Introduction
High rates of stress1,2 and burnout (a syndrome of emotional exhaustion3)4,5 are robustly demonstrated in persons working in veterinary medicine. Although large-scale efforts to identify and understand occupational distress abound,6–10 there is a paucity of rigorously tested interventions addressing stressors in the field.11–13 Recent research suggests that clients who are themselves distressed in the context of their pet’s illness (ie, experiencing “caregiver burden”)14 are likely to engage in challenging interactions with client-facing members of the veterinary healthcare team, leading to stress and burnout for the team.15,16 These interactions lay the foundation of the “burden transfer” theory, which posits that caregiver burden in the pet owner elicits various behaviors that contribute to difficult interactions with providers in veterinary medicine, transferring some of the client’s burden to the veterinary healthcare team.17,18 These difficult interactions, referred to as the burden transfer DANCE,18 (DANCE being an acronym for domains of interactions: Daily hassles, Affect, Nonadherent/inconsiderate behaviors, Confrontations, and Excess communications [Table 1], described in detail in prior work16–18) are strongly related to caregiver burden in the veterinary client and, in turn, correlate with stress and burnout for members of the veterinary healthcare team.16,19
Examples of the domains of the burden transfer DANCE.
| Domain | Examples |
|---|---|
| Daily hassles | Client wants impossible predictions |
| Client follows nonprofessionals’ advice about patient health needs | |
| Client shops around to compare costs | |
| Affect | Client requires euthanasia counseling |
| Client demonstrates anxiety | |
| Client demonstrates sadness or grief | |
| Nonadherent/Inconsiderate | Client declines recommended workup |
| Client refuses treatment | |
| Client no-shows for appointments | |
| Confrontation | Client becomes upset or blames |
| Client refuses to pay for services | |
| Client makes a complaint | |
| Excess communication | Client makes frequent phone calls |
| Client makes frequent email contact |
While the frequency of burden transfer DANCE interactions significantly predicts stress and burnout for persons working in the field, their reaction to these encounters, or how bothered the veterinary healthcare team member feels by the situation, is a far more important determinant of negative outcomes than just the frequency of the occurrence of these interactions.16,19 Reaction to a burden transfer DANCE encounter has thus been identified as a potential mechanism for negative outcomes in the field. Crucially, this means that burden transfer reaction is a modifiable risk factor for stress and burnout in the field.16–19
Recently, a randomized, controlled trial examined whether an acceptance and commitment training (ACT) program, developed to reduce this reaction, would reduce stress and burnout in veterinary personnel relative to a wait-list control.18 ACT incorporates 6 interrelated elements: being present (ie, consciously experiencing internal and external events as they occur), acceptance (ie, embracing inner struggles with one’s own thoughts, feelings, and urges), defusion (ie, unhooking from or viewing these inner processes of thoughts, feelings, and urges as being separate from the self), acknowledgment of the self-as-context (ie, observation of inner experiences as if from an outside perspective), values (ie, areas of importance to an individual that can be embraced as ways to guide action), and taking committed action (ie, choosing behaviors that align with a chosen value)20,21; this framework served as the basis of the program. This trial recruited participants across client-facing positions in veterinary medicine (ie, veterinarians, technicians, assistants, customer service representatives, and management) from 17 clinics, including general veterinary practices, specialty referral and emergency hospitals, and an academic medical center. Participants randomized to the intervention demonstrated high rates of acceptability, with 93% indicating the program was “useful” or “very useful.” They also showed frequent use of skills taught, averaging use of these techniques approximately 5 times daily at both post-testing and follow-up. Significantly reduced burden transfer reaction, stress, work-related burnout, and client-related burnout were demonstrated at post-testing and maintained at follow-up. With efficacy of the program demonstrated in a controlled trial, next steps must begin to target effectiveness in the field. To broadly disseminate the program, it is important to understand whetehr an asynchronous or self-paced version, accessed at the individual’s convenience through a learning platform, is as effective as the original live and interactive version.
The current study was a randomized, parallel-arms, noninferiority trial comparing the original live version of this ACT-based educational program to a self-paced format. Both were delivered to participants via in-house learning management systems of 2 large veterinary corporate groups. We hypothesized that (1) participants assigned to the self-paced condition would show better rates of program completion relative to the live condition; (2) participants would show similar accuracy and retention on a test of program knowledge, similar rates of program usefulness, and similar use of techniques taught in the program in daily life across both conditions; and (3) burden transfer, stress, and burnout (work and client related) would show similar decreases from pre- to postassessment, with maintenance of improvements at 1 month after and (for a subset of participants) at extended follow-up.
Materials and Methods
Participants
Employees from 2 nationwide (US) corporate veterinary groups were invited to participate. Hospitals and clinics included represented a variety of settings, including specialty referral, urgent care, emergency, and general practice. Inclusion criteria for the study were (1) age over 18 years and (2) ability to speak and comprehend English sufficiently to complete study measures and participate in the program. The sole research exclusion criterion was working in a position involving no client interaction.
