Substance Use, HIV Serostatus Disclosure, and Sexual Risk Behavior in People Living with HIV: An Event-Level Analysis

Abstract

People living with HIV (PLWH) face difficult decisions about disclosing their HIV status to new sexual partners. Alcohol and other drug use could impact these decision-making processes and subsequent sexual risk behavior. We sought to examine the event-level relationships between substance use, HIV disclosure, and condom use in PLWH and their first-time HIV-negative or unknown status sexual partners. Adult PLWH were recruited from care settings in a southeastern U.S. city. Participants reported their sexual behavior for 28 consecutive days via text message prompts. We employed multilevel covariation in a causal system to examine the event-level relations between substance use and condom use. We proposed that this relationship would be mediated by HIV disclosure and moderated by viral suppression status. A total of 243 participants (83% male, 93% Black) reported 509 sexual events with first-time HIV-negative/unknown status sexual partners. Substance use at the time of sex was negatively associated with disclosure in PLWH with suppressed viral load (OR 0.29, β = − 1.22, 95% CI [− 2.42, − 0.03], p = .045), but differentially associated with condom use in PLWH with detectable versus undetectable viral load. In PLWH with viral suppression, participants who always disclosed versus who never disclosed their HIV status were more likely to use condoms (β = 1.84, 95% CI [0.35, 3.53], p = .017), but inconsistent disclosers were less likely to use a condom after disclosing (OR 0.22, 95% CI [0.07, 0.68], p = .008). Event-level analysis offers a more nuanced understanding of the proximal (substance use, HIV disclosure) and person-level (substance use, viral load) determinants of HIV transmission risk behavior in PLWH.

Appendix: MPlus Code for Event-Level Models

Within-Person Effects Mediation Model (Results Reported in Table 3a, b)

INPUT INSTRUCTIONS
TITLE: Substance use, disclosure, and condom use mediation – Within-person
DATA: FILE IS Data.csv;
VARIABLE: NAMES ARE id condom_use substance_use disclosure partner_gender participant_gender age;
USEVAR ARE id condom_use substance_use disclosure_within partner_gender audit disclosure_between;
CATEGORICAL ARE disclosure_within condom_use;
CLUSTER = id;
WITHIN = substance_use partner_gender;
BETWEEN = audit disclosure_between;
Missing are all (-99);
ANALYSIS: TYPE = twolevel random;
ESTIMATOR = mlr;
MODEL: %WITHIN%
disclosure_within ON substance_use (awithin);
condom_use ON disclosure_within (bwithin);
condom_use ON substance_use;
condom_use ON partner_gender;
disclosure ON partner_gender;
%BETWEEN%
disclosure_within condom_use audit disclosure_between;
condom_use ON audit;
condom_use ON disclosure_between;
MODEL CONSTRAINT:
NEW (indw);
indw = awithin*bwithin;
OUTPUT: sampstat cinterval;

Between-Person Mediation Models (Results Reported in Table 4)

Four model versions alternately utilized between-person “substanceuse” terms derived from (a) AUDIT score, (b) DAST score, (c) number of positive drug screens at study entry, and (d) proportion of events using substances

INPUT INSTRUCTIONS
TITLE: Substance use, disclosure, and condom use mediation – Between-person
DATA: FILE IS Data.csv;
VARIABLE: NAMES ARE id condom_use substance_use disclosure_between partner_gender participant_gender age;
USEVAR ARE id condom_use substance_use disclosure_between partner_gender participant_gender age;
CATEGORICAL ARE condom_use;
CLUSTER = id;
WITHIN = ;
BETWEEN = substance_use disclosure_between participant_gender age;
Missing are all (-99);
ANALYSIS: TYPE = twolevel random;
ESTIMATOR = mlr;
MODEL: %WITHIN%
%BETWEEN%
substance_use disclosure_between condom_use gender age;
disclosure_between ON substance_use (abetween);
condom_use ON disclosure_between (bbetween);
condom_use ON substance_use;
condom_use ON participant_gender;
condom_use ON age;
disclosure_between ON participant_gender;
disclosure_between ON age;
MODEL CONSTRAINT:
NEW (indb);
indv = abetween*bbetween;
OUTPUT: sampstat cinterval;