Accounting for taste: preferences mediate the relationship between documentation time and ambulatory physician burnout

Abstract

Objectives

Physician burnout in the US has reached crisis levels, with one source identified as extensive after-hours documentation work in the electronic health record (EHR). Evidence has illustrated that physician preferences for after-hours work vary, such that after-hours work may not be universally burdensome. Our objectives were to analyze variation in preferences for after-hours documentation and assess if preferences mediate the relationship between after-hours documentation time and burnout.

Materials and Methods

We combined EHR active use data capturing physicians’ hourly documentation work with survey data capturing documentation preferences and burnout. Our sample included 318 ambulatory physicians at MedStar Health. We conducted a mediation analysis to estimate if and how preferences mediated the relationship between after-hours documentation time and burnout. Our primary outcome was physician-reported burnout. We measured preferences for after-hours documentation work via a novel survey instrument (Burden Scenarios Assessment). We measured after-hours documentation time in the EHR as the total active time respondents spent documenting between 7 pm and 3 am.

Results

Physician preferences varied, with completing clinical documentation after clinic hours while at home the scenario rated most burdensome (52.8% of physicians), followed by dealing with prior authorization (49.5% of physicians). In mediation analyses, preferences partially mediated the relationship between after-hours documentation time and burnout.

Discussion

Physician preferences regarding EHR-based work play an important role in the relationship between after-hours documentation time and burnout.

Conclusion

Studies of EHR work and burnout should incorporate preferences, and operational leaders should assess preferences to better target interventions aimed at EHR-based contributors to burnout.

Background and significance

Physician burnout in the United States has reached crisis levels, impacting more than 65% of physicians.¹ Burnout affects both physicians and patients, in the form of diminished physician job satisfaction, compromised quality of care,² turnover, and even departure from the workforce entirely,^3,4 creating high-cost,⁵ high-risk discontinuities in patient care.⁶ In light of this and the worsening primary care physician shortage,⁷ ameliorating physician burnout has become a major policy priority for US health care, with multiple national efforts aimed at identifying and addressing the antecedents to burnout by the National Academy of Medicine,^8,9 American Medical Association,¹⁰ and incorporation of “improving the work life of health care workers” into the Institute for Healthcare Improvement (IHI) Quadruple Aim, which now also includes advancing health equity (the Quintuple Aim).^11,12

The search for antecedents to physician burnout has yielded several prime suspects, with 2 of the most cited factors being burdensome documentation requirements imposed by the byzantine US billing and reimbursement system^13–15 and the low usability of electronic health records (EHR).^3,4,16–21 In particular, the intersection of these 2 factors manifests in extensive after-hours time spent in the EHR,²² which has been identified as a key risk factor for physician burnout.^16,23–25 However, recent qualitative evidence has shown that physicians vary in their preferred work styles, and some value the flexibility to adapt working hours to fit their lifestyle and preferences, including some instances of “after hours” work.²⁶ Indeed, the greatest amount of variation in EHR use occurs at the individual physician level, even within the same specialty,^27,28 illustrating that there are likely to be physician-level factors (eg, preferences) that influence the relationship between after-hours EHR use and burnout. It is possible that physicians with similar amounts of after-hours EHR time experience that time in vastly different ways given varied preferences for work flexibility. In turn, this range of preferences may translate to differences in burnout levels that cannot be captured solely in measures of after-hours EHR time. Understanding the mediating effects of preferences could guide interventions targeted for individuals who log significant amounts of EHR work in the evenings and feel acutely burdened by that work, making them more likely to reduce their clinical load or leave the organization. However, no studies to date have examined how after-hours EHR time may impact burnout in heterogenous ways after accounting for individual preferences.

Objective

The purpose of this study is to examine the extent to which the relationship between after-hours EHR time and burnout is mediated by physicians’ individual preferences regarding after-hours work. We combine survey data on physician wellbeing and work preferences with detailed EHR usage data capturing after-hours documentation time to fill this gap in the literature. We explore 3 research questions: first, how do physician preferences for after-hours documentation and other potentially burdensome scenarios vary? Second, what is the relationship between these preferences and burnout? And finally, do preferences related to after-hours documentation mediate the relationship between actual after-hours EHR documentation time and burnout?

