Autonomy in work location decision and burnout in behavioral health providers: Lessons learned from COVID-19☆ Authors: H Myra Kim, Peter P Grau, Rebecca K Sripada, Tony Van, Linda Takamine, Jennifer Burgess, Kara Zivin Published: 2023 ABSTRACT AbstractObjective:The ability to choose one’s work-location can influence burnout from employment. We sought to evaluate potential associations between autonomy in work-location decision and burnout.Methods:We used 2020 Veterans Health Administration Annual All Employee Survey data from behavioral health providers at 129 medical facilities. Based on responses to items in a COVID-19 module related to teleworking frequency and reasons for not teleworking both pre- and during COVID-19, we generated six groups representing potentially differing levels of work-location decision autonomy. We hypothesized that greater autonomy is associated with less burnout.Results:Psychologists reported highest burnout (40.1%, 1801/4494), followed by psychiatrists (35.6%, 703/1976) and social workers (31.0%, 3421/11,051). Adjusting for covariates including workload, providers newly approved for telework during COVID-19 showed 1.28 (95% CI=1.03–1.59), 1.71 (1.25–2.35), and 1.35 (0.86–2.12) times higher odds of burnout among social workers, psychologists, and psychiatrists, respectively, compared to providers chose not to telework during COVID-19.Conclusions:Providers who chose to keep coming into work in person reported less burnout than providers who previously could not telework and are teleworking during the pandemic. Greater autonomy in work-location decision might have greater importance in reducing burnout in behavioral health providers than having the ability to telework per se.Keywords: Autonomy in work-location decision, Tele-working, Burnout, Behavioral health providers Introduction During the COVID-19 pandemic, teleworking emerged as an effective way to keep people safe and productive by avoiding close contacts. With availability of networking tools, teleworking or flexible work arrangements that allow people to work from any location other than the traditional office became more common (Gartner, 2020). Well before the COVID-19 pandemic, research showed the positive impact of teleworking on productivity and well-being, though studies also noted potential limitations such as lack of social engagement and managerial access or support (Allen et al., 2015; Bosua et al., 2017). Depending on the type of job, working from home can provide more freedom, but data on whether telework increases or decreases stress or job productivity have not been consistent (Kniffin et al., 2021). The COVID-19 pandemic greatly increased amount of tele-mental health care provided nationally as well as within the Department of Veterans Affairs (VA), the nation’s largest health care system (Pierce et al., 2021). Psychologists working in VA medical centers reported a seven-fold increase in telepsychology use, which had a higher likelihood of occurring among providers who received dedicated telepsychology training and reported supportive organizational telepsychology policies (Pierce et al., 2021). In the VA, tele-mental health visits via both videoconferencing and telephone increased by over 400% during the first 6 weeks of the pandemic, while the in-person visits dropped by 81% (Connolly et al., 2020). Although the pandemic made teleworking opportunities more available, whether telework reduces or increases employee stress remains unknown. For example, a survey study of 209 Italians working at home after COVID-19 found self-leadership and job autonomy to have a positive association with job productivity, but not with stress of working from home (Galanti et al., 2021). Another study of 1285 teleworking Latin Americans found that remote work in pandemic times increased perceived stress, reduced work–life balance and work satisfaction, and increased productivity and engagement (Sandoval-Reyes et al., 2021). The latter study only covered the first two months period after the countries’ governments implemented measures to control COVID-19 (Sandoval-Reyes et al., 2021) and thus the increased perceived stress may