Medical Testing as Optimal Information Acquisition

Medical testing is necessary for informed treatment decisions. However, its costs make it suboptimal to test too often. This paper investigates the role of costly information acquisition through medical testing in a patient's health status. Using universal health screening data linked with matched provider-patient medical claims data, I find providers not keeping diabetic patients' blood sugar levels in control tend to under-test. Guided by the reduced-form findings, I develop and estimate a dynamic structural model of medical testing and prescription adjustments. The novel feature of the model is that prescription choices not accompanied by a costly blood sugar test can be inaccurate. I find that the higher cost of blood sugar testing leads to blood sugar levels that are more dispersed over time through ill-informed prescription adjustments. Counterfactual exercises show that performing an A1C test, the state-of-art method of blood sugar testing, every six months is the most cost-effective diabetes management, with an additional cost of $11,018.66 for one extra quality-adjusted life-year.