FAA Office of Aerospace Medicine
Civil Aerospace Medical Institute

Report No: DOT/FAA/AM-22/10

Title and Subtitle: Pilot Medical Certification Period Health State Forecasts

Report Date: September 2022

Authors: F. Bradbury; G. Chesterton; S. Chin; K. Sarkhel; D. Slater; Z. Slater; B. Wellner. All authors from The MITRE Corporation

Abstract: The Federal Aviation Administration (FAA) Office of Aerospace Medicine supports research to use available healthcare data to inform policies regarding pilots' medical certifications. The MITRE Corporation's Center for Advanced System Development (MITRE CAASD) was asked to examine methods in advanced data analytics and machine learning (ML) to inform such risk-based decision-making. As an initial step in assessing the potential predictive value of commercially available healthcare data, the FAA provided MITRE CAASD the IBM MarketScan dataset—a large set of commercial healthcare claims records. Using this dataset, we developed methods for identifying health status and changes in health status across conditions; for measuring changes in health status among enrollees with diabetes mellitus (DM); and for measuring the onset of new cases of DM, traumatic brain injury, sleep apnea, and chronic obstructive pulmonary disease. We developed a repeatable workflow and modeled these conditions using a wide range of ML methods. We conclude that ML-based predictive modeling of health conditions from IBM MarketScan data is feasible and informative. However, additional clinical information from commercially available electronic health records would likely improve accuracy and more closely align with future FAA needs.

Key Words: Aviation medicine, safety and health, medical claims data

No. of Pages: 39