have resulted from haphazard work coordination or other processes besides remote working. Teleworking can serve as an efficient means of improving productivity and reducing burnout for employees who commute long hours or need to devote greater amounts of time to domestic responsibilities (Kazekami, 2020). The ability to choose one’s work-location, including remote work through telework, can affect work satisfaction, job productivity, and burnout from work. Burnout rates may decline if employees can make autonomous work-location decisions with personal considerations for factors that may influence their individual work-related stress. Wider availability and increasing necessity of teleworking during the pandemic provided an opportunity to assess the relationship not only between teleworking and burnout, but between autonomy in the choice to engage in telework and work-related burnout in a large healthcare system. In the VA, pre-existing telehealth infrastructure facilitated rapid growth in tele-mental health visits (Connolly et al., 2020, 2022). Behavioral health providers (BHPs: psychiatrists, psychologists, or social workers) provide mental health treatments including counseling, group therapy, and medication. The VA also annually administers All Employee Survey (AES), and based on the AES, substantial rate of burnout was reported (31% to 38%) in the BHPs during 2015–2018 (Department of Veterans Affairs, 2020; Zivin et al., 2022). In this study, we used responses to the 2020 AES from BHPs in the VA medical facilities to assess if employee level burnout differed by not only their current teleworking status, but the level of autonomy in their work-location decisions. We considered the work-location decisions autonomous if the decision appeared self-determined, and non-autonomous if the decision appeared directed and hypothesized that burnout rate differs by the differing levels of autonomy in work-location decision. Methods Data sources The National Center for Organizational Development administers the AES to all Veterans Health Administration (VHA) employees as an annual organizational census of workplace perceptions and satisfaction (Department of Veterans Affairs, 2020; Rinne et al., 2020). AES data provide organizational feedback and lead to workplace improvement. In this study, we focused on BHPs, specifically respondents who self-identified as psychiatrists, psychologists, or social workers in their AES responses. In addition to questions on work-related burnout, the 2020 AES, administered in June of 2020, included COVID-19 related questions that asked how often the employee teleworked in an average week separately for both before the COVID-19 pandemic and during the COVID-19 pandemic. Measures Dependent variable: burnout We classified whether respondents indicated experiencing burnout or not according to methods used previously (Zivin et al., 2022; West et al., 2009). This approach used two burnout questions in the AES: emotional exhaustion (“I feel burned out from my work”) and depersonalization (“I worry that this job is hardening me emotionally”), each with a 7-point response (1=never; 2=A few times a year or less; 3=Once a month or less; 4=A few times a month; 5=Once a week; 6=A few times a week; 7=Every day). We generated a dichotomous variable of endorsing burnout based on a response of 5 or higher on either of the two questions. A prior study using the AES reported results from testing validity of the burnout variable cutoff by comparing it to another AES variable of turnover plans (“I plan to leave my job within the next six months”) which was highly correlated with burnout and the threshold value of ≥5 had the highest sensitivity among all thresholds, with a false positive rate <30% (Zivin et al., 2022). The two burnout questions were highly collinear with a correlation coefficient of 0.76, and we also used the average of the responses to the two questions as an alternate continuous measure of burnout. Autonomy in work-location decision (AWLD) We empirically defined six mutually exclusive AWLD groups related to current work status based on responses to teleworking status and to reasons for