Open AccessRecent Substance Use and Probability of Unsuppressed HIV Viral Load Among Persons on Antiretroviral Therapy in Continuity Care
Author(s) -
Catherine R. Lesko,
Alexander P. Keil,
Anthony T. Fojo,
Geetanjali Chander,
Bryan Lau,
Richard D. Moore
Publication year - 2019
Publication title -
american journal of epidemiology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.33
H-Index - 256
eISSN - 1476-6256
pISSN - 0002-9262
DOI - 10.1093/aje/kwz159
Subject(s) - medicine , viral load , heroin , credible interval , substance abuse , confidence interval , psychiatry , drug , environmental health , human immunodeficiency virus (hiv) , immunology
Among persons with human immunodeficiency virus (HIV) infection, illegal drug use and hazardous alcohol use are hypothesized to be strong risk factors for failure to achieve or maintain a suppressed HIV viral load, but accurate quantification of this association is difficult because of challenges involved in measuring substance use as part of routine clinical care. We estimated the associations of recent cocaine use, opioid/heroin use, and hazardous alcohol use with unsuppressed viral load among 1,554 persons receiving care at the John G. Bartlett Specialty Practice (Baltimore, Maryland) between 2013 and 2017. We accounted for measurement error in substance use using Bayesian models and prior estimates of the sensitivity and specificity of 2 imperfect measures of substance use derived from a previous analysis in this cohort. The prevalence difference for unsuppressed viral load associated with recent cocaine use was 11.3% (95% credible interval (CrI): 6.4, 17.0); that associated with recent opioid/heroin use was 13.2% (95% CrI: 6.6, 20.7); and that associated with recent hazardous alcohol use was 8.5% (95% CrI: 3.2, 14.4). Failure to account for measurement error resulted in clinically meaningful underestimates of the prevalence difference. Time-varying substance use is prevalent and difficult to measure in routine care; here we demonstrate a method that improves the utility of imperfect data by accounting for measurement error.