Article Title

Provider-Level Variation and Determinants of Outpatient Generic Drug Prescribing in a Mixed-Payer Health Care System

Publication Date



generic prescribing, provider variation


Background/Aims: While 86% of all drugs in the United States are dispensed as generics, this rate is not uniform across all therapeutic classes and drug products. The role of the health care provider in influencing generic use is not well understood. We sought to quantify variation between providers and provider-specific determinants of generic prescribing in an outpatient setting.

Methods: This was a cross-sectional analysis of electronic health record data from a mixed-payer health care system in northern California in 2013. We focused on 12 therapeutic classes that had the potential for suboptimal generic utilization, including stimulants, antidepressants, thyroids, vasodilators, vascular 5HT1 agonists, antiplatelets, sympathomimetics, smooth-muscle relaxers, eyes/ears/nose/throat drugs, estrogens, antigout agents and cardiac agents. Generic prescribing rates for each class were calculated, regardless of generic availability. For each class, we applied mixed-effect logit regression models with generic prescribed (yes/no) as the dependent variable and random effects (intercepts) for providers. Models included provider-level determinants: provider type (specialist vs. primary care), and continuous measures of average weekly prescription count and patient visits over the previous 3 months. We further adjusted for patient-level characteristics (age, sex, race/ethnicity, insurance, comorbidities, prescription burden). For each model we calculated the intraclass correlation (ICC) as a measure of random provider-level variation and adjusted odds ratio (OR) for provider-level fixed-effect covariates.

Results: We identified 162,819 prescriptions in 2013 for 150,881 patients in the 12 therapeutic classes of interest. Generic prescribing rates across therapeutic classes ranged from 9.3% (sympathomimetics) to 95.1% (antidepressants). After controlling for provider and patient characteristics, significant between-provider variation in generic prescribing was observed; ICC ranged from 7.6% (vasodilators) to 72.9% (eyes/ears/nose/throat drugs). Specialists had lower odds of prescribing generic antidepressants than primary care providers (OR: 0.48), but higher odds of prescribing generic estrogens (OR: 1.55) and sympathomimetic agents (OR: 2.53). Across classes, a provider’s prescription volume or patient volume had little or no effect on generic prescribing.

Conclusion: In this cross-sectional analysis of electronic health record data from a mixed-payer health system, we observed large heterogeneity across providers in outpatient generic prescribing for select therapeutic classes, even after accounting for patient characteristics. Health system initiatives may be warranted to better educate providers on generic prescribing.




June 30th, 2016


August 12th, 2016