not teleworking, separately asked for prior to and during the COVID-19 pandemic. Teleworking status question was “How often did you telework before the COVID-19 pandemic in an average week?” and “How often have you teleworked during the COVID-19 pandemic in an average week?” For each question, if employees responded as “never having teleworked” during the period, the next question asked whether they did not telework because they “chose not to”, “could not perform the job duties from home”, or “had not been approved to telework even though the job duties could be performed from home”. Table 1 shows the mutually exclusive six AWLD groups empirically defined, considering together the observed responses to both current and prior teleworking status and self-reported reasons for not teleworking. In assessing the relationship between autonomy and burnout using the 2020 AES survey data, we sought to examine the differences in burnout between groups with clear autonomy or lack of autonomy during COVID-19. For example, those providers who chose not to telework during COVID-19 (Group A) clearly had autonomy; while those groups who could not or did not have approval to work remotely during COVID-19 (Groups B and C) clearly had low autonomy. However, it remains unclear whether those providers who teleworked during COVID-19 (Groups D–F) did this by choice or due to employer requirement. Table 1Autonomy in work-location decision (AWLD) groups based on teleworking status and reasons for not teleworking in pre COVID-19 and during COVID-19.AWLD GroupNTeleworking status Before COVID-19During COVID-19Autonomy Level1A1235Chose not to, could not, or not approvedChose not to“High”B3042Could not or not approvedCould not or not approved“Low”C122Chose not toCould not or not approved“Low”D4021TeleworkedTelework“Moderate to High”E1197Chose not toTelework“Moderate to High”F7904Could not or not approvedTelework“Low”1The levels of autonomy in work location decision designated informally to each group. Autonomy in work-location decision (AWLD) groups based on teleworking status and reasons for not teleworking in pre COVID-19 and during COVID-19. The levels of autonomy in work location decision designated informally to each group. We used data from pre-COVID to identify potentially different levels of autonomy within this large group. That is, we assumed that those who tele-worked prior to COVID-19 did so by choice. It is unlikely that employers required anyone to telework prior to COVID-19, and those who “chose not to” telework prior to COVID-19 did so by choice. Hence, we considered providers in these two groups (Groups D and E) to have moderate to high levels of autonomy (not knowing their autonomous status during COVID-19). In contrast, we considered those who could not or did not have approval to telework pre-COVID to have low autonomy (Group F). In this way, we did not seek to designate and validate each AWLD group with specific levels of autonomy. Rather, we viewed the six groups to represent differing levels of autonomy with the current work-location decision. Column in Table 1 designates each group as “high”, “moderate”, or “low” autonomy. We note that although three groups are designated as “low” and two groups as “moderate”, the underlying level of autonomy in these groups are not similar enough to be combined in terms of their level of autonomy. For example, both Groups B and C were those who could not or were not approved to telework during COVID-19 and we considered both groups to have “low” autonomy. However, we did not think the group that could not or were not approved to telework pre-COVD (Group B) had the same low level of autonomy as those who chose not to telework pre-COVID (Group C). In defining the six AWLD groups, we considered any amount of teleworking as telework. Of note, 23.0% of the providers reported teleworking pre-pandemic and 74.9% during pandemic. Of those who teleworked during the pandemic, 9% teleworked for <1 day per week, 17.4% for 1–2 days, 32.8% for 3–4 days, and 40.8% for 5 days. We do not have data on the availability of facility teleworking options, but 95% (135/141) of facilities had providers in 