Determining the most likely source of infection: an application to Neisseria gonorrhoeae among men who have sex with men

The source of an infection is often unknown. To inform directed prevention measures, it is useful to know the location and partner type with the highest transmission risk. We developed a method to estimate infection risk of Neisseria gonorrhoeae per meeting location among men who have sex with men (MSM). In 2008-2009, we collected information from 2438 MSM attending the Sexually Transmitted Infections clinic of Amsterdam. For up to four partners per participant (8028 in total), we asked for details on meeting location, partner, and partnership characteristics. We used logistic regression to relate these to the participant's infection risk, accounting for unobserved transmission information in the likelihood. Based on the model estimates, we predicted the probability of a partner having N. gonorrhoeae. The probability that a partner was the source was proportional to his predicted infection risk. Each source was linked to the meeting location. We used a Bayesian method. Rectal N. gonorrhoeae was diagnosed in 157 MSM who reported data on 422 possible source partners, urethral N. gonorrhoeae in 126 reporting 285 possible sources, and pharyngeal N. gonorrhoeae in 162 reporting 451 possible sources. We estimated that most infections were acquired from long-lasting steady partners (21%; 95% CI 17-24). Partners met in an Amsterdam street with gay venues posed the highest transmission risk (13%; 95% CI 7.9-18). The presented method estimates the source of infection when there are multiple possible sources and enables the summation over various kinds of epidemiologic characteristics (here, meeting locations) that are relevant for prevention