We hypothesize that preferences mediate the relationship between after-hours EHR time and burnout; our study has important implications whether or not this hypothesis is supported. If our hypothesis is supported, our findings directly inform burnout reduction efforts by health systems and physician practices by making it imperative to systematically collect information about physician preferences to correctly identify the highest-risk physicians and tailor support according to these preferences. Conversely, if the relationship between after-hours EHR time and burnout is not mediated by preferences, efforts to ameliorate burnout should de-prioritize “bespoke” interventions for individual physicians and instead focus primarily on reducing after-hours EHR use in ways that impact broad swaths of the physician population, for example by platform-level improvements to EHR usability, organizational configuration of low-burden documentation workflows, or deployment of artificial intelligence tools for documentation support. In many ways, the policy levers in this second scenario are more central (and therefore more powerful) than the case wherein each physician’s individual preferences play an outsize role on the path from after-hours documentation time to burnout, which would require careful targeting and tailoring of burden-reduction interventions.

Materials and methods

We combined professional wellbeing survey data with EHR usage data to analyze the degree to which physician preferences related to outside-of-clinic documentation mediate the relationship between after-hours documentation time in the EHR and physician burnout. Our wellbeing survey was administered between March 27, 2023, and April 20, 2023, to front-line clinicians at MedStar Health, a large multi-specialty regional health system in the mid-Atlantic with more than 100 primary care and other ambulatory clinic locations. The survey had a 38% response rate among attending physicians (971 total responses from 2554 attending physician invitees). Invitees were emailed links to the wellbeing survey and sent 2 follow-up invitations during the survey administration period if they had not yet responded. The Georgetown University-MedStar Health Institutional Review Board approved this study, and we follow Strengthening Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

Measures: physician preferences and burnout

We developed a 10-item instrument to capture physician preferences for several EHR-related scenarios for which physicians may experience as more or less burdensome depending on their own preferences and workflows. The instrument, which we refer to as the “Burden Scenarios Assessment,” included 10 “scenarios” (eg, completing clinical documentation after clinic hours at home) that the respondent rated as (1) highly burdensome, (2) somewhat burdensome, (3) minimally burdensome, or (4) not burdensome at all (Figure 1). The question stem clarified that ratings of burden should account for the respondent’s own preferences and work style. There was also an option for respondents to indicate that a scenario did not apply to their role. Advanced practice providers and non-attending-level physicians (eg, residents and fellows) were also given the instrument but have high variability in their roles and types of employment across this health system and thus, were less likely to report that the scenarios were applicable to their current roles. Scenarios were developed in collaboration with clinicians and leaders in the MedStar Health Center for Wellbeing. For analysis, we constructed this measure in 2 ways: first, preserving all 4 response levels, and second, dichotomizing the measure into “not or minimally burdensome” (set to 0) and “somewhat or highly burdensome” (set to 1). Missing or “not applicable to my role” responses were dropped for analysis.

To measure physician burnout, we used the burnout index, a component of the Professional Fulfillment Index (PFI), a validated instrument for assessing dimensions of wellbeing.^29–31 The burnout index is computed from 2 subscales of the PFI capturing interpersonal disengagement (6 items) and work exhaustion (4 items). Responses to these 10 items are averaged and standardized into a continuous measure from 0 to 10. A burnout index value of ≥3.325 has been established as the threshold for a binary definition of burnout.³¹ We utilize both the binary and 0-10 continuous index operationalizations of this measure in our analyses.

Measures: after-Hours EHR documentation time

To measure actual after-hours documentation time, we leveraged user-hour active use log data from the Oracle Cerner Advance platform,³² which tracks hourly active time in the EHR for each user across 3 domains (documentation, chart review, and orders).³³ Active EHR use time is aggregated using a stopwatch-like method and is defined by Cerner as “the duration of activity involving three or more mouse clicks per minute, 15 or more keystrokes per minute, or mouse movement of 1700 pixels or more per minute. Activity separated by increments shorter than 45 seconds are considered ‘active’ use.”³³ We extracted this user-hour active use log data (termed “Timecards” in Oracle Cerner Advance) for all physicians at MedStar Health for the period coinciding with the survey administration. For our analysis of documentation preferences and documentation burden, we focused only on the active use time attributable to documentation in the EHR. We first calculated each physician survey respondent’s average total daily active documentation time logged after 7 pm and before 3 am the following day, limiting this to days during which the physician worked exclusively in the ambulatory setting via indicator variables for whether the physician logged any EHR usage in inpatient, emergency, and/or ambulatory settings. Last, we dichotomized average total daily active documentation time at the median, to create a binary measure indicating “below median” or “above median” average daily after-hours documentation time. Physician-days were included in the analysis if they included non-zero total active time (this is not specific to documentation), and physicians were included in the final sample only if they had at least 10 of these “active ambulatory practice days” during the survey administration period (Supplementary Appendix Figure S1).