5 or 6 different AWLD groups, indicating that most facilities had teleworking providers with differing AWLD levels. Covariates: demographic and other provider characteristics We used the socio-demographic characteristics data collected in the AES as covariates based on their potential influence on burnout and potential differences in distributions across the AWLD categories. They included sex, ethnicity, race, VHA tenure, level of supervisory responsibility, ever served as active duty, disabled, assigned to Patient Aligned Care Team (PACT; the VHA’s implementation of the Patient Centered Medical Home model for primary care), a supervisor of health professionals (yes, no), and faculty appointment (yes, no). Prior work showed reasonable workload as highly predictive of burnout among work environment characteristics (Zivin et al., 2022). We therefore adjusted for employee expressed level of workload according to a 5-point response to “My workload is reasonable” (1=“strongly disagree” and 5=“strongly agree”). Statistical analysis Since the relationship between AWLD groups and burnout can depend on occupation types (psychiatrists, psychologists, or social workers) with their differing roles and socio-demographic characteristics, we stratified all analyses by occupation types. We summarized the provider characteristics and calculated the percentage of BHPs who endorsed burnout in the six AWLD groups within each occupation type. To assess the relationship between AWLD and burnout, we used mixed effects logistic regression models with burnout as dependent variable, AWLD groups as the primary predictor of interest, and random intercepts for facilities to account for potential within-facility correlation of burnout. We used AWLD Group A that we considered to have “high” autonomy as the referent group and included in the model indicators for each of the five other AWLD groups. We included in the model provider characteristics potentially predictive of burnout but did not include age as it is highly correlated with VHA tenure. The parameter estimates of the AWLD group indicators provided the adjusted odds ratios (AOR) of burnout associated with the different AWLD groups relative to Group A. We did two additional analyses to assess robustness of our findings. We repeated the analysis of binary burnout adjusting for all covariates except workload to see if perceived workload changed the magnitude or the direction of the association between AWLD groups and burnout. We also repeated all analyses with continuous burnout scores with possible ranges from 1 to 7 as the dependent variable using linear mixed models. All analyses were done using Stata 16.0 (College Station, TX). Results Year 2020 AES provided data in 18,293 BHPs, and of the respondents, 17,599 (96.2%) providers had both burnout and AWLD exposure data. Respondents were from 141 VA facilities and included 72.4% female and 81.4% White. Social workers represented the largest occupation type (63.1%), followed by psychologists (25.6%), and psychiatrists (11.3%). By occupation type, 78.2% of social workers identified themselves as female, followed by 68.6% of psychologists and 48.6% of psychiatrists. Burnout rate was highest in psychologists at 40.1%, followed by 35.6% in psychiatrists, and 31.0% in social workers (Table 2). Table 2Characteristics of self-identified behavioral health providers in Veterans Health Administration 2020 All Employee Survey.Social WorkersPsychologistsPsychiatristsTotalCharacteristicsN=11,098 (63.1%)1N=4514 (25.6%)1N=1987 (11.3%)117,599Burnout3441 (31.0%)1809 (40.1%)708 (35.6%)5958 (33.9%)Male2358 (21.8%)1381 (31.4%)994 (51.4%)4733 (27.6%)Female8447 (78.2%)3021 (68.6%)941 (48.6%)12,409 (72.4%)Age in years <2688 (0.8%)3 (0.1%)2 (0.1%)93 (0.5%) 26–29467 (4.3%)103 (2.4%)6 (0.3%)576 (3.4%) 30–393248 (30.2%)1737 (39.8%)426 (22.1%)5411 (31.8%) 40–493464 (32.2%)1369 (31.4%)519 (26.9%)5352 (31.4%) 50–592331 (21.7%)706 (16.2%)504 (26.1%)3541 (20.8%) 60 years or older1150 (10.7%)444 (10.2%)471 (24.4%)2065 (12.1%)Hispanic964 (9.5%)324 (7.7%)166 (9.4%)1454 (9.0%)Race White / Caucasian8284 (79.1%)3871 (90.4%)1377 (73.1%)13,532 (81.4%) Black or African