Abstract
Introduction/Objective:
As fewer physicians enter primary care, Advanced Practice Providers (APPs) play an increasingly important role in meeting the demand for patient care. Factors contributing to work status changes among primary care APPs are not well studied. We aimed to evaluate differences in primary care APP work status across electronic health record (EHR) workload measures.
Methods:
We conducted a retrospective cohort study within a Midwestern healthcare system from 2020 to 2022, evaluating the relationship between primary care APP work status defined as retention at current full-time equivalent (FTE), reduction of FTE within primary care, or relocation outside of primary care and 3 time-based EHR workload metrics. Descriptive statistics, univariate, and multivariable analyses were reported.
Results:
There were no differences between work status (retained, reduced, or relocated) across the EHR workload measurements of in-basket time per appointment, system time per appointment and note time per appointment (P > .05) on univariate analysis. With multivariable analysis, there was no difference in likelihood of combined relocation/reduction adjusting for EHR workload variables (P > .05).
Conclusion:
The EHR workload was not associated with work status changes for primary care APPs in our study. Further exploration is needed to identify factors that may contribute to reduction in primary care APP workforce capacity.
Keywords
Introduction
In the contemporary health care environment, nurse practitioners (NPs) and physician assistants (PAs), together referred to as Advanced Practice Providers (APPs), are filling a critical role as primary care clinicians.1,2 APPs deliver primary care amid the national trend of fewer physicians entering primary care practices.3,4 While studies have not quantified the overall consequences of primary care APP-specific retention on outcomes such as care quality, cost, and workforce stability, understanding and addressing APP work status change remains essential to advancing the quadruple aim to improve patient experience, reduce healthcare cost, advance populations health, and improve provider experience.5-7
The turnover of primary care clinicians is complex and multifactorial. Burnout has been identified as a predictor of primary care clinician turnover. 8 Another proposed contributor is the increasing clerical burden,9-11 coinciding with the implementation of electronic health record (EHR) systems in healthcare. Melnick et al 12 examined turnover of ambulatory physicians in relation to objective measures of productivity and EHR use metrics. Counter to their hypothesis, they found that less time spent in the EHR was associated with increased turnover. Although a scoping review of 25 studies 13 including physicians and APPs demonstrated an association between EHR use and clinician burnout, these studies did not evaluate clinician turnover. To our knowledge, prior studies have not evaluated associations between EHR use and primary care AAP turnover.
In this study, we aim to address the gap in knowledge about primary care APP turnover by evaluating objective EHR workload and differences in turnover of primary care APP staff. We define turnover in our study as work status change with either a reduction in full-time equivalent (FTE) status within primary care or as a relocation outside of primary care. Both changes have a negative impact on primary care APP staffing, in contrast to retention which we define as remaining in one’s primary care position at current FTE. Our study dates overlap with the COVID-19 pandemic, a relevant factor given reported increases in intent to leave 14 and actual exiting 15 from healthcare worker roles during this time, especially among females. We hypothesize that there would be a difference in APP work status across the EHR workload variables with higher rates of reduction/relocation if greater EHR workload is noted.
Methods
This retrospective cohort study utilized data from the Epic Signal Platform to evaluate 3 EHR workload metrics among primary care APPs and their association with turnover. This study was reviewed and approved by the study site’s Institutional Review Board. The study included data from January 2020 through December 2022, capturing trends over time among primary care APPs within a large Midwestern healthcare system and academic medical center.
Study Sample
Family Medicine, Community Pediatrics, and Community Internal Medicine practices across a range of urban, suburban, and rural sites within the Midwest were included. Eligible participants were APPs employed at the institution during the study period who maintained a continuity panel of 200 or more patients in a primary care clinical practice setting. Those with panel sizes below 200 were excluded due to the likelihood of being new to practice or working in geriatric homebound or skilled nursing settings, which do not reflect typical daily clinical practice. Those who retired were excluded due to low numbers. Finally, APPs with managerial, temporary, float, or urgent care-only roles were also excluded.
