This retrospective cross-sectional study was approved by the University of Wisconsin’s institutional review board. We collected monthly EHR usage (Epic Systems; Verona, WI), scheduling, panel, and demographic data for all PCPs who practiced at UW Health—a large academic medical center serving over 290,000 medically homed patients—from April 2019 to March 2022. We calculated five measures of monthly EHR use for each physician and normalized each measure per 8 hours of scheduled clinic: total EHR time (EHR8), time in notes (Notes8), time in inbox (Inbox8), time in orders (Orders8), and time outside scheduled clinic hours or “Work Outside of Work” (WOW8).5

We performed unadjusted comparisons of physician demographics, workload, and EHR use with χ2 tests for count data and Mann-Whitney U tests for continuous variables. We used multivariable linear regression to perform adjusted comparisons of EHR use by gender. We modeled gender differences in EHR use during the last year of the study (April 2021–March 2022), controlling for physician specialty, duration of employment, panel size, monthly visit volume, normalized visit volume (Visit8), and combined normalized volume of four labor-intensive inbox message types: laboratory results management, medication management, telephone calls, and patient portal messages (Msg8). We modeled longitudinal changes in the gender gap by revising the above models to include data from all three study years (April 2019–March 2022), adding a term for the interaction between physician gender and study year, and controlling for year and month to account for potential seasonality. Each model used random effects to account for longitudinal correlation in individual physician’s EHR use. Data from March to May 2020 were excluded from analysis due to the transitory changes in EHR use early in the COVID-19 pandemic.6 Vendor-provided EHR usage data were missing for February and April 2021, so these months were likewise excluded.