American1775 (17.0%)222 (5.2%)95 (5.0%)2092 (12.6%) Asian260 (2.5%)159 (3.7%)394 (20.9%)813 (4.9%) Other2150 (1.4%)29 (0.7%)18 (1.0%)197 (1.2%)Veteran Health Administration tenure <6 months521 (4.8%)166 (3.8%)113 (5.8%)800 (4.7%) 6 months-1 year524 (4.8%)116 (2.6%)68 (3.5%)708 (4.1%) 1–2 years1066 (9.8%)331 (7.5%)168 (8.7%)1565 (9.1%) 2–5 years2164 (19.9%)994 (22.5%)430 (22.2%)3588 (20.9%) 5–10 years3312 (30.5%)1274 (28.9%)454 (23.4%)5040 (29.3%) 10–20 years2777 (25.6%)1273 (28.9%)487 (25.1%)4537 (26.4%) >20 years487 (4.5%)254 (5.8%)219 (11.3%)960 (5.6%)Supervisory role None8397 (76.7%)2943 (66.3%)1137 (58.5%)12,477 (72.0%) Team Leader1160 (10.6%)775 (17.5%)382 (19.7%)2317 (13.4%) First Line Supervisor884 (8.1%)388 (8.7%)241 (12.4%)1513 (8.7%) Manager/Executive507 (4.6%)335 (7.5%)183 (9.4%)1025 (5.9%)Prior active duty1742 (15.9%)312 (7.0%)196 (10.1%)2250 (13.0%)Disabled1660 (15.5%)373 (8.5%)118 (6.2%)2151 (12.7%)Patient Aligned Care Team member1476 (13.5%)484 (10.8%)110 (5.6%)2070 (11.9%)Supervise health professions trainees2535 (23.3%)2988 (67.4%)1403 (72.1%)6926 (40.1%)Faculty appointment303 (2.8%)1147 (25.7%)1187 (60.8%)2637 (15.2%)1Some characteristics are missing in 2020 AES respondents; Percentages are calculated among those with non-missing data for each characteristic. In the overall sample, highest missing values are seen for ethnicity (Hispanic vs. not) with 8.4% (N=1486) missingness.2Includes American Indian or Alaskan Native, Native Hawaiian or other Pacific Islander. Characteristics of self-identified behavioral health providers in Veterans Health Administration 2020 All Employee Survey. Some characteristics are missing in 2020 AES respondents; Percentages are calculated among those with non-missing data for each characteristic. In the overall sample, highest missing values are seen for ethnicity (Hispanic vs. not) with 8.4% (N=1486) missingness. Includes American Indian or Alaskan Native, Native Hawaiian or other Pacific Islander. The final analytic cohort included 17,521 respondents after excluding 78 respondents with unusual responses, those who teleworked before the pandemic, but during the pandemic, either chose not to telework (N=24) or could not or were not approved to telework (N=54). To ensure that the exclusions did not come from a specific occupation type or facilities, we checked their distributions; the excluded respondents included 47 social workers, 20 psychologists, and 11 psychiatrists and were from 55 different facilities. Burnout rates by occupation type in each of the six AWLD groups appear in Fig. 1 and Supplementary Table 1. Prior to the pandemic, only 23% (n=4021, Group D) teleworked with their reported current year burnout rate of 32.4%, while 74.9% (n=13,122, Groups D–F) teleworked during COVID with their current year burnout rate of 34.6%. The burnout rate in the combined Groups D–F was higher because it was higher in Group F. More importantly, burnout rates differed greatly by AWLD groups across and within each occupation type. Among social workers, burnout rates ranged from 24% (Group A and E) to 33% (Group F). Psychologists also showed burnout rate ranging from 29% (Group A) to 47% (Group C); 36% or higher endorsed burnout in all groups except Group A. Psychiatrists had the smallest sample size, but also showed a much higher rate of burnout in Group F compared with other groups. Of note, within each occupation type, across those who do not telework during the pandemic (Groups A–C), and across those who telework during the pandemic (Groups D–F), the percentage of respondents in each group endorsing burnout differed widely. Percentage endorsing burnout in each autonomy in work-location decision Groups (AWLD) by occupation type and across all occupation types. AWLD Group definitions are in Table 1. Briefly, Groups A, B, and C do not telework during the pandemic, and Groups D–F telework during the pandemic. Adjusting for covariates, AWLD groups remained significantly predictive of burnout among social workers (p=0.004) and psychologists (p=0.003), but not among psychiatrists (p=0.15, Table 3). Among social workers, compared with Group A that we considered as “high” in