Endogenous Screening and Health Behaviors

Our paper models the decision to seek low-dose CT (LDCT) lung cancer screening. Because 80$\%$ of lung cancers are caused by cigarette smoking, we build a dynamic structural model in which screening and smoking are chosen simultaneously. Screening has the potential to identify early-stage lung cancer, when it is potentially treatable, and it alleviates uncertainty about the true lung cancer state. At the same time, screening is costly in both pecuniary and non-pecuniary terms, and it is not perfect in the sense that false positives may lead to unnecessary care. Smoking generates utility, particularly for those with significant and recent smoking histories, but it also increases the risk for both lung cancer and other chronic conditions. Our goal is to estimate the model and to simulate policy counterfactual scenarios that address both costs and benefits of screening.

To directly estimate the model, we would need longitudinal information on screening and smoking behaviors, subjective beliefs, screening outcomes, and health transitions over many years. Because such data do not exist, our empirical approach combines data from several sources. For our primary data source, we plan to conduct a randomized information experiment that will evaluate how brief, randomized information about stigma, perceived quality, and price affect intentions to seek low-dose CT (LDCT) lung cancer screening within 12 months. Our population of interest are those above the age of 40 who have some cigarette smoking history. Before the information treatments, we will assess eligibility for $0 out-of-pocket (OOP) screening based on USPSTF 2022 guidelines. We intend to report the average treatment effects generated by our experiment and the heterogeneity in these effects by theoretically relevant factors at baseline, including stigma, perceptions of quality, and income. We will then use these treatment effects to calibrate our model. We will run our survey on Prolific.