Extracting phenotypes from patient claim records using nonnegative tensor factorization

Electronic health records (EHRs) are becoming an increasingly important source of patient information. Unfortunately, EHR data do not always directly and reliably map to medical concepts that clinical researchers need or use. Some recent studies have focused on EHR-derived phenotyping, which aims at mapping the EHR data to specific medical concepts; however, most of these approaches require labor intensive supervision from experienced clinical professionals. In this paper, we use Limestone, a nonnegative tensor factorization method to derive phenotype candidates from claims data with virtually no human supervision. Limestone represents the interactions between diagnoses and procedures among patients naturally using tensors (a generalization of matrices). The resulting tensor factors are reported as phenotype candidates that automatically reveal patient clusters on specific diagnoses and procedures. To the best of our knowledge, this is the first study that successfully extracts useful phenotypes by applying sparse nonnegative tensor factorization to a large, public-domain EHR dataset covering a broad range of diseases. Our experiments demonstrate the interpretability and the promise of high-throughput phenotypes generated from tensor factorization.