This proposal seeks to better understand the underlying microeconomic drivers of physician use of information technology (IT), with a focus on the provision of preventive care and its downstream population health impacts. To do so, we will leverage unique, rich data on physician use of IT (at the second-by-second level) to assess patient eligibility for preventive care, linked to traditional data sources (e.g. claims), and quasi-random variation
We will incorporate methods from machine learning and econometrics to i) tractably characterize use of IT based on numerous potential patterns of use and ii) understand heterogeneity in impacts of IT using datadriven approaches to estimation. We will be able to study the average impact of IT on the provision of preventive care. In addition, we will use natural variation in which incentives were moved from measures of the process of preventive care – screening for HbA1c level – to intermediate health outcomes – HbA1c levels under control. This will allow us to assess the impact of IT on physician and patient effort on preventive care as well as the impact on health itself. Our approach allows us to consider not only the mean impact of IT, but also the potential sources of its heterogeneous impacts on access and adherence to preventive care. For example,
among diabetic patients, is the primary determinant of appropriate preventive screening whether the patient visits the doctor or, conditional on visiting the doctor, whether the physician provides appropriate screening? How do these margins for care depend on access to IT, and physician, patient, and treatment heterogeneity? These distinctions have important implications for designing interventions and incentives to improve health
among the diabetic population – and prevention more broadly. IT also has the potential to provide targeted individual-level information and treatment guidelines.
Enhancing the scope of IT in this way could improve the effectiveness of targeted medical treatments in practice. However, how individuals and their doctors actually adopt IT and use it to respond to even simple guidelines is critical in assessing how to best disseminate patient-centered information and personalized treatment recommendations, with the primary goal of improving population health. We will identify the variance of health outcomes induced by a range of different medical treatments, in addition to the mean treatment impacts typically studied. We will use our results to assess the potential population health gains from IT-driven personalized treatment recommendations, informing policies designed to best use resources to improve the health care management and health of an aging population.