Available via license: CC BY-NC 4.0
Content may be subject to copyright.
1
High rates of stress1,2 and burnout (a syndrome of
emotional exhaustion3)4,5 are robustly demon-
strated in persons working in veterinary medicine.
Although large-scale eorts to identify and under-
stand occupational distress abound,6–10 there is a
paucity of rigorously tested interventions address-
ing stressors in the eld.11–13 Recent research sug-
gests that clients who are themselves distressed in
Self-paced acceptance and commitment training
reduces burden transfer, stress, and burnout
in veterinary healthcare teams
Mary Beth Spitznagel, PhD1*; John T. Martin, MA1; Christopher Was, PhD1; Alanna S. G. Updegra, PhD1;
Meg Sislak, DVM, DACVR2; Lisa Wiborg, LISW-S, VSW3; Christopher M. Fulkerson, MS, DVM, DAVCIM4,5;
Michael P. Twohig, PhD6
1Department of Psychological Sciences, Kent State University, Kent, OH
2MedVet, Akron, OH
3Healing Paws LLC, Cleveland, OH
4Department of Veterinary Clinical Sciences, College of Veterinary Medicine, Purdue University, West Lafayette, IN
5Purdue Institute for Cancer Research, West Lafayette, IN
6Department of Psychology, College of Education and Human Services, Utah State University, Logan, UT
*Corresponding author: Dr. Spitznagel (mspitzna@kent.edu)
Received January 23, 2023
Accepted February 24, 2023
doi.org/10.2460/javma.23.01.0037
the context of their pet’s illness (ie, experiencing
“caregiver burden”)14 are likely to engage in chal-
lenging 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 dicult interac-
OBJECTIVE
An acceptance and commitment training (ACT) educational program targeting reaction to dicult client interac-
tions recently demonstrated ecacy in reducing burden transfer, stress, and burnout in veterinary healthcare teams.
The current noninferiority trial compared eectiveness of the original program with a self-paced version.
SAMPLE
Employees of 2 corporate veterinary groups were randomized to live (n = 128) or self-paced (124) conditions. The
workshop and assessments were completed by 137 (55 live and 82 self-paced).
PROCEDURES
Asynchronous modules containing the same content as the original program were placed on in-house veterinary
clinic learning systems. Participants of this parallel arms trial completed pretest measures of burden transfer, stress,
and burnout. Following assessment, the 3-week ACT program was delivered via videoconferencing (live) or asyn-
chronous modules (self-paced). At post-test and 1-month follow-up, measures were repeated, with added assess-
ment of knowledge, helpfulness ratings, and usage of techniques. A subset (n = 33) of participants repeated mea-
sures 9 to 12 months as an extended follow-up.
RE SULTS
Program helpfulness was rated more highly by live versus self-paced participants. Self-paced showed better pro-
gram retention. No dierences in knowledge or use of program techniques (> 5 times daily) emerged. Relative to
pretest, both conditions showed reduced burden transfer, stress, and burnout at post-test and follow-up; no dier-
ences by condition emerged. Participants completing extended follow-up maintained improvement from baseline.
CLINICAL RELEVANCE
Findings suggest a learning system–based version of this program can improve occupational distress in veterinary
healthcare teams, with gains maintained over time. The exibility of this format promotes program completion and
allows broader dissemination.
