In a recent paper Dominici, working with other members of her group in the Department of Biostatistics at Harvard used the In policy research, it's crucial to understand how changes in a policy affect outcomes. This process, called shift-response function (SRF) estimation, can be complex. Existing neural network methods for this task need to be better tested theoretically and practically. To address this, Dominici and co-authors developed a new neural network method with proven reliability and efficiency for SRF estimation. The team applied this method to a large dataset (68 million people and 27 million deaths) to estimate the impact of a proposed change in U.S. air quality standards on mortality rates. They demonstrated that their new method, TRESNET, improves existing techniques by ensuring reliable and efficient results and handling various types of outcome data. The study also tested TRESNET in different scenarios to show its effectiveness and versatility.