| Summary: | Currently, some serious comorbidity-impacted chronic diseases have high incidence among older people in the U.S. Due to the incompleteness of the related clinical data, it is difficult to refine these diseases' staging and quantitatively assess risk effects of comorbidities on symptom progressions. Here, we used both electronic medical records (EMRs) and claims data to obtain a comprehensive data source in this paper. We adopted osteoarthritis (OA) as a demonstrated major disease. The key comorbidities and their risks for various OA stage-related progressions were estimated. We utilized the linked EMR-claims dataset of OA from 2007 to 2014. The EMR data provided pain scores and laboratory data, while claims data provided costs as a proxy for disease severity. Although both datasets contained diagnoses, procedures, and medications, the linked dataset included more distinct codes. We established a prototype to combine our developed relational dependency network (RDN) approach with Cox proportional models to extract and estimate key comorbidities' impacts on OA progression. We identified the key OA stage-related comorbidities. Our studies indicate that the combination of the EMR with claims data is a useful strategy for obtaining more accurate medical data sources from patients. The analyses of the impact of clinical factors on OA staging clarify the associations between key covariates and OA progression. These approaches can be generalized to summarize the impact of comorbidities on the development of various chronic diseases.
|
|---|