Statistical analysis

We merged after-hours documentation time data with survey responses at the physician level to facilitate our analysis. For our first research question, we computed descriptive statistics for all burden scenario responses across all physician respondents to illustrate variation in preferences across each scenario. To answer our second question, we visualized rates of burnout (using the binary definition of burnout) across all preference levels for each burden scenario to examine the relationship between these preferences and burnout. Finally, to explore the potential mediating effect of physician preferences, we visualized rates of burnout and the 0-10 burnout index stratified by both preferences and actual after-hours documentation time, focusing only on the after-hours clinical documentation at home scenario. Then, we conducted a mediation analysis to explore the degree to which preferences mediate the relationship between after-hours documentation time and burnout, using the 0-10 burnout index construction to preserve important variation in burnout levels that are masked in the binary burnout metric. We used 3 linear regression models to estimate (1) the relationship between actual after-hours documentation time (defined as above or below median) and our primary outcome of burnout; (2) the relationship between actual after-hours documentation time and our binary construction how “burdensome” the respondent found after-hours documentation (our hypothesized mediator); and (3) model 1 including our mediator as a covariate. If the mediator is significant in models 2 and 3 and actual after-hours documentation time is significant in model 1 but not in model 3, this indicates a mediated relationship between actual after-hours documentation time and burnout via physician documentation preferences.^34,35 We report regression outputs as well as estimates of the average direct relationship, average mediation relationship, and the proportion of the relationship that is mediated. We used an OLS linear regression approach for models 1 and 3, and a linear probability regression for model 2 for ease of interpretation across the models. In all models, we employ heteroskedasticity-robust standard errors. All analyses were conducted using R Statistical Software v4.3.2 and the tidyverse, fixest, and mediation packages, and this study was deemed exempt by the Georgetown University-MedStar Health Institutional Review Board due to the use of de-identified data for analysis.

Results

Our final matched sample consisted of 318 physician survey respondents for whom we also had sufficient EHR active use log data. Female physicians comprised 47.2% of the sample (n = 145), and male physicians comprised 42.7% (n = 131), with 10.1% (n = 31) preferring not to disclose their gender in the wellbeing survey (Table 1). Physicians represented an array of racial and ethnic groups as well as specialties. Most physicians indicated they were full-time employees (87.6%, n = 275), with 58% (n = 184) reporting their share of clinical time to be between 80% and 100%. The overall level of burnout in the sample was low (37.1%) relative to a reported national average of 62.8% in 2021.³⁶ Characteristics other than EHR-related experience and active EHR time did not vary across our categories of actual after-hours documentation time (Table 1). The overall average active after-hours documentation time was 4.0 min per day (SD: 6.3); “below median” physicians averaged 0.18 min (SD: 0.31) of active documentation time after 7 pm, while “above median” physicians averaged 7.9 min (SD: 7.0) of after-hours documentation time.

Preferences varied both within and across the 9 burden scenarios presented in the survey. Physicians reported post-hoc clinical documentation modifications in response to billing inquiries as the least burdensome, with only 16.7% of physicians indicating this scenario as highly burdensome, after accounting for their own preferences and work style (Figure 2). Finding relevant clinical information in the EHR also rated as relatively low burden, with only 18.8% of physicians indicating this scenario as highly burdensome. Across scenarios, completing clinical documentation after clinic hours while at home was the scenario rated most burdensome, with 52.8% of physicians rating this scenario as highly burdensome. Dealing with prior authorization also rated as burdensome, with 49.5% of physicians indicating this as highly burdensome (Figure 2). Burnout rates varied according to preferences, with respondents who indicated a given scenario as highly burdensome consistently having the highest rates of burnout. For example, 44.6% of physicians who indicated after-hours documentation at home as highly burdensome met the criteria for burnout, compared to 15.4% of physicians who indicated after-hours documentation at home as not burdensome at all (Figure 3).