Procedure
Procedures for this randomized, parallel-arms (live vs self-paced) trial were approved by the Institutional Review Board of Kent State University. The study was conducted between October 2021 and November 2022. To facilitate enrollment, the program was offered through the in-house learning systems of each group (one group using Dayforce Learning [Ceridian], the other using Docebo [Docebo Inc]). To ensure similarity across sites, researchers were involved in the setup and testing of the program’s functionality within each system. The program was offered as an educational training opportunity, and Registry of Approved Continuing Education credits were made available. Veterinary healthcare team members who enrolled in the program were presented with the option to also participate in research. Those who were not interested in research were permitted to complete the program just for the educational opportunity and/or Registry of Approved Continuing Education credit. For those agreeing to be research participants, following informed consent and completion of baseline pretest questionnaires (Qualtrics XM; Qualtrics), autorandomization assigned each participant to either the live or self-paced condition. Participants completed the baseline pretest assessment within 2 weeks prior to beginning the program, the post-test assessment within 2 weeks following program completion, and follow-up assessment 1 month later. To examine effects at an extended time point, additional funding was obtained to invite the first cohort of participants who completed the study (n = 51) to participate in an extended follow-up assessment conducted 9 to 12 months after program completion.
Sessions for the live condition were synchronously delivered by a member of the research team (MBS [licensed clinical psychologist who contributed to program development]) via a video-conferencing link placed in the learning system. During the 3-week period in which the 3-hour-long session program was offered, live condition sessions were made available 5 times throughout the day on dates identified prior to enrollment. Participants were aware of session availability at the time of enrollment; although asked to commit to a specific time, they were allowed to attend at a different time as needed. Modules for the self-paced condition were asynchronously delivered in lockstep with content from the live sessions: content delivered during week 1 for the live condition was covered in self-paced modules 1 and 2, which were assigned for completion in that same week. Similarly, modules 3 and 4 covered content from live sessions during week 2 and were assigned to be completed during week 2. Modules 5 through 7 covered content from week 3 live sessions and were assigned to be completed during week 3. Each module took approximately 20 to 30 minutes to complete, for a similar total time as the live condition (ie, approx 3 hours). Prior to use, all 7 self-paced modules were checked for fidelity to the ACT protocol by a separate member of the team (ASGU), who also checked 20% of the live program sessions. For all sessions, 100% fidelity to the protocol was achieved.
Pretest measures began with informed consent describing the study purpose, rights as a study participant, information related to institutional review board approval, and contact information for the principal investigator (MBS). The pretest protocol did not assess knowledge of program content, request ratings of program usefulness, or assess frequency of techniques taught during the program. Otherwise, the protocol was the same for all time points (see Measures section). All data were collected online. Participants were reimbursed for each survey completed via gift card for an online retailer.
ACT educational program
The ACT framework20,21 serves as a basis for the program and has been previously described in detail17,18; the program will thus be briefly summarized here. The program utilized here combines experiential exercises and specific skills for behavior change, encouraging a focus on reactivity to DANCE interactions. Techniques to enhance resilience in the context of challenging thoughts and feelings from difficult client interactions are modeled and practiced. Participants are asked to identify personal values and create individualized plans for concrete, values-guided action. “Homework” to practice these skills between sessions is assigned. Week 1 (live condition session 1; self-paced condition modules 1 and 2) includes an overview of the program rationale, skills for being mindfully present, and identification of DANCE encounters (ie, potential “points of burden transfer”). Week 2 (live condition session 2; self-paced condition modules 3 and 4) focuses on accepting and unhooking (ie, detaching) from difficult thoughts, feelings, and urges due to DANCE interactions and on separating the self from this struggle in an effort to better observe it. Week 3 (live condition session 3; self-paced condition modules 5 through 7) helps participants clarify the role of personal values in their work environment and develop concrete action plans that align with these values. All program sessions and materials (both live and self-paced) were accessed through the in-house learning management system of participating veterinary groups.
Measures
Components of the assessment protocol conducted that have been previously described17,18 will be briefly summarized here. The burden transfer inventory16 (BTI) measured burden transfer, asking participants about experiences across the 5 domains of client DANCE interactions. Higher BTI subscale scores reflect greater frequency of (BTI frequency) or reaction to (BTI reaction) DANCE encounters. This measure shows excellent internal consistency across subscales for frequency and reaction (α = 0.92 to 0.94 for combined subscales) as well as test-retest reliability (r = 0.78 to 0.82 for combined subscales). In the study reported here, BTI reaction was the primary variable of interest. The Perceived Stress Scale (PSS)22 was used to assess the degree to which one feels that life is unpredictable or overloaded. Higher PSS scores reflect greater current stress. This measure demonstrates internal consistency ranging from α = 0.68 to 0.78.23 The Copenhagen Burnout Inventory (CBI)24 was used to measure work- and client-related burnout, with higher CBI scores indicating greater current burnout for each domain. This measure demonstrates high internal consistency (α = 0.85 to 0.87).
In addition to the above previously validated measures, at post-test and follow-up time points, participants reported usefulness of the material presented (scale of 1 to 5, with higher numbers representing greater usefulness) and frequency of ACT techniques used during the preceding 2 weeks (again, higher numbers reflecting greater frequency of skill use). Additionally, a knowledge test (multiple choice) was developed for fact-based components of the program; the information covered by these questions was covered in both live and self-paced programs.
