PROJECT SUMMARY Medical and behavioral interventions are critical for preventing the spread of infectious diseases. But there are statistical challenges when estimating causal effects for such interventions. Most statistical methods assume the treatment of one individual does not affect outcomes of other individuals, i.e., that there is no interference or spillover effect. However, the effects of infectious disease interventions often include both direct effects on individuals receiving treatment and spillover effects in the community. An additional challenge for estimation of effects in infectious disease settings is external validity, as study samples often differ from the populations where interventions will be applied. Such differences can limit generalizability of study findings. Despite the importance in the field of infectious disease research, there has been no research at the intersection of interference and generalizability. The candidate's long-term career goal is to become an independent scientist and expert biostatistician working at the interface of causal inference, epidemiology, and infectious diseases. Therefore, the primary goal of this K01 award is to acquire training in infectious disease epidemiology, interference, and machine learning to develop and apply innovative statistical methods for infectious disease research. The candidate, Dr. Bonnie Shook-Sa, has assembled a strong team of mentors and collaborators who are leaders in each of these training areas. Training will support Dr. Shook-Sa's research aims which bring together the fields of interference and generalizability for the development of estimators of overall and spillover effects from infectious disease prevention studies that generalize to target populations that differ from study samples. Methods will first be developed in a standard interference setting, where individuals eligible for treatment are at risk of the outcome, e.g., for estimating the effect of child bed net use on incidence of malaria among children. Then, methods will be extended to the “bipartite interference” setting, where individuals eligible for treatment are distinct from those at risk of the outcome, e.g., for estimating the effect of universal HIV testing and treatment of persons with HIV on forward infection among persons without HIV. The proposed work will consider a variety of study designs, including cluster-randomized trials, probability-based sample designs, and convenience sample designs. Nonparametric estimators based on flexible machine learning methods will be proposed in each setting. The properties of proposed estimators will be derived, and estimators will be evaluated empirically through simulation studies. Proposed methods are motivated by and will be applied to data from the Uganda Malaria Indicator Survey, the HPTN 071: Population Effects of Antiretroviral Therapy to Reduce HIV Transmission (PopART) cluster-randomized trial, and the Hutterite Influenza Prevention Study.