Well-Being of Health Care Professionals Treating Opioid Use Disorder During the COVID-19 Pandemic: Results From a National Survey

The COVID-19 pandemic has had devastating impacts on global economies and health care (1–3). Less is known about impacts on the well-being of health care professionals (HCPs). Given historically high rates of burnout among HCPs (4, 5), the COVID-19 pandemic has likely affected HCP well-being. For example, among HCPs who treat patients with opioid use disorder, increasing demand of patients for mental health treatment and rising overdoses in the wake of COVID-19 may have contributed to negative impacts on well-being (6).

Before the COVID-19 pandemic, U.S. physicians, nurse practitioners, and physician assistants reported high rates of burnout (30%–50%) (7). Among physicians, 42% reported burnout immediately before the COVID-19 pandemic, with higher rates among women and 40- to 54-year-olds (4). Compared with the general population, physicians have increased risk for burnout and less likelihood of satisfaction with work-life balance, even after analyses are adjusted for age, sex, relationship status, and hours worked per week (8). Even before COVID-19, physicians who regularly prescribed buprenorphine reported less satisfaction with their role of treating opioid use disorder than with their role in general medical practice (9). Surveys comparing burnout among urban versus rural physicians have not shown significant differences in burnout between these two groups (10, 11). No surveys have compared measures of well-being during a public health emergency that has specifically affected major urban centers. Data from frontline physician trainees showed significantly higher stress and burnout among trainees exposed to COVID-19 (12), but less is known about changes to well-being for those providing remote patient care.

The COVID-19 public health emergency triggered rapid modifications to health care service delivery, particularly regarding telehealth regulations at the federal and state levels (13–15). For treatment of opioid use disorder, the Drug Enforcement Administration and the Substance Abuse and Mental Health Services Administration waived requirements for initial in-person visits for buprenorphine and allowed reduced in-person visits for patients receiving methadone (16). Implementation and expansion of telehealth for psychiatric services have been systematically studied and were found to be clinically and economically effective (17, 18). Less is known about telemedicine’s effectiveness for opioid use disorder treatment (19, 20), and studies have not systematically evaluated the impacts of telemedicine on HCP well-being. Various models of telehealth services exist; the patient and HCP may be located at their respective residences, elsewhere in the community, or both in a clinic. Variations in telehealth delivery have not been systematically studied, and specific impacts of these variations on HCP well-being are likely.

To better understand the impact of the COVID-19 pandemic on HCPs who treat patients with opioid use disorder, a consortium of stakeholder organizations (see an online supplement to this article) was rapidly identified, spearheaded by leadership of the American Academy of Addiction Psychiatry (AAAP). The consortium created the Opioid Use Disorder Provider COVID-19 Survey (see the online supplement) to explore practice changes, perspectives, and other impacts of the pandemic. To investigate impacts on the well-being of HCPs who treat patients with opioid use disorder, we explored associations among demographic variables, provider types, and practice settings, and we reported practice changes. Although we anticipated associations similar to those that have been previously described (i.e., differential impacts on well-being by gender and age), we also anticipated that demographic area would be associated with effects on HCP well-being, given more abrupt impacts of the pandemic on more densely populated U.S. cities between mid-March and mid-May (21).

Methods

Sample

Consortium members, including clinicians, educators, and policy experts, created the Opioid Use Disorder Provider COVID-19 Survey. This anonymous survey contained 31 items and was administered through the Qualtrics online survey platform. The survey included questions regarding HCP role and demographic characteristics, prepandemic clinical practices and postpandemic changes, HCP perception of these changes, and HCP well-being. It included a combination of multiple choice and open-ended questions. Questions were derived from a literature review, including a survey on behavioral health provider well-being (22). A smaller group of subject matter experts from the consortium member organizations reviewed and edited the survey (23, 24). Subject matter experts conducted two reviews of survey items and made revisions over three rounds of edits. The Yale University Institutional Review Board exempted the survey from review.

The total subscribership of all listservs was approximately 157,000 (for details, see online supplement). Among these organizations, membership type varied, including primarily nonaddiction specialist physicians (e.g., American Medical Association) and a combination of addiction specialist physicians with other health professionals, including advanced practice nurses, physician assistants, psychologists, social workers, and counselors (e.g., American Society of Addiction Medicine [ASAM], AAAP, and Addiction Technology Transfer Center [ATTC]). Among AAAP listserv subscribers, approximately 40% were prescribers (physicians or advanced practice providers). A similar breakdown by prescriber or nonprescriber was not available for ASAM or ATTC listservs. Memberships of these organizations and subscriptions to their relevant listservs were not mutually exclusive and may have had significant overlap. We could not discern the number of listserv subscribers who were active prescribers of medications for opioid use disorder. Survey responses were received from July 14 to August 15, 2020.