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
2
tions with providers in veterinary medicine, trans-
ferring some of the client’s burden to the veterinary
healthcare team.17,18 These dicult interactions, re-
ferred to as the burden transfer DANCE,18 (DANCE
being an acronym for domains of interactions: Daily
hassles, Aect, Nonadherent/inconsiderate behav-
iors, 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
as the basis of the program. This trial recruited par-
ticipants 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 accept-
ability, with 93% indicating the program was “useful”
or “very useful.” They also showed frequent use of
skills taught, averaging use of these techniques ap-
proximately 5 times daily at both post-testing and
follow-up. Signicantly reduced burden transfer reac-
tion, stress, work-related burnout, and client-related
burnout were demonstrated at post-testing and
maintained at follow-up. With ecacy of the program
demonstrated in a controlled trial, next steps must
begin to target eectiveness in the eld. To broadly
disseminate the program, it is important to under-
stand whetehr an asynchronous or self-paced ver-
sion, accessed at the individual’s convenience through
a learning platform, is as eective 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 partici-
pants via in-house learning management systems
of 2 large veterinary corporate groups. We hypothe-
sized that (1) participants assigned to the self-paced
condition would show better rates of program com-
pletion relative to the live condition; (2) participants
would show similar accuracy and retention on a test
of program knowledge, similar rates of program use-
fulness, 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. Hospi-
tals and clinics included represented a variety of set-
tings, including specialty referral, urgent care, emer-
gency, and general practice. Inclusion criteria for the
study were (1) age over 18 years and (2) ability to
speak and comprehend English suciently to com-
plete 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 Insti-
tutional Review Board of Kent State University. The
study was conducted between October 2021 and No-
vember 2022. To facilitate enrollment, the program
Table 1—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
Aect Client requires euthanasia counseling
Client demonstrates anxiety
Client demonstrates sadness or grief
Nonadherent/ Client declines recommended workup
Inconsiderate 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 Client makes frequent phone calls
communication Client makes frequent email contact
While the frequency of burden transfer DANCE
interactions signicantly predicts stress and burnout
for persons working in the eld, their reaction to these
encounters, or how bothered the veterinary healthcare
team member feels by the situation, is a far more im-
portant 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 identied as a potential mechanism for neg-
ative outcomes in the eld. Crucially, this means that
burden transfer reaction is a modiable risk factor for
stress and burnout in the eld.16–19
Recently, a randomized, controlled trial exam-
ined whether an acceptance and commitment train-
ing (ACT) program, developed to reduce this reaction,
would reduce stress and burnout in veterinary person-
nel relative to a wait-list control.18 ACT incorporates 6
interrelated elements: being present (ie, consciously
experiencing internal and external events as they oc-
cur), 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 in-
dividual 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
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
3
was oered through the in-house learning systems
of each group (one group using Dayforce Learn-
ing [Ceridian], the other using Docebo [Docebo
Inc]). To ensure similarity across sites, research-
ers were involved in the setup and testing of the
program’s functionality within each system. The
program was oered as an educational training
opportunity, and Registry of Approved Continuing
Education credits were made available. Veterinary
healthcare team members who enrolled in the pro-
gram were presented with the option to also partici-
pate 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 in-
formed consent and completion of baseline pretest
questionnaires (Qualtrics XM; Qualtrics), autoran-
domization 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 ef-
fects at an extended time point, additional funding
was obtained to invite the rst 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 synchro-
nously 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 oered, live condition sessions were made avail-
able 5 times throughout the day on dates identied
prior to enrollment. Participants were aware of ses-
sion availability at the time of enrollment; although
asked to commit to a specic time, they were al-
lowed to attend at a dierent time as needed. Mod-
ules for the self-paced condition were asynchro-
nously 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 con-
tent from live sessions during week 2 and were as-
signed 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 condi-
tion (ie, approx 3 hours). Prior to use, all 7 self-paced
modules were checked for delity to the ACT proto-
col by a separate member of the team (ASGU), who
also checked 20% of the live program sessions. For all
sessions, 100% delity to the protocol was achieved.
Pretest measures began with informed consent
describing the study purpose, rights as a study par-
ticipant, information related to institutional review
board approval, and contact information for the prin-
cipal 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 com-
pleted 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 de-
tail17,18; the program will thus be briey summarized
here. The program utilized here combines experiential
exercises and specic skills for behavior change, en-
couraging a focus on reactivity to DANCE interactions.
Techniques to enhance resilience in the context of chal-
lenging thoughts and feelings from dicult client in-
teractions are modeled and practiced. Participants are
asked to identify personal values and create individual-
ized plans for concrete, values-guided action. “Home-
work” to practice these skills between sessions is as-
signed. 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 identication of DANCE encounters (ie, potential
“points of burden transfer”). Week 2 (live condition
session 2; self-paced condition modules 3 and 4) fo-
cuses on accepting and unhooking (ie, detaching) from
dicult thoughts, feelings, and urges due to DANCE in-
teractions and on separating the self from this struggle
in an eort 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 con-
ducted that have been previously described17,18 will
be briey summarized here. The burden transfer in-
ventory16 (BTI) measured burden transfer, asking par-
ticipants about experiences across the 5 domains of
client DANCE interactions. Higher BTI subscale scores
reect greater frequency of (BTI frequency) or reac-
tion to (BTI reaction) DANCE encounters. This mea-
sure shows excellent internal consistency across sub-
scales 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
reect greater current stress. This measure demon-
strates 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 burn-
out, with higher CBI scores indicating greater current
burnout for each domain. This measure demonstrates
high internal consistency (α = 0.85 to 0.87).