autonomy, significantly higher odds of burnout was seen in Group B (AOR=1.40; p=0.01) and Group F (AOR=1.28; p=0.03). Among psychologists, burnout was significantly more likely in nearly all AWLD groups when compared to Group A. Among psychiatrists, although AWLD groups were not significantly predictive of burnout, the AOR associated with Group F was 1.35 (p=0.19), suggesting higher burnout for Group F than Group A. Analysis of burnout adjusting for all provider characteristic variables except for perceived workload showed the association between AWLD groups and burnout to be not affected by whether workload was adjusted for or not (Supplementary Table 2). Analyses using continuous burnout scores also gave substantively identical findings to significant associations seen with dichotomized burnout in terms of relative magnitudes and directions of associations (Supplementary Table 2). Table 3Adjusted odds ratio (AOR) of burnout associated with autonomy in teleworking decision, stratified by occupation, each based a mixed effects logistic regression model.Social WorkerN=9327PsychologistsN=3877PsychiatristsN=1589 CharacteristicsAOR95% CIAOR95% CIAOR95% CIAWLD: Group A1.01.01.0 Group B 1.40 * (1.11–1.77) 1.81 ^ (1.26–2.60)1.02(0.62–1.67) Group C1.18(0.58–2.39) 2.72 * (1.18–6.27)0.61(0.17–2.17) Group D1.14(0.91–1.44)1.37(0.98–1.93)0.98(0.62–1.56) Group E0.93(0.68–1.27) 1.58 * (1.08–2.32)0.87(0.52–1.46) Group F 1.28 * (1.03–1.59) 1.71 ^ (1.25–2.35)1.35(0.86–2.12)Male0.91(0.80–1.03)0.84(0.71–0.99)0.87(0.69–1.11)Hispanic0.99(0.83–1.18)0.95(0.71–1.28)0.94(0.62–1.44)Race: White1.01.01.0 Black 0.71 ^ (0.61–0.82)0.87(0.62–1.21) 0.42 * (0.23–0.77) Asian1.02(0.74–1.41)1.07(0.72–1.59) 0.64 * (0.47–0.87) Other1.00(0.66–1.52)1.84(0.78–4.33)0.66(0.16–2.63)Tenure: <6 months1.01.01.0 6 months-1 year 1.80 ^ (1.20–2.68)1.85(0.95–3.64)1.36(0.60–3.09) 1–2 years 2.41 ^ (1.69–3.44) 2.26 ^ (1.29–3.96)1.10(0.56–2.18) 2–5 years 3.09 ^ (2.21–4.32) 3.37 ^ (2.01–5.62)1.46(0.80–2.65) 5–10 years 3.31 ^ (2.37–4.61) 3.64 ^ (2.19–6.08) 2.11 * (1.16–3.84) 10–20 years 3.12 ^ (2.23–4.36) 2.73 ^ (1.63–4.56)1.63(0.89–2.98) >20 years 2.24 ^ (1.50–3.37) 2.09 * (1.16–3.77)1.41(0.73–2.72)Supervisory: None1.01.01.0 Team Leader1.02(0.87–1.20)0.99(0.81–1.21)1.07(0.78–1.46) Supervisor0.97(0.81–1.18)0.97(0.74–1.26)1.12(0.77–1.64) Manager/Executive0.93(0.72–1.18)1.31(0.98–1.76)0.87(0.56–1.34)Prior active duty0.98(0.84–1.15)0.83(0.61–1.13)0.88(0.58–1.31)Disabled 1.55 ^ (1.34–1.80) 1.43 * (1.09–1.88)1.20(0.73–1.96)PACT member1.00(0.86–1.16)0.82(0.64–1.04)0.90(0.54–1.52)Supervise HT1.02(0.90–1.15)0.97(0.82–1.15)0.84(0.61–1.16)Faculty appointment1.07(0.79–1.44)0.90(0.75–1.08)1.03(0.77–1.38)Reasonable workload: 11.01.01.0 2 0.35 ^ (0.28–0.43) 0.38 ^ (0.27–0.54)0.62(0.33–1.18) 3 0.17 ^ (0.14–0.22) 0.24 ^ (0.17–0.34) 0.28 ^ (0.15–0.51) 4 0.08 ^ (0.06–0.09) 0.08 ^ (0.06–0.11) 0.11 ^ (0.06–0.20) 5 0.03 ^ (0.03–0.04) 0.03 ^ (0.02–0.05) 0.04 ^ (0.02–0.08)AWLD group effect§p-valueχ2(df=5)p-valueχ2(df=5)p-valueχ2(df=5).00417.0.00318.0.158.1Abbreviation: AWLD is autonomy in work-location decision; HT is health trainees; AA is African American; PACT is Patient Aligned Care Team; AOR is adjusted odds ratio; CI is confidence interval.*p<0.05.^p<0.01.§Likelihood ratio test for differences in burnout rates across 6 AWLD groups.(AWLD Group A Before/During) Chose not to, could not, or not approved / Chose not to(AWLD Group B Before/During) Could not or not approved / Could not or not approved(AWLD Group C Before/During) Chose not to / Could not or not approved(AWLD Group D Before/During) Teleworked / Telework(AWLD Group E Before/During) Chose not to / Telework(AWLD Group F Before/During) Could not or not approved / Telework. Adjusted odds ratio (AOR) of burnout associated with autonomy in teleworking decision, stratified by occupation, each based a mixed effects logistic regression model. Abbreviation: AWLD is autonomy in work-location decision; HT is health trainees; AA is African American; PACT is Patient Aligned Care Team; AOR is adjusted odds ratio; CI is confidence interval. p<0.05. p<0.01. Likelihood ratio test for differences in burnout rates across 6 AWLD groups. (AWLD