Dependent Variables

Dependent variables measured HCP well-being as functioning and work-life balance. We used two responses to the following survey question: “Please indicate any changes to staffing and provider well-being you have experienced as a result of COVID-19.” Functioning was captured with the response option “My anxiety level about COVID-19 has impacted my functioning at home and/or work,” and work-life balance with the response option “I am having a more difficult time than usual balancing work and home life.” For each question, participants could select all options that applied. A response option was available for recording “no changes,” which helped to distinguish nonresponders (i.e., missing data) from people who did not experience any changes. Respondents who checked a box for a response option were coded 1 for the response. Respondents who checked at least one box for the question, but left a response option blank, were coded 0 for that particular option. Respondents who left all options blank were coded as missing data.

Independent Variables

Seven response options assessed organizational practices and related HCP experiences, which came from responses to the same survey question as dependent variables and were coded in the same manner.

Covariates

Covariates included respondent self-reported race (White, Black, or Asian), ethnicity (Hispanic or not), gender, age, demographic area (rural, suburban, urban, or other), board certification in addiction, practice setting, and whether most of their patients with opioid use disorder had Medicaid as their insurance type. Options for gender included “man,” “prefer not to say,” “prefer to self-describe,” and “woman.” Because of the small number of responses in the “prefer not to say” and “prefer to self-describe” categories (15 people across both categories), respondents were coded 1 if they selected “man” and 0 if they selected any other category. Survey response options for age in years were categorical (<30, 30–39, 40–49, 50–59, 60–69, ≥70) and treated as continuous in analysis. Board certification in addiction was reported as yes or no. Practice setting included 10 options and “other”; multiple responses were permitted.

To simplify analysis and interpretation, we used “multiple settings” as one category and collapsed categories with a small number of responses (Veterans Health Administration, Indian Health Service, emergency setting, prison or jail, and other) into a single “other only” category. Only the top three categories (multiple settings, private practice only, and opioid treatment program only) were included in the model. State was reported and included options for all 50 U.S. states, Puerto Rico, and “I do not reside in the U.S.” Provider type was dichotomized as physician or advanced practice provider (nurse practitioner or physician assistant) and described; however, it was not included in the model because it had a larger missing portion (12%, N=94 of 812), a small proportion of advanced practice providers, and collinearity between this variable and other covariates, particularly gender.

Statistical Analysis

Descriptive statistics (frequencies and percentages) were used across the following variables: race, ethnicity, gender, age, demographic area, board certification, prescriber type, practice setting, Medicaid acceptance, and organizational practices and experiences. We then examined relationships between HCP well-being (described earlier in the Dependent Variables section) and organizational practices and related HCP experiences by using multilevel multivariate logistic regression models, with random effects at the state level. We used likelihood-ratio tests to compare multilevel models in addition to ordinary logistic regression. Models were adjusted for previously described covariates. Stata (version 16) was used for analysis (25).

Results

Between July 14 and August 15, 2020, a total of 1,109 individuals answered the first survey question, 832 completed the entire survey, and 812 completed the survey item relevant to the current analysis. The response rate could not be calculated because of the overlap among organizations’ e-mail listservs and because the number of listserv recipients who were active prescribers of medications for opioid use disorder was unknown. The rate of missing data among variables included in regression models was <1%. The survey respondents came from all 50 states except South Dakota. Eight respondents were from outside the United States (<1% of the sample), and one was from Puerto Rico. The top five states by number of survey participants were Colorado (N=108), California (N=50), New York (N=49), Massachusetts (N=43), and New Mexico (N=35). Of the remaining states, 22 had fewer than 10 respondents.