When comparing burnout rates (binary measure) and the burnout index (continuous 0-10 measure) across preferences and levels of after-hours documentation time, we found that physicians with below-median actual after-hours documentation time who also indicated that after-hours documentation was not burdensome had lower rates of burnout (21.4%) compared to physicians with similar levels of after-hours time but indicated that after-hours clinical documentation was burdensome (39.4%) (Figure 4). The same was true among physicians with above-median after-hours time (35.3% of the “not or minimally burdensome” group illustrating burnout vs 41.8%). When we operationalized the burnout score as an index from 0 to 10, we found similar results (Figure 4). Levels of actual after-hours documentation time across burden scenario preferences can be found in Supplementary Appendix Figure S2.

In our mediation analysis, above-median actual after-hours EHR time was associated with a 0.76-point greater score on the burnout index (95% CI, 0.30-1.21, P = .001) and with a 24-percentage point greater likelihood of indicating that after-hours documentation was somewhat or highly burdensome (B: 0.24, 95% CI, 0.15-0.33, P < .001) (Figure 5). When controlling for both variables, the relationship between actual after-hours documentation time decreased (B: 0.49, 95% CI, 0.02-0.96, P = .04), and we found that rating after-hours documentation as burdensome was associated with a 1.09-point higher score on the burnout index (95% CI, 0.57-1.61, P < .001). These findings are consistent with a partially mediated relationship; we estimated that 35% of the relationship between after-hours documentation time and burnout was mediated by preferences related to after-hours documentation (95% CI, 15%-86%, P < .001). Mediation analysis estimates with an alternative specification of our mediator (highly burdensome vs all other preferences) can be found in Supplementary Appendix Figure S3.

Discussion

We found wide variation in physician preferences both within and across EHR-based work scenarios, suggesting that even within the same organization and within a given scenario such as after-hours documentation, physicians do not uniformly experience certain aspects of EHR work as burdensome.²⁶ Between 5% and 20% of physicians rated each scenario as “not burdensome at all,” a non-trivial proportion of the physician population that might be inadvertently prioritized for interventions if those interventions are designed without accounting for varied preferences. We also find that after-hours EHR use time alone does not fully or sufficiently correlate with burnout; rather, variation in preferences for after-hours EHR documentation partially mediates this relationship. This finding supports our hypothesis posited in the introduction and makes it clear that systematic collection of information on physician preferences is crucial to informing burden-reduction efforts both nationally and at individual organizations. The design and implementation costs to organizations for interventions that are tailored to physicians according to their preferences are surely greater than one-size-fits-all interventions like EHR user interface design improvement by vendors or organization-level configurations for documentation workflows. However, these costs may be worth taking on if they translate into improved efficiency (eg, via focus on the highest-risk physicians) and efficacy (eg, via material reductions in burdensome after-hours documentation time and resultant wellbeing improvements) of these efforts. Pragmatically, both levels of intervention are important to pursue; however, our findings may help policymakers and resource-constrained organizations prioritize burden reduction efforts.

Holistically, this study makes 3 important contributions to the physician wellbeing and documentation burden literature. First, we developed a prototype survey instrument, the Burden Scenarios Assessment tool, to systematically collect information about physician preferences regarding different potential sources of burden at scale. This instrument can be modified by interested researchers in 2 ways: by expanding the selection of scenarios and/or by translating it to different clinician populations. For example, in our wellbeing survey, we asked nurse respondents different burden scenarios questions that were pertinent to nursing practice. One can easily imagine a future library of Burden Scenarios Assessment items specific to numerous clinical roles. We encourage health care organizations and researchers to further develop and use this instrument in future work analyzing differences in clinician preferences, to both explore the generalizability of this instrument as instantiated in our study and to assess preferences across new scenarios and populations.