Participants also self-reported demographic information, specifically gender (multiple choice), age (numeric slide bar), race or ethnicity (multiple choice), nature of employment (multiple choice), and years in the field (numeric slide bar).
Power analysis
Based upon data from a previous trial,18 an effective curve reliability of 0.60 was estimated for the same outcome measures based on changes following the live format of the intervention used in the present study.25 Effective curve reliability is calculated as the proportion of slope variance not due to error and is an index that ranges from 0 to 1, becoming larger as the study design has reduced error. It can thus be interpreted as a standardized effect size; in this case, a moderate effect size was expected. Regarding maximum likelihood estimates for longitudinal growth curve models, a minimum of 5 to 10 observations/item or parameter is recommended.26 In light of attrition rates of approximately 40% from initial enrollment to actual study completion noted in prior similar work,18 we planned to recruit 250 participants. The final sample of 137 in the present report yields approximately 20 observations/parameter to be estimated—well within recommendations.26
Statistical analyses
Statistical analyses were conducted using commercially available software.27,28 First, independent samples t tests were used to examine for any pretest differences between participants retained versus removed from analyses. An intent-to-treat approach was then taken for analyses. To examine whether type of employment should be considered in analyses, repeated-measures analyses of variance were conducted. No differences emerged on effects of employment type on PSS, CBI—client-related burnout, CBI—work-related burnout, and BTI reaction; this variable was thus not considered further in analyses. Next, demographic information and primary variables were characterized using descriptive statistics (percent for categorical data, and mean, SD, and minimum-maximum for numeric data). To explore whether barriers to workshop completion differed significantly between the live and self-paced programs, 2 proportion z tests were completed. Descriptive statistics (percent for categorical data, and mean, SD, and minimum-maximum for numeric data) were used to characterize acceptability, knowledge retention, and frequency of technique use. Independent-sample t tests were used to examine group differences in program acceptability and knowledge retention.
To examine the effect of the ACT program on primary outcomes, latent growth curve (LGC) analysis was used. Latent growth curve represents repeated measures of a given variable as a function of time. Each time point measurement contributes to 2 indicators of growth: initial status (intercept) and change over time (slope) on which individuals may vary. Because initial status is analogous to the intercept in a regression equation, in each of the tested models, unstandardized loadings of all indicators (the 3 primary time points for each of the measured constructs) on initial status were fixed to 1. To specify linear trends, the loading of time 1 (pretest) was fixed to 0, time 2 (post-test) was fixed at 1, and time 3 (follow-up) was fixed to 2. This is analogous to centering in hierarchical linear modeling. To examine effects of the ACT program, condition was included as a fixed covariate, and the variances from the latent variables (intercept and slope) were allowed to covary. This allowed for the examination of the relationship between levels of the dependent variable in the model and rate of change across time. As certain fit indices are influenced by sample size and we hypothesize there will be no differences in rate of change between groups, in the event of suboptimal model fit, we included repeated-measures ANOVA to examine change in each outcome over the course of the ACT program; this also allowed examination of effect size.
Multiple imputation of missing data was accomplished with the R multivariate imputation via chained equation (MICE) package. Assuming the data are missing at random, MICE imputes data on a variable-by-variable basis. Linear regression is used to predict continuous missing values, and multiple data sets are generated, in this case 5. Missing values were replaced by means of the values generated in these 5 data sets. Each model was assessed with the following fit indices: χ2 (χ2), χ2/df, root mean square error approximation (RMSEA), the standardized root mean square residual, comparative fit index (CFI), and the Tucker-Lewis index (TLI). Although χ2 is commonly reported, it is sensitive to sample size and will more often than not return a significant value. χ2/df provides an alternative to χ2 as a measure of model fit, with values < 5 considered an adequate fit. Root mean square error approximation is an absolute fit index, with values < 0.08 considered adequate fit and less than 0.05 considered good fit.29 Comparative fit index and TLI are incremental fit indices; Bentler and Bonett30 recommend that TLI > 0.90 is acceptable fit, whereas Hu and Bentler31 suggest that CFI and TLI > 0.95 indicate good model-data fit.
Of those who were included in the above analyses, 33 completed measures at extended follow-up. Data were aggregated across conditions before paired samples t tests were conducted to determine whether burden transfer, stress, and burnout (work and client related) would continue to show decreases from pretest to extended follow-up. Procedures described above for exploration of knowledge retention and frequency of technique use were repeated within this subsample.
Results
Participants
Of the 429 individuals who indicated interest in the program, 274 completed informed consent, agreeing to participate. Of those who agreed to participate, 22 did not return to complete pretest measures or program sessions. Of the remaining 252 participants, 105 completed pretest questionnaires but did not participate in any part of the program and were dropped from the analyses. Of the 137 individuals who were included in analyses, 99 completed the entire intervention, and 23 completed at least 2 of the 3 weeks; all were retained in analyses given the intent-to-treat approach to analysis. Imputed data account for no more than 7.8% of the data in any given analysis. Independent-sample t tests indicate no difference between individuals retained in analyses versus those removed due to failure to return on scores of the pretest PSS (P = .31), CBI—work-related burnout (P = .30), CBI—client-related burnout (P = .83), and BTI reaction (P = .83).
