Physician Burnout and Timing of Electronic Health Record Use

1Department of Medicine, University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin, United States 2Primary Care Academics Transforming Healthcare Collaborative, UW Health, Madison, Wisconsin, United States 3Department of Family Medicine and Community Health, University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin, United States 4Division of Hospital Medicine, Department of Pediatrics, University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin, United States 5Department of Biostatistics and Medical Informatics, University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin, United States 6Department of Family and Community Medicine, University of New Mexico Health Sciences Center, Albuquerque, NewMexico, United States


Background and Significance
Burnout among primary care physicians is a major problem in the United States. Forty-eight percent of internal medicine physicians reported at least one symptom of burnout in 2017, compared with 43.9% of physicians overall. 1 In addition to negatively impacting physician health, burnout is also associated with reduced quality of patient care and increased physician turnover, further contributing to health care costs and the shortage of physicians. [2][3][4][5][6] While many issues can contribute to physician dissatisfaction and burnout, the electronic health record (EHR) increasingly is implicated as a major source of dissatisfaction. [7][8][9][10] Although the EHR offers some benefits to practice, it also pulls physicians' time away from meaningful face-toface patient interactions toward data entry and asynchronous tasks, such as telephone/patient-portal care. 7 Recent studies confirm that the EHR occupies a large portion of physicians' lives, both during and after traditional workhours. One study found that physicians spent almost half of their work-hours engaged with the EHR. 11 Another study found that primary care physicians spent almost 6 hours in the EHR per day, including 4.5 hours during, and 1.4 hours outside of, normal clinic hours. 12 While prior research has identified the link between perceived EHR workload and burnout, it is not clear whether actual patterns of physicians' EHR use are associated with burnout. If such a relationship exists, it would help clarify the role that changing EHR use could have on reducing burnout and perhaps illuminate specific patterns that are particularly related to physician stress.

Objectives
We conducted a study to determine the association between physician burnout and timing of EHR use in an academic internal medicine primary care practice. We hypothesized that increased time spent in the EHR, particularly outside of normal work-hours, would be associated with higher levels of burnout.

Study Population
Eligible participants included all 58 primary care internal medicine physicians who practiced and managed a primary care clinic panel at 9 of the 10 University of Wisconsin (UW) Health General Internal Medicine (GIM) clinics in the Madison, Wisconsin area. One rural clinic was excluded from this study due to its distinctly different practice environment and workflows. All physicians at these clinics have academic positions at UW-Madison, with practices ranging from full-time clinical work to hybrid practices with variable amounts of teaching, research, and administrative responsibilities. Physicians' clinical work is structured within half-day sessions. At the time of this study, full-time clinicians (1.0 full-time equivalents or [FTE]) were expected to see patients for 27 hours per week in 9 half-day sessions averaging 2.0 patients per hour, with one half-day administrative time per week; providers working clinically less than full-time have reductions in increments of half-day sessions. Full-time physicians are expected to care for a weighted panel 13 of 1,800 patients, which is adjusted downward proportional to their in-clinic time. Panel is defined as the group of patients assigned to a specific physician as their primary care provider, and for whom that physician team is expected to provide appropriate primary care, including office visits and non-face-to-face care (i.e., responding to telephone and patient-portal messages). Patients are included on a physician's panel as long as they have at least one clinical encounter within the organization in the previous 3 years. The study was approved by the UW-Madison Health Sciences Institutional Review Board.