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
4
In addition to the above previously validated mea-
sures, at post-test and follow-up time points, partici-
pants 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 num-
bers reecting greater frequency of skill use). Addition-
ally, a knowledge test (multiple choice) was developed
for fact-based components of the program; the in-
formation covered by these questions was covered in
both live and self-paced programs.
Participants also self-reported demographic
information, specically gender (multiple choice),
age (numeric slide bar), race or ethnicity (multiple
choice), nature of employment (multiple choice),
and years in the eld (numeric slide bar).
Power analysis
Based upon data from a previous trial,18 an eec-
tive 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
Eective curve reliability is calculated as the propor-
tion 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 eect size; in this case, a moderate ef-
fect size was expected. Regarding maximum likelihood
estimates for longitudinal growth curve models, a mini-
mum of 5 to 10 observations/item or parameter is rec-
ommended.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 nal sample of 137 in the present re-
port yields approximately 20 observations/parameter
to be estimated—well within recommendations.26
Statistical analyses
Statistical analyses were conducted using com-
mercially available software.27,28 First, independent
samples t tests were used to examine for any pretest
dierences between participants retained versus re-
moved from analyses. An intent-to-treat approach
was then taken for analyses. To examine whether type
of employment should be considered in analyses, re-
peated-measures analyses of variance were conduct-
ed. No dierences emerged on eects 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, demo-
graphic information and primary variables were char-
acterized using descriptive statistics (percent for cat-
egorical data, and mean, SD, and minimum-maximum
for numeric data). To explore whether barriers to
workshop completion diered signicantly between
the live and self-paced programs, 2 proportion z tests
were completed. Descriptive statistics (percent for
categorical data, and mean, SD, and minimum-max-
imum for numeric data) were used to characterize
acceptability, knowledge retention, and frequency of
technique use. Independent-sample t tests were used
to examine group dierences in program acceptabil-
ity and knowledge retention.
To examine the eect of the ACT program on pri-
mary 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 indi-
cators of growth: initial status (intercept) and change
over time (slope) on which individuals may vary. Be-
cause initial status is analogous to the intercept in a
regression equation, in each of the tested models, un-
standardized loadings of all indicators (the 3 primary
time points for each of the measured constructs) on
initial status were xed to 1. To specify linear trends,
the loading of time 1 (pretest) was xed to 0, time
2 (post-test) was xed at 1, and time 3 (follow-up)
was xed to 2. This is analogous to centering in hi-
erarchical linear modeling. To examine eects of the
ACT program, condition was included as a xed co-
variate, and the variances from the latent variables
(intercept and slope) were allowed to covary. This al-
lowed for the examination of the relationship between
levels of the dependent variable in the model and rate
of change across time. As certain t indices are inu-
enced by sample size and we hypothesize there will
be no dierences in rate of change between groups,
in the event of suboptimal model t, we included re-
peated-measures ANOVA to examine change in each
outcome over the course of the ACT program; this
also allowed examination of eect size.
Multiple imputation of missing data was ac-
complished 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 t indices: χ2 (χ2), χ2/df, root mean square
error approximation (RMSEA), the standardized root
mean square residual, comparative t index (CFI),
and the Tucker-Lewis index (TLI). Although χ2 is com-
monly reported, it is sensitive to sample size and will
more often than not return a signicant value. χ2/df
provides an alternative to χ2 as a measure of model
t, with values < 5 considered an adequate t. Root
mean square error approximation is an absolute t
index, with values < 0.08 considered adequate t and
less than 0.05 considered good t.29 Comparative t
index and TLI are incremental t indices; Bentler and
Bonett30 recommend that TLI > 0.90 is acceptable t,
whereas Hu and Bentler31 suggest that CFI and TLI >
0.95 indicate good model-data t.