The final analytic sample was 137 (live, n = 55; self-paced, 82; Figure 1). Both groups were comprised primarily of females (94.9%) identifying as Caucasian or white (92.7%), averaging approximately 38 years of age. Employment types included veterinarians (21.2%), technicians (27.7%), assistants (5.8%), customer service representatives (17.5%), management (18.2%), and other (9.6%). Years of employment in the field ranged from 1 to 41 years, averaging about 12 years. No demographic variables differed significantly between those assigned to self-paced versus live conditions (Table 2).
Enrollment flow diagram for a study comparing effectiveness of live and self-paced versions of an acceptance and commitment training educational program on outcomes of burden transfer, stress, and burnout in small animal veterinary hospital employees.
Citation: Journal of the American Veterinary Medical Association 261, 6; 10.2460/javma.23.01.0037
Demographics of participants in a study comparing effects of live and self-paced versions of an acceptance and commitment training program on burden transfer, stress, and burnout among small animal veterinary hospital employees.
| Variable | Group | |
|---|---|---|
| Live | Self-paced | |
| Age (y) | ||
| Mean (SD) | 38.3 (10.2) | 38.4 (11.1) |
| Range | 22–65 | 22–65 |
| Race-ethnicity | ||
| White | 51 (92.7) | 76 (92.7) |
| Hispanic or Latino | 1 (1.8) | 4 (4.9) |
| Black or African American | 1 (1.8) | 1 (1.2) |
| Asian American or Pacific Islander | — | 1 (1.2) |
| Native American | 1 (1.8) | — |
| Other | 1 (1.8) | — |
| Gender | ||
| Female | 52 (94.5) | 78 (95.1) |
| Male | 3 (5.5) | 4 (4.9) |
| Experience (y) | ||
| Mean (SD) | 13.3 (8.2) | 11.8 (9.5) |
| Range | 1–37 | 1–41 |
| Position | ||
| Veterinarian | 12 (21.8) | 17 (20.7) |
| Technician | 14 (25.4) | 24 (29.3) |
| Assistant | 3 (5.5) | 5 (6.1) |
| Customer service | 6 (10.9) | 18 (22.0) |
| Management | 15 (27.3) | 10 (12.2) |
| Other | 5 (9.1) | 8 (9.8) |
Participants were randomly assigned to the live (n = 55; synchronous delivery) or self-paced (82; asynchronous delivery) condition. Unless otherwise specified, data are given as number (percentage).
Comparison of program completion and barriers to completion
Participants assigned to the self-paced condition showed better rates of full program completion relative to the live condition (self-paced, 48.4%; live, 30.5%; z = –2.91; P < .05). For those who did not complete the full program, a majority cited unexpected life circumstances as the primary barrier, though a small proportion described difficulties using their learning management system. Participants assigned to the live condition were more likely to not complete the program due to unpredictable events (z = 2.76; P < .05), whereas those assigned to the self-paced condition more often cited feeling overwhelmed or too busy as a reason for noncompletion (z = –3.01; P < .05; Table 3).
Completion and barriers to program completion for live and self-paced versions of an acceptance and commitment training program addressing burden transfer, stress, and burnout among small animal veterinary hospital employees.
| Completion | Group | P value | |
|---|---|---|---|
| Live | Self-paced | ||
| Full workshop completion | 39 (30.5) | 60 (48.4) | < .05 |
| Partial workshop completion | 62 (48.4) | 85 (68.5) | < .05 |
| Enrolled, but did not complete any part of workshop | 66 (51.6) | 39 (31.5) | < .05 |
| Barrier | |||
| Felt too busy/overwhelmed | 5 (21.7) | 14 (66.7) | < .05 |
| Scheduled conflict (eg, scheduled work) | 8 (34.8) | 4 (19.0) | ns |
| Unpredictable event (eg, work emergency, illness) | 7 (30.4) | 0 | < .05 |
| Technical problems (eg, learning system login) | 2 (8.7) | 2 (9.5) | ns |
| Other | 1 (4.3) | 1 (4.8) | ns |
Participants were randomly assigned to the live (synchronous delivery) or self-paced (asynchronous delivery) condition. Data are given as number (percentage).
ns = Non-significant.
Comparison of acceptability, knowledge retained, and frequency of techniques used
On measures completed within 2 weeks after the intervention period (ie, post-test assessment), more participants assigned to the live condition rated the program as “helpful” or “very helpful” compared with the self-paced condition (live, 94.5%; self-paced, 86.1%; t(132) = 3.07; P < .01). Accuracy of overall program knowledge averaged 82.5% at both post-test (± 13.1%; minimum to maximum, 35.0% to 100.0%) and follow-up (± 14.1%; minimum to maximum, 20.0% to 100.0%). Program knowledge did not differ significantly across conditions (P > .05), reflecting good retention of learned information, regardless of whether the ACT program was completed live or self-paced. Reported frequency of technique use for the 2 weeks preceding evaluation was 5.72 times/d (80.1 ± 43.6; minimum to maximum, 10 to 184 times) at post-test and 5.37 times/d (75.2 ± 48.5; minimum to maximum, 4 to 195 times). Frequency of technique use did not differ significantly across conditions (P > .05).