Among the respondents, most identified their race as White (80%, N=650 of 809), gender as male (53%, N=430 of 810), and profession as physician (75%, N=536 of 718). Participant age was equally distributed across categories, except for those ages <30 years representing 1% (N=12 of 810) and those ≥70 years representing 11% (N=92 of 810) of the survey sample. Nearly half (46%, N=376 of 812) reported working in an urban area, and the remaining respondents were split between suburban (28%, N=228 of 812) and rural (23%, N=186 of 812) areas. Only slightly more than one-third of participants (38%, N=309 of 812) were board certified in addiction. More than half (57%, N=464 of 811) reported Medicaid as their patients’ primary insurance. Practice settings were diverse, with 24% (N=192 of 812) reporting multiple settings, 17% (N=136 of 812) private practice only, and 11% (N=92 of 812) opioid treatment program only. Only 16% (N=131 of 812) reported that all staff were working onsite as usual. Overall, 17% (N=136 of 812) of respondents reported that COVID-19 had caused function-impairing anxiety at work and home, and 28% (N=229 of 812) reported that COVID-19 had disrupted their work-life balance (Table 1).

Bivariate Pearson’s correlation tests indicated that about half of pairwise comparisons of organizational characteristics were statistically significant, although these correlations were mostly weak (r=−0.46 to 0.30) (see online supplement). Two factors were associated with lower odds of reporting functional impairment because of anxiety about COVID-19: male gender (odds ratio [OR]=0.60, p=0.01) and having no staff changes at work (OR=0.51, p=0.01). Two factors were associated with higher odds of reporting functional impairment because of anxiety about COVID-19: staff sick with COVID-19 (OR=2.59, p<0.01) and feeling close to patients (OR=1.97, p=0.01) (Table 2).

Male gender was associated with lower odds of having difficulty balancing work and home life (OR=0.56, p<0.01), as was being older (OR=0.76, p<0.01) and reporting no staff changes at their organization (OR=0.54, p=0.01). Characteristics associated with higher odds of difficulty included addiction board certification (OR=1.88, p<0.01), working in multiple practice settings (OR=1.78, p=0.01), layoffs or furloughs or reduced hours (OR=1.66, p=0.01), staff sick with COVID-19 (OR=1.66, p=0.01), and group meetings to check in on staff well-being (OR=1.85, p=0.01) (Table 3). Likelihood-ratio tests revealed no significant variation at the state level. No associations were found between demographic area and reporting disruptions to functioning or work-life balance.

Discussion

Our findings provide insights into potential impacts of COVID-19 on HCP well-being after the transition from in-person to remote treatment of opioid use disorder. Only 16% of the HCPs were working onsite as usual during the COVID-19 pandemic period surveyed (in summer 2020), likely a dramatic change because these HCPs previously provided only minimal remote treatment. A minority reported negative impacts on well-being because of the pandemic. Disparities were found by gender and age, with male and older HCPs apparently being less affected by the pandemic. Additionally, significant associations of reported difficulties were found with addiction board certification, practice setting, organizational factors, and HCP experiences. Demographic, provider-level, and organizational-practice variables warrant further investigation of the impact of COVID-19 on HCP well-being.

The small fraction of respondents reporting functional impairment because of anxiety about COVID-19 may be explained by the timing of the survey. Associations with reduced staffing and negative impacts on work-life balance are consistent with previous research reporting that longer hours are associated with burnout (26, 27), because affected HCPs likely have had increased clinical and administrative demands. Telework was not specifically assessed, but the rapid transition to telework was likely disruptive to HCPs’ lives.

Our gender- and age-related findings are consistent with previous research showing associations between gender and age with burnout among HCPs. We did not ask respondents about having children in the home; however, associations between demographic variables (e.g., gender or age) may be explained by this factor. Our study is the first to document differences in the well-being of HCPs by addiction board certification, with those reporting either addiction medicine or addiction psychiatry board certification having higher odds of reporting difficulty with work-life balance. A previous study of physician job satisfaction reported no differences between addiction or nonaddiction specialties or between buprenorphine-waivered and nonwaivered prescribers (9). Board-certified professionals may be treating a larger number of patients with more severe substance use and co-occurring psychiatric disorders, which have worsened during the pandemic.

HCPs working in multiple practice settings were more likely to report negative impacts on work-life balance, possibly because of increased administrative demands. Feeling close to patients was associated with functional impairments from anxiety. Although this association was unanticipated, the survey item was included to understand the impact on rapport between HCPs and patients. It is possible that increased rapport coincided with increased worry for patient well-being. Associations between difficulty with work-life balance and meeting to check in on well-being as a group might be related to perceived infringement of work-life barriers or conversely may indicate an increased need for well-being check-ins. However, no significant association was found between functional impairments and well-being check-in meetings or between feeling supported by the organization and either well-being metric.

