Predicting Provider Specialty For Nurse Practitioners And Physician Assistants

Medicare claims do not include descriptive provider specialties for Nurse Practitioners (NPs) and Physician Assistants (PAs). Missing NP/PA provider specialty information is an issue for both researchers and Medicare’s Accountable Care Organization (ACO) patient attribution logic. Researchers are sometimes forced to exclude NP/PA visits from analyses because it is unclear if the visit is clinically relevant to the research objective. Inpatient attribution, all NP/PAs are considered primary care practitioners even though 73% of NPs/PAs at our local health system work in specialty care. Given that specialty care NPs/PAs are included in attribution, patients can be attributed to the incorrect ACO. NPs/PAs are increasingly caring for patients in both primary and specialty care, so these issues for researchers and ACOs will only continue to proliferate. This analysis had two aims: 1) use Medicare claims data to create a collection of models to predict the provider specialty for NPs/PAs using information from our local hospital’s electronic health record and 2) create a single, easy-to-use model that can be used to find most of the NPs/PAs that work in primary care. The predicted specialties from either aim could be used to update the attribution algorithm to ensure that specialty and primary care NPs/PAs are included only when appropriate. The modeling strategy from aim one predicted the specialty for 80.4% of individual NPs/PAs, 97.4% of individual primary care NPs/PAs, and 94.6% of the NPs/PAs’ associated Medicare claims. The model from aim two identified 84.6% of NPs/PAs that practiced in primary or non-primary care.