Change in BTI reaction
To examine change in burden transfer reaction, the LGC model was specified with pretest, post-test, and follow-up as the 3 time points to determine intercept and slope, with condition (live vs self-paced) as a fixed covariate. Fit indices suggest the model provided a good fit to the data (χ2 [4] = 5.53; P = .24; χ2 /df = 1.38; CFI = 0.99; TLI = 0.97; RMSEA = 0.05 [90% CI, 0.00 to 0.15]). In this model, conditions did not differ in initial reported burden transfer (intercept; β = 0.02; P = .88), nor did condition predict slope (β = 0.12; P = .55). Regardless of whether the ACT program was completed live or self-paced, participants reported less BTI reaction over time. Slope and intercept were not associated (r = 0.08; P = .79), suggesting that initial levels of BTI reaction were not associated with rate of change over time (Figure 2). To ensure significant change in BTI reaction over time, we conducted a repeated-measures ANOVA with condition as a between-subjects variable. Significant results were observed for the main effect of time (F[2,270] = 54.46; P < .001; partial η2 = 0.29). The main effect of condition was not significant (F < 1), and the interaction was also not significant (F[2,270] = 1.31; P = .27; η2 = 0.01). Whether the ACT program was completed live or self-paced, participants reported lower BTI reaction over time.
Line graphs of (A) Mean Burden Transfer Inventory—reaction, (B) Perceived Stress Scale, (C) Copenhagen Burnout Inventory—work-related burnout, and (D) Copenhagen Burnout Inventory—client-related burnout scores by group (acceptance and commitment training live vs self-paced) and time (before, 1 week after, and 1 month after). Error bars represent within-subject 95% CIs.
Citation: Journal of the American Veterinary Medical Association 261, 6; 10.2460/javma.23.01.0037
Change in PSS
To examine change in perceived stress, the LGC model was specified as described above. Fit indices suggest the model did not provide a good fit to the data (χ2 [4] = 32.32; P < .001; χ2 /df = 8.08; CFI = 0.81; TLI = 0.71; RMSEA = 0.23 [90% CI, 0.16 to 0.30]). Although χ2 is commonly reported, it is sensitive to sample size and will more often than not return a significant value; RMSEA tends to be inflated with low degrees of freedom. Poor model fit might also be driven by the lack of condition contributing to variance in the intercept and slope. Conditions did not differ in initial reported PSS (intercept; β = 0.05; P = .62), nor did condition predict slope (β = 0.13; P = .37). Slope and intercept were not associated (r = −0.38; P = .08), suggesting that initial PSS levels were not associated with rate of change over time. To ensure significant change in PSS over time, we conducted a repeated-measures ANOVA with condition as a between-subjects variable. Significant results were observed for the main effect of time (F[2,270] = 28.57; P < .001; partial η2 = 0.18). The main effect of condition was not significant (F[1,135] = 1.28; P = .26; partial η2 = .01) and the interaction was also not significant (F < 1). Whether the ACT program was completed live or self-paced, participants reported lower PSS over time (Figure 2).
Change in CBI—work-related burnout
To examine change in work-related burnout, the LGC model was specified as described above. Some fit indices suggest the model provided a good fit to the data (χ2 /df = 4.41; CFI = 0.91). Others (χ2 [4] = 19.04; P < .01; TLI = 0.87; RMSEA = 0.16 [90% CI, 0.09 to 0.23]) did not. Based on the biases inherent to some fit indices, we are confident in the specified model. In this model, condition did not predict slope (β = 0.35; P = .06), indicating groups did not differ in change in CBI—work-related burnout over time. Condition also did not predict the intercept (β = 0.01; P = 0.92), indicating that participants in the 2 conditions did not differ in initial CBI—work-related burnout. Slope and intercept were not associated (r = −0.25; P = .41), suggesting that initial levels of CBI—work-related burnout were not associated with a greater rate of change over time. As some fit indices did not suggest adequate fit, repeated-measures ANOVA was conducted with condition as a between-subjects variable. A significant main effect of time (F[2,270] = 30.85; P < .001; partial η2 = 0.19) was observed. The main effect of condition was not significant (F[1,135] = 1.26; P = .27; partial η2 = 0.01), and the interaction was also not significant (F[2,270] = 2.30; P = .10; η2 = 0.02). Whether the ACT program was completed live or self-paced, participants reported lower CBI—work-related burnout over time (Figure 2).
Change in CBI—client-related burnout
To examine change in burnout related to clients, the LGC model was specified as described above. Some fit indices suggest the model provided a good fit to the data (χ2 /df = 3.90; CFI = 0.95; TLI = 0.93), whereas others did not (χ2 [4] = 15.61; P = .004; RMSEA = 0.13 [90% CI, 0.06 to 0.19]). Based on the biases inherent in some fit indices, we are confident in the specified model. In this model, condition did not predict slope (β = 0.30; P = .31), indicating groups did not differ in change in client-related burnout over time. Condition also did not predict the intercept (β = 0.07; P = .42), indicating that participants in the 2 conditions did not differ in initial CBI—client-related burnout. Slope and intercept were not associated (r = −0.37; P = .43), suggesting that initial levels of CBI—client-related burnout were not associated with a greater rate of change over time. As some fit indices did not suggest adequate fit, repeated-measures ANOVA was conducted with condition as a between-subjects variable. A significant main effect of time (F[2,270] = 41.50; P < .001; partial η2 = 0.24) was observed. The main effect of condition was not significant (F[1,135] = 3.04; P = .08; partial η2 = 0.02), and the interaction was also not significant (F[2,270] = 1.79; P = .17; η2 = 0.02). Whether the ACT program was completed live or self-paced, participants reported lower CBI—client-related burnout over time (Figure 2).