We hypothesized more negative impacts on well-being among those working in urban settings because of the precipitous impacts of COVID-19 in large U.S. cities early in the pandemic. However, we found no significant associations between demographic area and either measure of HCP well-being after accounting for clustering of HCPs within states. The reason for this finding likely is that our measure was overly simplified because some urban centers were significantly more affected by the pandemic early on (28). With a larger sample, an analysis comparing highly affected regions with less affected ones may provide a better understanding of the impact of infection rate severity.

A limitation of our survey was that we recruited a convenience sample through e-mail listservs and had a low number of responses relative to the total listserv subscribership and to the total number of known buprenorphine-waivered prescribers; this low response rate limited generalizability of our findings to all HCPs who treat patients with opioid use disorder. HCPs who were differentially affected by the COVID-19 pandemic may have varied in their willingness or ability to complete the survey. However, we believe that the large number of responses within a relatively small subspeciality in a short time frame provides important insights to better understand the well-being of HCPs who have treated patients with opioid use disorder during the pandemic. Findings related to gender and age were consistent with those of previous research, increasing our confidence in the reliability of the results, which highlight that women and younger HCPs were particularly affected by the pandemic.

Additionally, a low number of responses were received from providers of color, with Black and Hispanic HCPs making up only 3% and 6% of the respondents, respectively. Although our survey attempted to reach a diverse group of prescribers of medications for opioid use disorder through the listservs of large organizations, we relied on voluntary participation. The percentage of Black and Hispanic physicians in the United States is approximately 5% and 6%, respectively (29), which was approximated by our sample. This convenience sample provides early insights that should be further evaluated among HCPs who treat patients with opioid use disorder.

We note that the changing course of the COVID-19 pandemic and survey timing may have affected our results. Survey responses were obtained during a summer 2020 COVID-19 peak in national case numbers. This surge was followed by a period of relative stability in the fall of 2020, followed by another surge in the winter, with a 7-day average more than three times that in the summer (28). Notably, the winter surge coincided with COVID-19 vaccine rollouts in the United States. Numerous variables may have affected HCP well-being at various time points during the pandemic, and our survey provides only a snapshot of these impacts.

We also measured well-being broadly and did not use validated burnout scales. Our intention with the survey was to begin to understand HCP practice changes related to COVID-19 as well as HCP perceptions and impacts of changes. In this analysis, “functioning” and “work-life balance” were identified as potential proxies for HCP well-being. We recognize that well-being and burnout are multifactorial and are not fully encompassed by these dependent variables and therefore warrant further investigation with validated instruments.

Our findings highlight the importance of emphasizing HCP well-being and strategies to reduce burnout. A meta-analysis found that organization-directed interventions had significantly greater effects than physician-directed interventions (30). Effective organization-directed interventions included reducing or modifying work or on-call shifts and initiating targeted quality-improvement projects (26, 27, 31), and effective physician-directed strategies included mindfulness-based interventions (32, 33). A narrative review identified several organizational approaches to prevent or manage burnout among staff: developing and publicizing employee assistance programs; scheduling genuine break time; and having regular meetings to discuss goals, hours, tasks, and fairness (34). Five physician-level principles were also identified: establishing work-life balance; making career decisions by first identifying what is energizing or else draining; nurturing well-being strategies, social networks, and self-care; becoming engaged or reengaged; and building resilience.

Conclusions

This multiorganizational collaborative study sought to understand the impacts of organizational changes related to the COVID-19 pandemic on the well-being of HCPs who treat patients with opioid use disorder. We identified an impact on HCPs’ ability to balance professional and personal lives adequately during the pandemic. Our results should inform organizational leadership of what pandemic-related changes are associated with changes in HCP well-being. We propose that employers should consider the specific needs of women, younger HCPs, and HCPs who work in multiple settings; employers should also carefully evaluate the intention of well-being check-ins and consider the structure and frequency to focus on genuine well-being needs and avoid infringement on work-life balance. Further research should study trends to determine whether COVID-19–related functionally impairing anxiety and work-life balance disruptions persist and what interventions should be implemented to mitigate harms; a particular focus should be given to HCPs who are Black, Indigenous, and other people of color, given known differential impacts on and disparities in health care delivery for these groups before and during the COVID-19 pandemic (35–38).