Abstract
New combination human acquired deficiency (HIV) prevention strategies that include biomedical and primary prevention approaches add complexity to the task of measuring sexual risk. Latent transition models are beneficial for understanding complex phenomena; therefore, we trialed the application of latent class and latent transition models to HIV surveillance data. Our aims were to identify sexual risk states and model individuals' transitions between states. A total of 4,685 HIV-negative men who have sex with men (MSM) completed behavioral questionnaires alongside tests for HIV and sexually transmissible infections at one of 2 Melbourne, Victoria, Australia, general practices (2007-2013). We found 4 distinct classes of sexual risk, which we labeled "monogamous" (n = 1,224), "risk minimizer" (n = 1,443), "risk potential" (n = 1,335), and "risk taker" (n = 683). A positive syphilis, gonorrhea, or chlamydia test was significantly associated with class membership. Among a subset of 516 MSM who had at least 3 clinic visits, there was general stability across risk classes; MSM had on average a 0.70 (i.e., 70%) probability of remaining in the same class between visits 1 and 2 and between visits 2 and 3. Monogamous MSM were one exception; the probability of remaining in the monogamous class was 0.51 between visits 1 and 2. Latent transition analyses identified unobserved risk patterns in surveillance data, characterized high-risk MSM, and quantified transitions over time.