Study Measures and Data Collection
Burnout and sociodemographic data (age, marital status, years since residency, years employed at UW) was collected via an electronic survey sent to all eligible physicians in November 2017. Two reminders were sent approximately 1 and 3 weeks later. Burnout was measured with a single item asking physicians to use their own definition of burnout and rate their feelings on a 5-point scale. Response options included the following: (1) I enjoy my work. I have no symptoms of burnout; (2) I am under stress, and don't always have as much energy as I did, but I don't feel burned out; (3) I am definitely burning out and have one or more symptoms of burnout (e.g., emotional exhaustion); (4) the symptoms of burnout that I am experiencing won't go away. I think about work frustrations a lot; and (5) I feel completely burned out. I am at the point where I may need to seek help. This single-item measure was chosen for several reasons. First, it is identical to item used to measure burnout in the Physicians Worklife Survey 14 as well as the Minimizing Error, Maximizing Outcome (MEMO) study, a multisite study that showed significant relationships between burnout and lower job satisfaction, greater time pressure, poorer work control, more workplace chaos, lower work-life balance, and greater intent to leave the practice. 3,15 This measure of burnout was also found in another study to be associated with physicians leaving their institution in the subsequent 2 years. 6 Second, this single-item measure correlates with the emotional exhaustion component of the Maslach Burnout Inventory (MBI). [16][17][18] While it does not correlate strongly with the other two subscales of the MBI (depersonalization and personal accomplishment), 16 prior studies found that the emotional exhaustion subscale, and not the depersonalization subscale, was associated with physicians reducing their work effort 4 and leaving practices. 19 Lastly, this singleitem measure has been promoted by the American Medical Association as part of their STEPS Forward program as a method for organizations to measure burnout. 20 EHR time and visit volumes were determined for all eligible physicians using retrospective automated data from UW Health's EHR system (Epic Systems Corporation, Verona, Wisconsin, United States). One EHR database, event logging records, automatically stores information each time a clinician accesses or moves between modules in the EHR, including the time that the specific process occurs. Sequential time stamps can then be used to determine the time providers spend in the EHR. Time for each individual process was truncated at 90 seconds to minimize inclusion of idle time when the EHR was not engaged by the user. 12 EHR data were captured for the 12 months prior to the burnout survey administration. Since we were interested in the work done by primary care outpatient physicians, we included only EHR data when physicians were logged into their outpatient clinic environment, and excluded data directly related to inpatient practice.
EHR time was categorized as occurring either during the workdays, after-hours-weekdays, or weekends/holidays. The workday was defined as 7:30 am to 5:29 pm, to capture variation in clinic start times (ranging from 7:50 am to 8:00 am) and end times (no patients are scheduled after 5 pm), and included some time before/after clinic to review and finish charting. Workdays were further subdivided into morning (7:30 am to 12:29 pm) and afternoon (12:30 pm to 5:29 pm) half-day sessions. Billing data were used to categorize the time for each half-day session as occurring in-clinic ( 1 billed office visit during the half-day period) or out-of-clinic (no office visit billed during that half-day period). Visits done by residents were assigned to the supervising (billing) attending. The average EHR time per in-clinic and out-of-clinic session was then calculated by dividing in-clinic and out-of-clinic EHR times by the number of in-clinic and out-of-clinic sessions, respectively. Similarly, the average EHR time for after-hours-weekdays and weekend/holidays was divided by the number of weekday and weekend/holiday days, respectively. The number of days was further adjusted for physicians who started working at UW Health, or took a leave of absence, within the EHR time period of interest.
Administrative records were used to collect physicians' sex and clinical FTE (cFTE). For simplicity, cFTE was defined as the proportion of 10 half-day sessions a provider was expected to be in clinic seeing patients per week.

Analyses
Descriptive statistics were used to describe participants' demographic and practice characteristics, degree of burnout, and timing of EHR use. Burnout was treated as an ordinal (ordered) variable. While most previous studies used burnout as a dichotomous variable, we chose to use it as an ordinal variable to retain potentially important differences between providers with varying levels of responses, and to maintain power in our analysis. 21 Bivariate analyses using nonparametric testing (Spearman's rho for two continuous variables, Mann-Whitney U-test for continuous-dichotomous variables) examined the unadjusted relationship between burnout/EHR use and demographic and practice characteristics. Ordinal regression was then used to determine the bivariate and adjusted relationships between burnout (dependent variable) and EHR times. The adjusted model included all EHR time categories (excluding total EHR time) as independent variables, as well as potential confounders. Age and sex were automatically included as potential confounders as they were frequently associated with burnout in prior studies. 2 Visit volume, defined as the total number of visits, was included to adjust for workload; visit volume was felt to adjust for workload better than cFTE, as it changes based on time away from clinic (i.e., due to vacations or conferences), whereas cFTE does not. We included additional demographic/practice characteristics as covariates only if they were significantly related to EHR use or burnout in bivariate analyses (p < 0.05). The analyses used EHR data in the 3 months prior to the survey, as this time period was expected to be more relevant to our one-time measurement of burnout, and long enough to reduce susceptibility to month-to-month schedule variations. In addition, sensitivity analyses using different EHR time periods (1, 6, and 12 months prior to the survey) showed similar results. Analyses were conducted using SPSS v. 25