Data Sources and Measures
Human Resources provided datasets identifying APPs employed in primary care, their FTE, and work status patterns. Turnover, or a change in work status, was defined as either a reduction in FTE status or a relocation. A reduction in FTE referred to APPs who decreased their work hours during the study period but remained in primary care, whereas relocation encompassed transfers to other departments within or outside of the organization and termination. When multiple events occurred within the same year (eg, an FTE reduction followed by relocation), assignment was based on the first observed event. Additionally, APP retention was assessed and defined as maintaining both FTE status and primary care position throughout the study period. Panel size for APPs was obtained from the institution’s Primary Care Population Health database. Human Resources data and Population Health data were then merged to create a comprehensive dataset capturing APPs with continuity panels and their employment status during the study period.
Relevant EHR workload metrics were collected using Signal data, an Epic Systems tool designed to capture clinician EHR use. Inclusion of Signal data required the providers to meet minimum thresholds of EHR activity. For basic inclusion, providers needed at least 1 scheduled appointment and at least 1 login to the EHR per month. Calculation of per-appointment metrics required a minimum of 5 scheduled appointments per week. Data were summarized in monthly increments and were derived from several sources including the User Action Log (UAL), scheduling records, note attribution and composition, and in-basket activity. The UAL captures user actions such as mouse clicks and keystrokes, as well as time spent across different EHR activities throughout the day.
Clinical workload variables were collected monthly from the Epic Signal database and included (1) in-basket time per appointment, (2) note time per appointment, and (3) system time per appointment. Averages were calculated over the calendar year or up until the APP reduced or relocated for each measure.
In-basket time per appointment was calculated monthly by summing the total minutes spent in in-basket activities (numerator) and dividing by the total number of scheduled appointments during the same period (denominator). Providers were included if they had at least 5 appointments scheduled per week.
Note time per appointment is the average number of minutes a provider spent writing, viewing, and editing clinician-authored patient documentation within the notes activity per scheduled appointment. Inclusion criteria mirrored those used for in-basket metrics. Monthly values were calculated by dividing the number of minutes of providers spent in notes-related activities by the number of scheduled appointments.
System time per appointment is the average number of minutes the provider actively interacted with any aspect of the EHR per appointment, including in-basket and notes plus all other categories of EHR work. Active time was captured in minutes, reflecting direct engagement with the EHR. Monthly values were calculated by dividing the number of total active minutes providers spent in the EHR by the total number of scheduled appointments.
Statistical Analysis
APPs were grouped each year into 1 of the 3 outcome categories (retained, reduced, and relocated), and those work status descriptions were reported by frequency and percentage of the total sample that year. In-basket time per appointment, note time per appointment, and system time per appointment were reported by retained, reduced, and relocated outcomes for each year using mean and standard deviation (SD). Kruskal-Wallis tests were used to compare any differences between the 3 work status groups each year across the EHR workload variables.
To compare whether these EHR workload per appointment measurements had an impact on whether someone retained, reduced, or relocated, generalized linear mixed models were completed with the three-group outcomes collapsed into a binary outcome of either reduced/relocated or retained. We collapsed reduction and relocation for this analysis given the small number of events per category, to simplify the interpretation of the outcome, and because they were both considered to have a negative impact on APP staffing. Each time per appointment measurement was fitted in its own model due to multicollinearity issues (Variance Inflation Factor greater than 5 for system time per appointment and note time per appointment and correlations –0.7 and below for the same variables). Panel size was adjusted as a fixed effect. The repeated measures for APPs over the 3 years were taken into account with a random effect and a covariance structure of compound symmetry was utilized. Odds ratios and 95% confidence intervals were reported. P-values less than .05 were considered statistically significant. Analysis was completed using SAS (SAS Institute Inc., version 9.4).
Results
The number of APPs varied across each year of the study period (Table 1). The FTE- reduced and relocated groups were smaller each year compared to the retained group. The average panel size for each primary care APP was 753 patients (SD = 383.5) for 2020, 756 patients (SD = 399.5) for 2021, and 659 patients (SD = 377.2) for 2022. Across each studied year, approximately 80% of APPs maintained their FTE status annually. Among turnover types, FTE reduction was most common. The percentage of APP FTE reduction by year was 10.9%, 12.7%, and 17.0% respectively for 2020, 2021, and 2022. The average FTE reduction for those that reduced FTE was –0.2 for every year (SD = 0.17 for 2020, SD = 0.17 for 2021, SD = 0.16 for 2022).
APP Work Status Demographics.
