Comorbidities as Drivers of Patient Healthcare Utilization Patterns: Uses of Administrative Data in Modeling Disease-Mediated Interactions

Steady increases in healthcare expenditures in the United States have prompted a variety of studies, pilot projects, and initiatives aimed at understanding and influencing the mechanisms that impact healthcare decisions and costs. Likewise, much effort has been spent understanding how individual diseases and comorbidities affect health outcomes. Summary measures such as hospital utilization rate lack granularity as a tool for clinicians or administrators. This work proposes a novel approach for modeling the direct relationship between comorbidities and healthcare utilization patterns. A model that directly interprets sequences of patient visits through the lens of patient diagnoses provides actionable, granular patient-level information that can directly inform clinical decisions, case management, and healthcare system design. Administrative data obtained from a healthcare insurance provider was used as the input for constructing the model. Modeling showed that patient movements between providers were highly patterned, and distinct groups of utilization patterns were observed within comorbidity clusters. Across all morbidities, increases in overall utilization were highly correlated with strings of repeated visits to medical specialists, although this trend was not as strong for psychiatric diagnoses. Lastly, a descriptive modelling framework was created as a potential schematic for clinical usage embedded within patient Electronic Health Record.