Characterization of Electronic Health Record Use
Of the 58 eligible physicians for this study, most were female (62%) with cFTEs ranging from 0.6 to 0.8. EHR data were available for all 58 physicians, while only 34 (59%) completed the survey. There were no significant differences in sex, cFTE, number of clinic visits, and EHR use between those that did and did not complete the survey (►Table 1).
The majority of EHR time occurred when physicians were in clinic, where they used the EHR for an average of 141 minutes (2.4 hours) per each half-day session (►Table 1). Physicians saw an average of 5.6 (standard deviation [SD]: 0.8) patients per session. EHR usage for other time categories was positively skewed, and was relatively low during weekdays when physicians were not in clinic (median: 27 minutes per half-day session, range: 1-129), after-hours during the week (median: 17 minutes per weekday, range: 0-109), and after-hours on weekends/holidays (median: 16 minutes per weekend/holiday day, range: 0-186). While it is difficult to determine the time spent in EHR on a typical weekday because of the variability in physicians' schedules, we estimate that physicians spent approximately 299 minutes (5 hours) on the EHR during a day when they were in clinic for both morning and afternoon sessions (average: 141 minutes for each half-day session, plus a median of 17 minutes per weekday after-hours).
EHR use varied according to physicians' volume in practice, with higher cFTEs and higher visit volumes associated with more total EHR time, half-day in-clinic session EHR time, and weekend/holiday EHR time (►Table 2). Out-ofclinic session EHR time was significantly associated with a higher number of visits, but not with cFTE. No significant relationships were seen between EHR time categories and physicians' age, sex, marital status, or years in practice.

Burnout and Electronic Health Record Use
Among the 34 survey respondents, the average burnout rating was 2.3 (SD: 0.9) on the 1 to 5 scale where higher ratings indicate more burnout. Most respondents scored a 2 (n ¼ 17, 50%) on the scale, while only one physician endorsed a 5 (the highest level of burnout). In bivariate analyses, there were no significant relationships (p < 0.05) between burnout and sex, age, marital status, years since clinical training, ACI Open Vol. 4 No. 1/2020 years at UW Health, cFTE, and visit volume. In unadjusted analyses, higher levels of burnout were associated with more total and in-clinic EHR time (►Table 3 and ►Fig. 1). In models adjusting for age, sex, cFTE, visit volume, and other categories of EHR time, burnout was significantly associated only with in-clinic EHR time and not with out-of-clinic or after-hours EHR time (►Table 3).

