Correlates of Annual Testing for Sexually Transmitted Infections (STIs) in an Online Sample of Men Who Have Sex with Men (MSM): Study Sample Validity, Measure Reliability, and Behavioral Typologies

Abstract

Objective: Testing for STIs has been prioritized as part of a comprehensive HIV/AIDS prevention plan. Internet-based studies of STI testing among men who have sex with men (MSM) are efficient methods of recruiting non-clinic samples from diverse geographic areas. However, online survey methods raise unique concerns regarding threats to the validity of study samples and unknown measurement properties. Thus, this dissertation had two aims. The first was to examine methods related to online survey research by evaluating a protocol to detect invalid survey entries and determining the test-retest reliability of online measures of sexual behavior and STI testing. The second aim was to use the validated sample and reliable measures to examine correlates of STI testing in the year prior to the survey. Methods: In Manuscript 1, survey submissions were classified as valid and invalid according to a de-duplication and cross-validation protocol. Logistic regression models were used to determine associations between invalidity and key demographic and behavioral variables. In Manuscript 2, test-retest reliability over one week was evaluated using intraclass correlation coefficients (ICCs) and kappa statistics for measures of sexual behavior, HIV status, HIV testing, and STI diagnoses. Finally, in Manuscript 3, the valid sample from Manuscript 1 and measures that were evaluated in Manuscript 2 were used to examine the clustering and correlates of STI testing behaviors. Results: In Manuscript 1, three components of the protocol for detecting invalid submissions were responsible for identifying the most invalid survey submissions: duplicate IP address, changed eligibility responses, and duplicate payment name. A total of 146 (11.6%) of the submissions were identified as invalid. Invalid submissions had lower odds of reporting HIV testing in the past year. Hispanic/Latino identity, age, and HIV status were also significantly associated with invalidity. In Manuscript 2, counts of sexual partners (three months), HIV status, HIV testing, and STI diagnoses were found to have substantial (0.61-0.80) to almost perfect (0.81-1.00) seven-day test-retest reliabilities, according to commonly used cutpoints. Partner-specific data, however, were only fairly or moderately reliable (0.21-0.60). Finally, in Manuscript 3, a latent class analysis indicated five STI testing classes: no STIs, all STIs, bacterial STIs and hepatitis, bacterial STIs only, and hepatitis only. The largest class was no STIs, indicating that 45.8% of the validated sample had not been tested for STIs in the past year. Predictors of being in a testing class versus no STI testing included age, education, outness about having sex with men, HIV status, and having a sexual partner in the last three months. Conclusions: This dissertation served two primary aims. The first was to evaluate sample validity and measure reliability in an online study of MSM. The second was to apply the information from those analyses to examine the presence and correlates of a latent variable of STI testing. Across all three manuscripts, online survey research appears to be a viable method of studying STI testing in Internet-based samples of MSM.