Across each of the 3 years, mean EHR workload metrics (Table 2) were similar among APPs who retained FTE, reduced FTE, or relocated, with no statistically significant differences observed (all P > .05). Although relocated APPs had slightly higher note time in 2021 and 2022 compared to other groups, these differences did not reach significance. Overall, workload metrics were stable with minor year-to-year variations but no clear differences across work status groups.
Differences in APP Work Status across Electronic Health Record Workload.
Kruskal-Wallis test p-values.
In-basket time, system time, and note time are all in minutes.
System time/Appt and Note time/Appt were missing for 1 patient, so the N = 180 in those cases;
The generalized linear mixed models (Table 3) showed that there was no significant impact on whether an APP was retained or relocated/reduced for any of the three time per appointment EHR workload variables (all P > .05). Panel size was also shown to not be a significant factor.
Likelihood of Reduction/Relocation Adjusting for Electronic Health Record Workload Variables.
Generalized linear mixed models; Outcome was binary for reduced/relocated or retained.
Discussion
We examined change in work status patterns of retained, reduced, and relocated FTE in relation to objective EHR workload metrics among primary care APPs at a large Midwestern academic medical center. We observed no statistically significant differences in EHR workload measures across different work status categories, even after adjusting for panel size.
The study period encompassed the first 3 years of the COVID-19 pandemic, a time marked by heightened workforce strain across healthcare settings. Early in the pandemic, national survey data indicated substantial intent among APPs to alter their work arrangements, with 28.9% of APPs reporting an intent to reduce hours and 33.0% reporting an intent to leave their current practice within 2 years. 16 Consistent with these concerns, national longitudinal data from 2010 to 2021 demonstrated substantial APP mobility: 14.4% of NPs and 15.4% of PAs changed practice settings within 1 year of first billing Medicare, and nearly 30% relocated practices within 3 years. 17 In contrast, observed turnover in our institution was considerably lower, with fewer than 7.0% of APPs leaving primary care during the study period.
Our study results suggests that factors beyond measurable EHR workload may play a substantial role in retention. Prior survey-based studies have highlighted the importance of broader contextual factors including practice environment, administrative support, physician relationships, and opportunities for professional autonomy in shaping APP satisfaction and retention.18,19 Consistent with this literature, a multi-state survey by Poghosyan et al 20 reported that although 22% of primary care NPs intended to leave their job within 1 year, more supportive work environments were associated with higher job satisfaction and lower turnover intention. 20
Strengths of this study include the use of multi-year data and reliance on objective EHR-generated workload metrics rather than self-reported measures. Work status changes were clearly defined and systematically categorized, enhancing interpretability and reproducibility. In addition, this study contributes to a relatively limited body of literature examining workload and staffing patterns among primary care APPs.
Limitations of this study include the small number of APPs in the reduced and relocated FTE groups which may have limited statistical power to detect differences in these work status categories. The study was conducted within a single academic health system, which may limit the generalizability of findings. Demographic information of study participants was not collected, limiting assessment of whether work status changes differed between demographic groups. Finally, while EHR time metrics provide objective measures of workload, they do not capture all dimensions of clinician workload such as cognitive burden, interruptions, or emotional demands of patient care.
Future research should explore additional predictors of APP turnover, such as practice environments, team structures, clerical burden, and career development which have been identified in prior studies as influential determinants of APP satisfaction and turnover.20,21 Qualitative studies may also offer deeper insight into the organizational and relational factors that shape APP retention.
Conclusion
Understanding work status change within the primary care APP workforce is critical for identifying opportunities to strengthen clinician retention. Unexpectedly, we observed no differences in multiple EHR time-based workload measurements across the different work status categories, suggesting further research is warranted to explore the potential drivers of turnover in this subset of the primary care workforce.
Footnotes
Acknowledgements
None.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This publication was made possible by the Mayo Clinic CTSA through grant number UL1TR002377 from the National Center for Advancing Translational Sciences (NCATS), a component of the National Institutes of Health (NIH).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Ethical Considerations
The Mayo Clinic Institutional Review Board reviewed this study #23-000486 and determined it to be exempt.
Informed Consents
The Mayo Clinic Institutional Review Board reviewed the study and determined it to be exempt #23-000486. Therefore informed consent was not required.
Consent for Publication
Consents for publication are not applicable.
Data Availability Statement
Data may be shared upon reasonable request of the corresponding author*.