Discussion
In this small study of physicians in an academic primary care internal medicine practice, we found that burnout was associated with the amount of EHR use that occurred during clinic sessions. This finding may indicate that physicians with more in-clinic EHR use feel a greater intrusion of the EHR  pulling them away from meaningful face-to-face interactions toward computer-time and data-entry tasks. 7 Our result is consistent with prior research reporting that increased clerical burden is associated with burnout and lower job satisfaction. 22 Certainly, some of the variability in physicians' in-clinic EHR use may be related to different workloads in face-to-face and asynchronous care, which is reflected in our finding that a higher visit volume was associated with more EHR time in most categories. Efforts to understand the amount of EHR time required to care for specific panels of patients may help normalize productivity expectations, and could prevent overly burdensome workloads. However, variability in EHR use may also be related to inefficiencies in EHR-related proficiency and workflows, even among physicians in the same practice setting. Further research could help determine best practices associated with EHR efficiency, as well as identify specific EHR tasks that particularly increase in-clinic EHR time and burnout. This could lead to targeted physician-and organization-level interventions focused on improving EHR efficiency, offloading work from physicians, and better integrating computer-oriented tasks during face-to-face encounters. Indeed, physician satisfaction has been shown to improve with interventions that distribute EHR tasks to other members of the health care team, 23,24 enhance physicians' EHR efficiency through proficiency training, 25,26 and improve note documentation. 27 Our results did not show a consistently significant relationship between after-hours EHR work (often referred to as "pajama time") and burnout. These results are not consistent with prior findings that self-reported time spent at home on work-related tasks negatively impacts physician satisfaction and burnout. 2,28 Possible reasons for not finding a significant relationship include our low statistical power due to the low number of participants, and our lack of inclusion of personal information-such as the presence of work-home conflictsthat may impact the consequence of after-hours work in individual physicians. 2,29 It may also be related to the relatively low amount of EHR time we observed in the after-hours periods compared with prior studies. 11,12 Some of this difference may be attributed to differences in methodologies, as one study used self-report to determine afterhours work, 11 which might be an inaccurate estimate of objectively measured EHR time outside of clinic. 30 Another study normalized EHR time to a full-time physician and included weekday and weekend use in after-hours totals. 12 Nonetheless, physicians in this study had relatively low faceto-face time and patients-per-hour requirements compared with other organizations, which may have resulted in lower after-hours EHR use and limit the generalizability of our findings to settings with higher productivity requirements. Further studies including a larger number of participants from diverse clinical settings, and perhaps examining for threshold effects or looking at maximum (rather than average) after-hours EHR times, are needed to further define the relationship between after-hours EHR work and burnout.
There are several limitations to our study. First, while the response rate to our survey was adequate (59%), our sample size was low and might have limited our power to detect meaningful relationships. While the demographic characteristics and EHR use of survey responders and nonresponders were not significantly different, the possibility that responders were less-or more-burned out than respondents could affect our results and generalizability. Second, we measured burnout with a single-item question, which may have less validity and discriminatory ability compared with the complete 22-item MBI. We chose a single validated measure to encourage a higher response rate among our small group of physicians. Third, our measure of EHR use was not an exact measure of physician's active engagement with EHR, but rather based on user access logs which cannot discriminate between active or idle time. Fourth, EHR measurements intentionally included only time spent when physicians were logged in to their outpatient environment. Our EHR metric was therefore not a reflection of total EHR work, and some EHR work may have been misclassified if physicians performed inpatient work while logged into their outpatient environment, or vice versa. Nonetheless, we believe this effect would have minimal impact on our results, as only eight respondents (24%) did inpatient work. Fifth, our EHR metric did not subdivide the type of EHR work performed, such as work for office visit documentation, order entry, billing, telephone calls, or patient-portal messages. Future studies including EHR task categorizations could help identify which EHR tasks may particularly contribute to burnout. Lastly, our sample was restricted to academic internal medicine physicians in one university-based practice setting, so may not be applicable to other academic or nonacademic practices.

Conclusion
In summary, our results show that burnout among physicians is associated with higher levels of in-clinic EHR use, but not with after-hours EHR use. It is important for future studies to further clarify the relationship between burnout and afterhours EHR use in larger populations and other settings. It is also imperative to better understand the factors involved with high-intensity and intrusive EHR use, particularly during clinic sessions when physicians are required to do multiple computer-related tasks while seeing patients. Improved understanding of the dynamics of in-clinic EHR use, including which specific EHR tasks are particularly related to burnout, may lead to focused interventions at the organizational and individual level, such as improved EHR training, 25,26 the adoption of scribes, 23,24 or the reduction of waste in data entry and documentation processes. [31][32][33] These strategies may ultimately help promote better quality of care and a happier and more stable physician workforce.

Clinical Relevance Statement
This article presents our findings of an observational cohort study using novel ways to objectively characterize primary care physicians' EHR use, to determine the relationship between burnout and their use of the EHR. We found that burnout is primarily related to EHR use during clinic sessions, and not related to after-hours EHR use. We believe our results may help refine efforts to improve physician burnout, by focusing attention on strategies that specifically help integrate EHR demands with in-clinic care.

Protection of Human and Animal Subjects
This study was performed in compliance with the Belmont Report and the Common Rule (45 CFR 46), and was approved by the UW-Madison Health Sciences Institutional Review Board.