Modifiable Arial Unit Problems for Infectious Disease Cases Described in Medicare and Medicaid Claims, 2016-2019

Abstract

Abstract Introduction: Modifiable Arial Unit Problems are a major source of spatial uncertainty, but their impact on infectious diseases and epidemic detection is unknown. Methods: CMS claims (2016-2019) which included infectious disease codes learned through SNOMED CT were extracted and analyzed at two different units of geography; states and ‘home to work commute extent’ mega regions. Analysis was per member per month. Rolling average above the series median within geography and agent of infection was used to assess peak detection. Spatial random forest was used to assess region segmentation by agent of infection. Results: Mega-regions produced better peak discovery for most, but not all agents of infeciton. Variable importance and Gini measures from spatial random forest show agent-location discrimination between states and regions. Conclusions: Researchers should defend their geographic unit of report used in peer review studies on an agent-by-agent basis.