prevalent in RML states and increase the risk of cannabis use by 1.08 to 1.11 times to explain observed results, indicating results that are sensitive to omitted variables. No associations were found among the respondents aged 18 to 25 years. Among respondents 26 years or older, past-month marijuana use after RML enactment increased from 5.65% to 7.10% (OR, 1.28; 95% CI, 1.16-1.40), past-month frequent use from 2.13% to 2.62% (OR, 1.24; 95% CI, 1.08-1.41), and past-year CUD from 0.90% to 1.23% (OR, 1.36; 95% CI, 1.08-1.71); these results were more robust to unmeasured confounding. Among marijuana users in this age group, past-month frequent marijuana use and past-year CUD did not increase after RML enactment. Conclusions and Relevance/UNASSIGNED:This study's findings suggest that although marijuana legalization advanced social justice goals, the small post-RML increase in risk for CUD among respondents aged 12 to 17 years and increased frequent use and CUD among adults 26 years or older in this study are a potential public health concern. To undertake prevention efforts, further studies are warranted to assess how these increases occur and to identify subpopulations that may be especially vulnerable.

Racial/ethnic homophily in sexual partnerships (partners share the same race/ethnicity) has been associated with racial/ethnic disparities in HIV. Structural racism may partly determine racial/ethnic homophily in sexual partnerships. This study estimated associations of racial/ethnic concentration and mortgage discrimination against Black and Latino residents with racial/ethnic homophily in sexual partnerships among 7847 people who inject drugs (PWID) recruited from 19 US cities to participate in CDC's National HIV Behavioral Surveillance. Racial/ethnic concentration was defined by two measures that respectively compared ZIP code-level concentrations of Black residents to White residents and Latino residents to White residents, using the Index of Concentration at the Extremes. Mortgage discrimination was defined by two measures that respectively compared county-level mortgage loan denial among Black applicants to White applicants and mortgage loan denial among Latino applicants to White applicants, with similar characteristics (e.g., income, loan amount). Multilevel logistic regression models were used to estimate associations. Interactions of race/ethnicity with measures of racial/ethnic concentration and mortgage discrimination were added to the final multivariable model and decomposed into race/ethnicity-specific estimates. In the final multivariable model, among Black PWID, living in ZIP codes with higher concentrations of Black vs. White residents and counties with higher mortgage discrimination against Black residents was associated with higher odds of homophily. Living in counties with higher mortgage discrimination against Latino residents was associated with lower odds of homophily among Black PWID. Among Latino PWID, living in ZIP codes with higher concentrations of Latino vs. White residents and counties with higher mortgage discrimination against Latino residents was associated with higher odds of homophily. Living in counties with higher mortgage discrimination against Black residents was associated with lower odds of homophily among Latino PWID. Among White PWID, living in ZIP codes with higher concentrations of Black or Latino residents vs. White residents was associated with lower odds of homophily, but living in counties with higher mortgage discrimination against Black residents was associated with higher odds of homophily. Racial/ethnic segregation may partly drive same race/ethnicity sexual partnering among PWID. Future empirical evidence linking these associations directly or indirectly (via place-level mediators) to HIV/STI transmission will determine how eliminating discriminatory housing policies impact HIV/STI transmission.