Of those who were included in the above analy-
ses, 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 decreas-
es from pretest to extended follow-up. Procedures
described above for exploration of knowledge reten-
tion and frequency of technique use were repeated
within this subsample.
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
5
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 com-
pleted 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 analy-
sis. Imputed data account for no more than 7.8% of the
data in any given analysis. Independent-sample t tests
indicate no dierence between individuals retained in
analyses versus those removed due to failure to return
on scores of the pretest PSS (P = .31), CBI—work-re-
lated burnout (P = .30), CBI—client-related burnout
(P = .83), and BTI reaction (P = .83).
The nal analytic sample was 137 (live, n = 55;
self-paced, 82; Figure 1). Both groups were com-
prised primarily of females (94.9%) identifying as
Caucasian or white (92.7%), averaging approxi-
mately 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 em-
ployment in the eld ranged from 1 to 41 years, av-
eraging about 12 years. No demographic variables
diered signicantly between those assigned to self-
paced versus live conditions (Table 2).
Figure 1—Enrollment ow diagram for a study comparing eectiveness of live and self-paced versions of an accep-
tance and commitment training educational program on outcomes of burden transfer, stress, and burnout in small
animal veterinary hospital employees.
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
6
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 circumstanc-
es as the primary barrier, though a small proportion
described diculties 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 non-
completion (z = –3.01; P < .05; Table 3).
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 rat-
ed the program as “helpful” or “very helpful” com-
pared 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 maxi-
mum, 20.0% to 100.0%). Program knowledge did
not dier signicantly across conditions (P > .05),
reecting good retention of learned information, re-
gardless of whether the ACT program was complet-
ed live or self-paced. Reported frequency of tech-
nique 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). Fre-
quency of technique use did not dier signicantly
across conditions (P > .05).
Change in BTI reaction
To examine change in burden transfer reaction,
the LGC model was specied with pretest, post-test,
and follow-up as the 3 time points to determine inter-
cept and slope, with condition (live vs self-paced) as
a xed covariate. Fit indices suggest the model pro-
vided a good t 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
dier 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, partici-
pants reported less BTI reaction over time. Slope and
intercept were not associated (r = 0.08; P = .79), sug-
gesting that initial levels of BTI reaction were not as-
sociated with rate of change over time (Figure 2).
To ensure signicant change in BTI reaction over
time, we conducted a repeated-measures ANOVA
with condition as a between-subjects variable. Sig-
nicant results were observed for the main eect of
time (F[2,270] = 54.46; P < .001; partial η2 = 0.29).
The main eect of condition was not signicant
Table 2—Demographics of participants in a study comparing
eects 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.
Group
Variable 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 1 (1.8) 1 (1.2)
African American
Asian American — 1 (1.2)
or Pacic Islander
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; synchro-
nous delivery) or self-paced (82; asynchronous delivery) condition.
Unless otherwise specied, data are given as number (percentage).
Table 3—Completion and barriers to program completion for live and self-paced versions of an acceptance and commit-
ment training program addressing burden transfer, stress, and burnout among small animal veterinary hospital employees.
Group
Completion Live Self-paced P value
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 conict (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-signicant.
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
7
(F < 1), and the interaction was also not signicant
(F[2,270] = 1.31; P = .27; η2 = 0.01). Whether the
ACT program was completed live or self-paced, par-
ticipants reported lower BTI reaction over time.
Change in PSS
To examine change in perceived stress, the LGC
model was specied as described above. Fit indi-
ces suggest the model did not provide a good t
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 sensi-
tive to sample size and will more often than not re-
turn a signicant value; RMSEA tends to be inated
with low degrees of freedom. Poor model t might
also be driven by the lack of condition contributing
to variance in the intercept and slope. Conditions did
not dier 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 signicant change in PSS over time, we con-
ducted a repeated-measures ANOVA with condition
as a between-subjects variable. Signicant results
were observed for the main eect of time (F[2,270]
= 28.57; P < .001; partial η2 = 0.18). The main eect
of condition was not signicant (F[1,135] = 1.28;
P = .26; partial η2 = .01) and the interaction was also
not signicant (F < 1). Whether the ACT program
was completed live or self-paced, participants re-
ported lower PSS over time (Figure 2).
