(P145) CAN PARTNER NOTIFICATION REDUCE POPULATION-LEVEL GONORRHEA INCIDENCE? MODEL-BASED ANALYSIS OF KEY DETERMINANTS OF INTERVENTION IMPACT

Background:
Limited estimates exist on population-level impact of partner notification (PN) for gonorrhea with uncertainty in the influence of local epidemiology on PN effectiveness. An ecological study in New York found a 6% reduction in diagnoses with a 10% increase in PN coverage. We modelled reductions in gonorrhea incidence through use of PN under varying epidemiological conditions to compare effect sizes to this prior estimate and understand key determinants of variation.

Methods:
We developed a stochastic network model of men who have sex with men and calibrated gonorrhea transmission dynamics to varied epidemiological conditions. Calibration targets were prevalence, incidence, diagnosis rate, and proportion diagnosed with symptoms (2-10% prevalence and 2-23% incidence). We captured parametric and stochastic uncertainty. Population-level impact of increasing PN was summarized by incidence rate ratios (IRRs), and relative importance of explanatory variables (including network density, baseline burden, natural history parameters) was assessed via linear regression modeling of IRR, considering calibrated parameters as independent variables, and bootstrapping to evaluate uncertainty in estimation.

Results:
Modeled effect sizes were compatible overall with the previous ecological study. We estimated IRR of 0.97 (95% range 0.93­-0.99) for a 10% relative increase in PN coverage, comparable to the IRR of 0.94 (0.91-­0.97) identified in the prior study. PN retained effectiveness under diverse epidemiological conditions. The strongest correlates of variation in the estimated IRR were parameters governing index case testing probability, with higher testing probability associated with higher impact. In univariate sensitivity analysis maximum testing frequency for symptomatic or asymptomatic patients yielded estimated IRRs of 0.92 (0.90-0.94) or 0.93 (0.93-0.94), respectively. Network density, baseline incidence and other natural history parameters had relatively modest effects on IRR.

Conclusion:
Improved understanding of the contextual factors that influence the effectiveness of PN can inform more effective programming, and be useful in benchmarking impact across varied settings.