Extended follow-up
For the subset of individuals who completed extended follow-up, overall program knowledge averaged 83.1% (± 10.2%; minimum to maximum, 65.0% to 100.0%). Reported frequency of technique use for the 2 weeks preceding evaluation was 4.08 times/d (57.1 ± 43.4 times over 2 weeks; range, 0 to 142 times). A comparison of pretest to extended follow-up was conducted across primary outcome measures within the subset of individuals who provided extended follow-up data. Relative to pretest, at extended follow-up, participants reported lower BTI reaction (t[32] = 5.19; P < .001), PSS (t[32] = 3.56; P < .001), CBI—work-related burnout (t[32] = 3.03; P = .005), and CBI—client-related burnout (t[32] = 4.33; P = .005).
Discussion
This randomized parallel-arms trial compared live (synchronous and interactive) and self-paced (asynchronous and independent) versions of an ACT-based educational program designed to reduce reactivity to DANCE interactions with veterinary clients. Although a higher percent of participants described the live program as “helpful” or “very helpful” compared with the self-paced version, those assigned to the self-paced version were more likely to complete the program during the study period. The 2 formats showed no difference in accuracy on a knowledge-based test or use of techniques taught. Importantly, improvements in burden transfer reactivity, stress, and both work- and client-related burnout were shown in both conditions, with large effects observed and gains being maintained 1 month later; no differences between groups emerged in these outcomes. Finally, the subset of individuals examined 9 to 12 months later demonstrated continued benefit of the program relative to baseline.
A past randomized controlled trial examining a live version of this program compared with wait-list control18 demonstrated that 93% of participants endorsed the program as helpful. Although the live version of the program in the current study showed similar program acceptability (94.5% endorsing it as “helpful” or “very helpful”), this number was 86.1% for self-paced. While lower, this acceptability is still an improvement over the general statistic for the field of < 50% finding well-being and mental health resources useful.11 Moreover, actual rates of program completion were better for individuals assigned to the self-paced format, with fewer individuals citing barriers to completion. At both post-test and follow-up, average knowledge maintained from the program showed no differences between formats, and utilization of techniques learned in the program was more than 5 times/d for both groups. Importantly, both versions of the program significantly reduced reaction to DANCE encounters, stress, and both work- related and client-related burnout in participants. The effect sizes of these improvements were large and were maintained 1 month after the program ended in the full sample. The large effects noted here suggest that the program leads not just to statistically significant improvements, but to meaningful change for individuals who complete it. Overall, it appears that the interactive component of the previously reported live program18 is not essential to the success of the program. An asynchronous version is acceptable, adequately disseminates knowledge, promotes use of techniques that enhance resilience, and can provide similar overall benefits.
Additionally, data from the subset of individuals who completed extended follow-up measures 9 to 12 months after the initial program suggest good retention of information learned in the program, continued daily use of skills taught, and sustained improvement in burden transfer reaction, stress, and burnout relative to baseline. Because participants reported using program skills an average of approximately 4 times/d at that extended time point (relative to > 5 times daily at proximal time points), it is worth considering whether a booster program would further enhance skill use and outcomes. This subsample was small and not part of the initially planned data collection; as such, future work should examine extended effects of this program over time. However, the current data are promising.
Implications for scalability and sustainability of this program are clear. The study was designed to be ecologically valid, with utilization of the in-house learning management systems of our partnering corporate veterinary groups. A small number of individuals (4 total) reported experiencing technological difficulties preventing program completion within the time-limited study period; however, the learning systems were relatively new for both partners at the time of data collection. It is expected that this barrier would be reduced with eased demands for completion rate (weekly modules were assigned for research purposes to keep self-paced and live versions in lockstep with each other) and with greater learning system familiarity on the part of participating employees. In combination, our results supported that an asynchronous, learning system–based version of this ACT program could be an effective way to share its content, making its benefits readily available to individuals working in the field of veterinary medicine. Efforts must now focus on creating a version appropriate for broad dissemination. Moving forward, due to the better completion rates observed with the self-paced format, providing this program in a manner that allows individuals to complete it at their convenience will be important. Because ratings of helpfulness were greater with the live format, it is worth considering whether this is due to the higher level of engagement required when the program involves a live instructor. Future work might aim to replicate the engagement of live instruction through a requirement of active responding during the program to determine whether ratings of helpfulness can be increased.