Change in CBI—work-related burnout
To examine change in work-related burnout, the
LGC model was specied as described above. Some
t indices suggest the model provided a good t 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 t indices, we are condent in the specied
model. In this model, condition did not predict slope
Figure 2—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.
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
8
(β = 0.35; P = .06), indicating groups did not dier in
change in CBI—work-related burnout over time. Condi-
tion also did not predict the intercept (β = 0.01; P = 0.92),
indicating that participants in the 2 conditions did not
dier 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 t indices did not sug-
gest adequate t, repeated-measures ANOVA was
conducted with condition as a between-subjects
variable. A signicant main eect of time (F[2,270] =
30.85; P < .001; partial η2 = 0.19) was observed. The
main eect of condition was not signicant (F[1,135]
= 1.26; P = .27; partial η2 = 0.01), and the interaction
was also not signicant (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 specied as described above.
Some t indices suggest the model provided a good
t to the data (χ2 /df = 3.90; CFI = 0.95; TLI = 0.93),
whereas others did not (χ2 [4] = 15.61; P = .004; RM-
SEA = 0.13 [90% CI, 0.06 to 0.19]). Based on the bi-
ases inherent in some t indices, we are condent in
the specied model. In this model, condition did not
predict slope (β = 0.30; P = .31), indicating groups
did not dier 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 dier 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 t indi-
ces did not suggest adequate t, repeated-measures
ANOVA was conducted with condition as a between-
subjects variable. A signicant main eect of time
(F[2,270] = 41.50; P < .001; partial η2 = 0.24) was
observed. The main eect of condition was not sig-
nicant (F[1,135] = 3.04; P = .08; partial η2 = 0.02),
and the interaction was also not signicant (F[2,270]
= 1.79; P = .17; η2 = 0.02). Whether the ACT program
was completed live or self-paced, participants re-
ported lower CBI—client-related burnout over time
(Figure 2).
Extended follow-up
For the subset of individuals who completed ex-
tended follow-up, overall program knowledge aver-
aged 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 con-
ducted across primary outcome measures within the
subset of individuals who provided extended follow-up
data. Relative to pretest, at extended follow-up, par-
ticipants 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-relat-
ed 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 reac-
tivity to DANCE interactions with veterinary clients.
Although a higher percent of participants described
the live program as “helpful” or “very helpful” com-
pared 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 dierence 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 eects ob-
served and gains being maintained 1 month later; no
dierences between groups emerged in these out-
comes. Finally, the subset of individuals examined 9
to 12 months later demonstrated continued benet
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
eld of < 50% nding 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 dierences between formats, and utiliza-
tion of techniques learned in the program was more
than 5 times/d for both groups. Importantly, both
versions of the program signicantly reduced reac-
tion to DANCE encounters, stress, and both work-
related and client-related burnout in participants.
The eect sizes of these improvements were large
and were maintained 1 month after the program
ended in the full sample. The large eects noted here
suggest that the program leads not just to statisti-
cally signicant improvements, but to meaningful
change for individuals who complete it. Overall, it
appears that the interactive component of the previ-
ously 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 benets.
Additionally, data from the subset of individuals
who completed extended follow-up measures 9 to 12
months after the initial program suggest good reten-
tion of information learned in the program, continued
daily use of skills taught, and sustained improvement
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
9
in burden transfer reaction, stress, and burnout rela-
tive to baseline. Because participants reported us-
ing 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 con-
sidering whether a booster program would further
enhance skill use and outcomes. This subsample was
small and not part of the initially planned data col-
lection; as such, future work should examine extend-
ed eects 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 indi-
viduals (4 total) reported experiencing technological
diculties preventing program completion within
the time-limited study period; however, the learn-
ing 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 ver-
sions in lockstep with each other) and with greater
learning system familiarity on the part of participat-
ing employees. In combination, our results support-
ed that an asynchronous, learning system–based
version of this ACT program could be an eective
way to share its content, making its benets readily
available to individuals working in the eld of veteri-
nary medicine. Eorts 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 pro-
gram 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 dierences were observed be-
tween individuals retained in analyses versus those
removed due to failure to return, the phenomenon
of low completion rate is not attributed to psycho-
logical state (ie, participants with higher baseline
distress did not show poorer program completion).