Our second contribution is to illustrate empirically that considerable variation in perceived burden and preferences exists both within and across EHR-based work scenarios. Ratings of “highly burdensome” varied across scenarios, and no scenario had universal “highly” or even “somewhat burdensome” ratings (Figure 2). The fact that perceived burden and preferences vary across scenarios further supports our understanding that different aspects of EHR use impose different costs on physician wellbeing; both organizational leaders and policymakers should therefore prioritize those scenarios that impose the highest costs for intervention. Our findings suggest that the most burdensome scenarios, on average, are after-hours documentation at home and dealing with prior authorization. The fact that no scenario was universally highly burdensome for all physicians also underscores existing evidence that individual physicians vary greatly in both their EHR use patterns^27,28 and their experiences of burden,²⁶ even for the same EHR-based work scenario. This makes it increasingly clear that physician EHR use and experiences exhibit extreme heterogeneity, contrary to claims that physician practice patterns have become overly standardized.³⁷ This finding also suggests that successful interventions to improve workflows, reduce burden, and improve physician wellbeing will likely need to be tailored to individual users in ways that account for extant workflows, preferences around a given EHR work scenario, and personal priorities. For example, initiatives to universally eliminate evening documentation may not be a net wellbeing benefit to physicians who would rather prioritize afternoon family responsibilities (eg, childcare) in exchange for some evening work.

Our third contribution is to show that after-hours documentation time does not fully explain physician burnout, and information on physician preferences around these potentially burdensome work scenarios is crucial to identifying physicians at the greatest risk of burnout. Numerous studies have found a relationship between components of burnout and after-hours documentation time as measured in EHR active use logs^16,17; others have found no relationship.^3,20 We find there to be an independent relationship that is partially mediated by physician preferences, suggesting that analyses that use only actual after-hours documentation time to predict burnout may be subject to non-trivial rates of both false positives (ie, physicians who prefer to document after clinic hours and/or do not find it to be burdensome and are thus not experiencing burnout) and false negatives (ie, physicians who have low levels of after-hours documentation but find it particularly burdensome and therefore are experiencing burnout). Going forward, we encourage researchers and operational leaders aiming to ameliorate EHR-derived physician burnout to collect information about physician preferences to assess more accurately who is at the greatest risk.

Limitations

Our study has several important limitations. Primarily, we study burnout and EHR use among a relatively small sample of physicians in a single, albeit geographically diverse, health care organization, so understanding the generalizability of these findings will require additional research across organizations and in larger samples of physicians. Second, this sample of ambulatory physicians is limited to respondents to MedStar Health’s annual wellbeing survey, which may introduce response bias, recall bias, salience bias, and social desirability bias into our measures of preferences and physician burnout. Third, our Burden Scenarios Assessment tool is not a formally validated instrument for capturing physician preferences and will require validation if it is to be more widely adopted. While we focused on a physician population for this tool, future studies could consider a similar approach to develop and validate burden scenarios relevant to advanced practice providers (eg, nurse practitioners, physician assistants), resident physicians, and/or physicians in non-ambulatory care settings. Finally, our analyses are all associational, thus preventing causal interpretation of our estimates for the relationships between actual after-hours documentation time and preferences as well as between preferences and physician burnout.

Conclusion

We studied physicians’ preferences regarding a set of potentially burdensome EHR-based work scenarios and found wide variation both within and across scenarios. In the context of after-hours documentation time (the most burdensome scenario), physician burnout was correlated with these preferences such that preferences mediated more than one-third of the direct relationship between actual after-hours EHR documentation time and burnout. Our findings suggest that efforts to ameliorate burnout via reductions in after-hours documentation time should take physician preferences into account, as this work is not universally burdensome. Accounting for taste can more efficiently allocate scarce resources to the most burdened physicians who are at the highest risk of burnout.

Author contributions

All authors contributed to the study conception, design, and manuscript preparation. Daniel Marchalik, Heather Hartman-Hall, and Nate C. Apathy facilitated the survey data collection; Nate C. Apathy performed all analyses; Nate C. Apathy and Dae Hyun Kim interpreted results.

Supplementary material

Supplementary material is available at Journal of the American Medical Informatics Association online.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Conflicts of interest

The authors have no competing interests to declare.

Data availability

The data underlying this article cannot be shared publicly, as it was collected as a part of the MedStar Health annual wellbeing survey and can only be reported via aggregate statistics due to its sensitive nature. The data may be shared on reasonable request to the corresponding author.

References