Similar to previous work, high rates of attrition occurred from initial recruitment to actual program completion. Because no demographic or baseline outcome measure differences were observed between individuals retained in analyses versus those removed due to failure to return, the phenomenon of low completion rate is not attributed to psychological state (ie, participants with higher baseline distress did not show poorer program completion). Rather, it is likely that this occurred secondary to our efforts to recruit participants in an ecologically valid manner. Specifically, the program was offered as an educational opportunity for all employees, with the option to participate in research being made available for anyone enrolling in the program. This method was employed to enhance generalizability of results, but may have led to a lower sense of obligation, if participants enrolled primarily for the educational opportunity rather than as a commitment to participate in research. Despite attrition rates, engagement was similar to that found in our prior work.18 Because the majority of individuals who dropped out did so prior to attending the first session, it is possible that the remaining sample comprised individuals who, in general, would be more inclined toward adherence to program recommendations. If that is the case, the current sample may include individuals who are more likely to enact techniques taught compared with the general population of persons working in veterinary medicine, and thus more likely to show benefit. Additionally, it is noted that outcomes, including frequency of techniques utilized, are limited to self-report. Although skills taught in this program are largely cognitive in nature, making objective observation difficult to gauge, any construct being assessed via self-report could be subject to biases of the individual providing that information. Another limitation is found in low representation of individuals who are often underrepresented in veterinary medicine. While the current sample does reflect the lack of diversity in veterinary healthcare teams in this country in general,32 it does not represent everyone working in the field. Continued work is needed to ensure this program works across cultures and backgrounds.
In summary, the current work compared the effectiveness of live and self-paced versions of an ACT-based educational program designed to reduce reactivity to difficult client interactions occurring in veterinary healthcare. Findings demonstrated that participants found both versions of the program helpful, retained similar amounts of information regardless of format, and did not differ in the degree to which they incorporated techniques taught in the program into their daily lives. The program significantly improved outcomes of burden transfer reactivity, stress, and burnout, regardless of format. Within the full sample, gains were maintained 1 month later. Within a subsample, gains were still observed up to 1 year later. Future directions must now focus on efforts to create materials appropriate for broad dissemination of this program.
Acknowledgments
The project described in this article was primarily supported by the Clinical Scholars program of the Robert Wood Johnson Foundation, with additional funding (extended follow-up data collection) from the Applied Psychology Center at Kent State University. The funding sources had no involvement in the study design, data analysis and interpretation, or writing and publication of the manuscript.
The authors declare that there were no conflicts of interest.
The authors would like to acknowledge Drs. Mark D. Carlson, Elizabeth Strand, and Aviva Vincent for their helpful comments during program development, as well as Mission Veterinary Partners and MedVet veterinary groups for their support of this work, particularly Karlene Belyea, Eric Bookmeyer, Courtney Choate, Elizabeth Knudsen, Dr. Michael Podell, and Stephanie Resnick. Finally, the authors wish to acknowledge the assistance of Gabrielle Heminger, BS; Andrea Kohut, RVT; and Karlee Patrick, BS, for their invaluable assistance in creating the asynchronous version of the program.
References
- 1.↑
Bartram DJ, Baldwin DS. Veterinary surgeons and suicide: influences, opportunities and research directions. Vet Rec. 2008;162(2):36–40. doi:10.1136/vr.162.2.36
- 2.↑
Nett RJ, Witte TK, Holzbauer SM, et al. Risk factors for suicide, attitudes toward mental illness, and practice-related stressors among US veterinarians. J Am Vet Med Assoc. 2015;247(8):945–955. doi:10.2460/javma.247.8.945
- 4.↑
Ouedraogo FB, Lefebvre SL, Hansen CR, Brorsen BW. Compassion satisfaction, burnout, and secondary traumatic stress among full-time veterinarians in the United States (2016–2018). J Am Vet Med Assoc. 2021;258(11):1259–1270. doi:10.2460/javma.258.11.1259
- 5.↑
Hayes GM, LaLonde-Paul DF, Perret JL, et al. Investigation of burnout syndrome and job-related risk factors in veterinary technicians in specialty teaching hospitals: a multicenter cross-sectional study. J Vet Emerg Crit Care (San Antonio). 2020;30(1):18–27. doi:10.1111/vec.12916
- 6.↑
Platt B, Hawton K, Simkin S, Mellanby RJ. Suicidal behaviour and psychosocial problems in veterinary surgeons: a systematic review. Soc Psychiatry Psychiatr Epidemiol. 2012;47(2):223–240. doi:10.1007/s00127-010-0328-6
- 7.
Bartram DJ, Baldwin DS. Veterinary surgeons and suicide: a structured review of possible influences on increased risk. Vet Rec. 2010;166(13):388–397. doi:10.1136/vr.b4794
- 8.
Brannick EM, DeWilde CA, Frey E, et al. Taking stock and making strides toward wellness in the veterinary workplace. J Am Vet Med Assoc. 2015;247(7):739–742. doi:10.2460/javma.247.7.739
- 9.
Moses L, Malowney MJ, Wesley Boyd J. Ethical conflict and moral distress in veterinary practice: a survey of North American veterinarians. J Vet Intern Med. 2018;32(6):2115–2122. doi:10.1111/jvim.15315
- 10.↑
Vande Griek OH, Clark MA, Witte TK, Nett RJ, Moeller AN, Stabler ME. Development of a taxonomy of practice-related stressors experienced by veterinarians in the United States. J Am Vet Med Assoc. 2018;252(2):227–233. doi:10.2460/javma.252.2.227
- 11.↑
Volk JO, Schimmack U, Strand EB, et al. Executive summary of the Merck Animal Health Veterinarian Wellbeing Study III and Veterinary Support Staff Study. J Am Vet Med Assoc. 2022;260(12):1547–1553. doi:10.2460/javma.22.03.0134
- 12.
