Health Services/Policy
Quantifying the primary care workforce in the U.S: A validation study with and without an imperfectly measured referent standard Nicole Rafalko* Nicole Rafalko Scott Siegel Paul Yerkes Jan Eberth Igor Burstyn Neal Goldstein
Purpose: To validate the ability of a commonly used data source, the National Provider Identifier (NPI), to identify primary care physicians, physician assistants (PAs), and nurse practitioners (NPs).
Methods: Validation studies were conducted calculating the sensitivity, specificity, and associated 95% confidence intervals for physicians, PAs, and NPs. For physicians, Medicare claims data was used as an imperfectly measured referent standard. For PAs and NPs, we used a simulation-based method to estimate sensitivity, specificity, and associated 95% credible intervals that assumes the NPI and Medicare claims are equally error-prone.
Results: Using the Medicare claims as the referent standard for physicians yielded a sensitivity and specificity of 0.9 (95% CI: 0.8, 1.0) and 0.8 (95% CI: 0.7, 0.8), respectively. Using the simulation-based method yielded a sensitivity and specificity of 0.6 (95% CI: 0.1, 1.0) and 0.5 (95% CI: 0.1, 1.0), respectively for physician assistants and 0.6 (95% CI: 0.1, 1.0) and 0.6 (95% CI: 0.1, 1.0), respectively for nurse practitioners.
Conclusions: Our validation results varied by provider role, with the highest accuracy observed for physicians further highlighting the challenges in quantifying PAs and NPs. Failure to consider potential misclassification in the NPI may result in biased health services research.