Group A Before/During) Chose not to, could not, or not approved / Chose not to (AWLD Group B Before/During) Could not or not approved / Could not or not approved (AWLD Group C Before/During) Chose not to / Could not or not approved (AWLD Group D Before/During) Teleworked / Telework (AWLD Group E Before/During) Chose not to / Telework (AWLD Group F Before/During) Could not or not approved / Telework. Discussion In this study, we evaluated a relationship between autonomy in decisional involvement on work-location and work-related burnout using an annual survey administered to all VHA employees. In March 2020, work environment changed unexpectedly due to the COVID-19 pandemic and provided an opportunity to explore if autonomous decisional involvement in where to work is associated with employee work-related burnout. Using responses to two pandemic-related questions added to the survey in year 2020, we grouped providers into six groups representing potentially differing levels of autonomy in work-location decisions and explored if burnout among behavioral health providers was associated with the groups. First, we found high percentage of behavioral health providers endorsing burnout, and the percentage differed greatly across the three occupation types; approximately 31% of social workers, 36% of psychiatrists, and 40% of psychologists experienced burnout. We also found that burnout rate did not vary by their current teleworking status, and it instead varied significantly across AWLD groups within those who telework during the pandemic (Groups D–F) and within those who do not telework during the pandemic (Groups A–C). Our data also showed that across occupation types, BHPs who chose to keep coming into work in person (Group A) reported less burnout than those who previously were not able to or not approved to telework and now have the flexibility to telework (Group F). This finding supports the notion that it may not be telework, but rather the autonomy in the work-location decision that predicts burnout. We found several provider characteristics predicted burnout, and in particular, burnout was less likely with increasing level of reasonable workload, consistent with a prior study based on the AES (Zivin et al., 2022). Even after adjusting for provider characteristics, the odds ratios of burnout associated with the AWLD groups, when compared with the Group A, considered to have more autonomy, ranged from 1.37 to 2.72 among psychologists and from 0.93 to 1.40 among social workers. Among psychiatrists, potentially due to a smaller sample size, we did not find burnout rates to significantly vary across the six AWLD groups after adjusting for covariates, but a relatively higher level of burnout was seen in Group F with lack of autonomy than in Group A with more autonomy. Previous studies have generally shown teleworking to have benefits of increased job satisfaction and productivity (Troup and Rose, 2012; Maruyama and Tietze, 2012). These studies have taken place among those who are already teleworking or had teleworking as an available option. In this study, we explored if the differing levels in autonomy in work-location decision was associated with burnout in BHPs. We have no data directly obtained from providers regarding autonomy in their work-location decisions. Nonetheless, we found significant variation in burnout across the different groups. Even across the three groups of teleworking providers, their burnout levels differed by the implied autonomy in the decisional involvement in their prior work-location. Autonomy is a key driver of human motivation, performance, and fulfillment (Reisinger and Fetterer, 2021). Enabling and empowering employee autonomy is said to be the single most important enabler of flexibility, and a culture of autonomy and flexibility can benefit not only individuals but their teams and organizations (Reisinger and Fetterer, 2021). In health care settings, a culture of autonomy can benefit providers, teams, and health care systems and can therefore benefit patients. Autonomy can encompass many different aspects, and in this study, we focused on autonomy in work-location decision and