Moir FM, Van den Brink A. Current insights in veterinarians’ psychological wellbeing. N Z Vet J. 2020;68(1):3–12. doi:10.1080/00480169.2019.1669504
- 13.↑
Rohlf VI. Interventions for occupational stress and compassion fatigue in animal care professionals—a systematic review. Traumatology (Tallahass Fla). 2018;24(3):186–192. doi:10.1037/trm0000144
- 14.↑
Spitznagel MB, Jacobson DM, Cox MD, Carlson MD. Caregiver burden in owners of a sick companion animal: a cross-sectional observational study. Vet Rec. 2017;181(12):321. doi:10.1136/vr.104295
- 15.↑
Spitznagel MB, Cox MD, Jacobson DM, Albers AL, Carlson MD. Assessment of caregiver burden and associations with psychosocial function, veterinary service use, and factors related to treatment plan adherence among owners of dogs and cats. J Am Vet Med Assoc. 2019;254(1):124–132. doi:10.2460/javma.254.1.124
- 16.↑
Spitznagel MB, Ben-Porath YS, Rishniw M, Kogan LR, Carlson MD. Development and validation of a Burden Transfer Inventory for predicting veterinarian stress related to client behavior. J Am Vet Med Assoc. 2019;254(1):133–144. doi:10.2460/javma.254.1.133
- 17.↑
Spitznagel MB, Updegraff A, Twohig MP, Carlson MD, Fulkerson CM. Reducing occupational distress in veterinary medicine personnel with acceptance and commitment training: a pilot study. N Z Vet J. 2022;70(6):319–325. doi:10.1080/00480169.2021.1938270
- 18.↑
Beth Spitznagel M, Updegraff ASG, Was C, et al. An acceptance and commitment training program reduces burden transfer, stress, and burnout among veterinary healthcare teams. J Am Vet Med Assoc. 2022;260(12):1554–1561. doi:10.2460/javma.22.05.0196
- 19.↑
Spitznagel MB, Martin JT, Carlson MD, Fulkerson CM. Validation of the Burden Transfer Inventory-abbreviated and examination across veterinary medicine positions and settings in the United States. Vet Rec Open. 2022;9(1):e46. doi:10.1002/vro2.46
- 20.↑
Hayes SC. Acceptance and Commitment Therapy and the New Behavior Therapies: Mindfulness, Acceptance and Relationship. Guilford Press; 2004.
- 21.↑
Hayes SC, Strosahl K, Wilson KG. Acceptance and Commitment Therapy: An Experiential Approach to Behavior Change. Guilford Press; 1999.
- 22.↑
Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J Health Soc Behav. 1983;24(4):385–396. doi:10.2307/2136404
- 23.↑
Perera MJ, Brintz CE, Birnbaum-Weitzman O, et al. Factor structure of the Perceived Stress Scale-10 (PSS) across English and Spanish language responders in the HCHS/SOL Sociocultural Ancillary Study. Psychol Assess. 2017;29(3):320–328. doi:10.1037/pas0000336
- 24.↑
Kristensen TS, Borritz M, Villadsen E, Christensen KB. The Copenhagen Burnout Inventory: a new tool for the assessment of burnout. Work Stress. 2005;19(3):192–207. doi:10.1080/02678370500297720
- 25.↑
Brandmaier AM, von Oertzen T, Ghisletta P, Lindenberger U, Hertzog C. Precision, reliability, and effect size of slope variance in latent growth curve models: implications for statistical power analysis. Front Psychol. 2018;9:294. doi:10.3389/fpsyg.2018.00294
- 26.↑
Hamilton J, Gagne PE, Hancock GR. The effect of sample size on latent growth models. Paper presented at: Meeting of the National Council on Measurement in Education; 2003; Chicago, IL.
- 27.↑
Rosseel Y. Lavaan: an R package for structural equation modeling. J Stat Softw. 2012;48(2):1–36. doi:10.18637/jss.v048.i02
- 28.↑
IBM SPSS Statistics for Windows. Version 28.0. Accessed January 2023. https://www.ibm.com/support/pages/downloading-ibm-spss-statistics-28010
- 29.↑
Kenny DA, Kaniskan B, McCoach DB. The performance of RMSEA in models with small degrees of freedom. Sociol Methods Res. 2015;44(3):486–507. doi:10.1177/0049124114543236
- 30.↑
Bentler PM, Bonett DG. Significance tests and goodness of fit in the analysis of covariance structures. Psychol Bull. 1980;88(3):588–606. doi:10.1037/0033–2909.88.3.588
- 31.↑
Hu LT, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Modeling. 1999;6(1):1–55. doi:10.1080/10705519909540118
- 32.↑
Elmore RG. The lack of racial diversity in veterinary medicine. J Am Vet Med Assoc. 2003;222(1):24–26. doi:10.2460/javma.2003.222.24