its association with provider burnout. Our study confirms that VA BHPs not only commonly experienced burnout during the first year of the COVID-19 pandemic, but their burnout levels showed sizable differences across the six empirically defined AWLD groups based on their teleworking status and reasons for not teleworking. Autonomous decision-making regarding work-location allows providers to make a decision that considers one’s own preferred method of managing domestic needs, achieving work-life balance, and addressing health concerns, and this helps alleviate burnout. A similar pattern of associations occurred across different occupation types despite their potentially differing roles and responsibilities, which strengthens the potential importance of improving autonomy in work-location decision and in crafting one’s own work location in reducing work related burnout. Study limitations First, the AES did not ask direct and specific questions about work-location decisional involvement. We instead extracted the underlying decisional involvement data for current work status from responses to working status both prior to and during the pandemic, particularly in the case of those who teleworked during the pandemic. In this case we could not necessarily assume teleworking during the pandemic was by choice; so, we relied on the pre-COVID responses to categorize level of autonomy, recognizing that pre-COVID responses may show differing levels of autonomy in work location decision and differing association with burnout in 2020. The associations seen between differing burnout levels and differing AWLD in our data suggest that it is worth considering improved autonomy in work-location decision. Second, COVID-19 may have put unique strains and stressors on health systems and providers, so findings may not represent relationship in non-pandemic times. Third, within and across each AWLD group, we expected variations in family conflicts or level of social isolation to explain partial variation in burnout, but these were not measured. Lastly, our data does not allow us to speculate on how role and autonomy impact burnout and why psychiatrists do not experience the same pattern of burnout as the other behavioral health occupations. Conclusions Given the challenges in conducting a randomized controlled trial to determine if decisional involvement in work-location affects employee outcomes including burnout, data from a natural experiment that influences a change in decision provides an alternative to assess the role of AWLD on employee (provider) experiences. This study and the pandemic provided an opportunity to assess outcomes when many providers had the opportunity to telework due to the pandemic. Lack of AWLD appears to have an association with increased burnout in BHPs, particularly in psychologists and social workers. Provider burnout may improve if they can have greater involvement in deciding their own work-location. Supplementary Material Acknowledgments Funding information. This work was supported in part by the Health Services Research and Development, Grant/Award Numbers: VA IIR 17–262, VA. The funding source did not have any role in conception, design, writing or editing of the paper. Footnotes Previous presentation: The paper has been presented at the VA HSR&D 2023 National Meeting. Declaration of generative AI in scientific writing None. The authors take full responsibility for the content of the publication. CRediT authorship contribution statement H. Myra Kim: Conceptualization, Formal analysis, Writing – original draft, Writing – review & editing. Peter P. Grau: Conceptualization, Writing – review & editing. Rebecca K. Sripada: Conceptualization, Writing – review & editing. Tony Van: Data curation, Formal analysis. Linda Takamine: Conceptualization, Writing – review & editing. Jennifer Burgess: Project administration, Writing – review & editing. Kara Zivin: Conceptualization, Funding acquisition, Writing – review & editing. Declaration of Competing Interest None. Supplementary materials Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.jadr.2023.100652. References Associated Data Supplementary Materials