Rather, it is likely that this occurred secondary to
our eorts to recruit participants in an ecologically
valid manner. Specically, the program was oered
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 generaliz-
ability of results, but may have led to a lower sense
of obligation, if participants enrolled primarily for
the educational opportunity rather than as a com-
mitment 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 rst
session, it is possible that the remaining sample
comprised individuals who, in general, would be
more inclined toward adherence to program recom-
mendations. If that is the case, the current sample
may include individuals who are more likely to enact
techniques taught compared with the general pop-
ulation of persons working in veterinary medicine,
and thus more likely to show benet. Additionally, it
is noted that outcomes, including frequency of tech-
niques utilized, are limited to self-report. Although
skills taught in this program are largely cognitive
in nature, making objective observation dicult to
gauge, any construct being assessed via self-report
could be subject to biases of the individual provid-
ing that information. Another limitation is found
in low representation of individuals who are often
underrepresented in veterinary medicine. While the
current sample does reect the lack of diversity in
veterinary healthcare teams in this country in gen-
eral,32 it does not represent everyone working in the
eld. Continued work is needed to ensure this pro-
gram works across cultures and backgrounds.
In summary, the current work compared the
eectiveness of live and self-paced versions of an
ACT-based educational program designed to re-
duce reactivity to dicult client interactions oc-
curring in veterinary healthcare. Findings demon-
strated that participants found both versions of the
program helpful, retained similar amounts of infor-
mation regardless of format, and did not dier in
the degree to which they incorporated techniques
taught in the program into their daily lives. The pro-
gram signicantly improved outcomes of burden
transfer reactivity, stress, and burnout, regardless
of format. Within the full sample, gains were main-
tained 1 month later. Within a subsample, gains were
still observed up to 1 year later. Future directions
must now focus on eorts to create materials appro-
priate for broad dissemination of this program.
Acknowledgments
The project described in this article was primarily sup-
ported by the Clinical Scholars program of the Robert Wood
Johnson Foundation, with additional funding (extended fol-
low-up data collection) from the Applied Psychology Center
at Kent State University. The funding sources had no involve-
ment in the study design, data analysis and interpretation, or
writing and publication of the manuscript.
The authors declare that there were no conicts 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 assis-
tance of Gabrielle Heminger, BS; Andrea Kohut, RVT; and Kar-
lee Patrick, BS, for their invaluable assistance in creating the
asynchronous version of the program.
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
10
References
1. Bartram DJ, Baldwin DS. Veterinary surgeons and suicide:
inuences, 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 sui-
cide, 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
3. Maslach C. Burnout—The Cost of Caring. Prentice-Hall
International; 1982.
4. Ouedraogo FB, Lefebvre SL, Hansen CR, Brorsen BW. Com-
passion 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. Investiga-
tion 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 behav-
iour 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 inuences 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 con-
ict and moral distress in veterinary practice: a sur-
vey 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 Unit-
ed 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 sum-
mary of the Merck Animal Health Veterinarian Wellbe-
ing Study III and Veterinary Support Sta 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 veterinar-
ians’ 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 com-
passion 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 ani-
mal: 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 cli-
ent behavior. J Am Vet Med Assoc. 2019;254(1):133–144.
doi:10.2460/javma.254.1.133
17. Spitznagel MB, Updegra 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, Updegra ASG, Was C, et al. An accep-
tance 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. Vali-
dation of the Burden Transfer Inventory-abbreviated and
examination across veterinary medicine positions and set-
tings 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 Com-
mitment 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. Fac-
tor 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 as-
sessment 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 eect 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 eect of sample
size on latent growth models. Paper presented at: Meet-
ing 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 Jan-
uary 2023. https://www.ibm.com/support/pages/down-
loading-ibm-spss-statistics-28010
29. Kenny DA, Kaniskan B, McCoach DB. The perfor-
mance 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. Signicance tests and goodness
of t 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. Cuto criteria for t indexes in co-
variance 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
Unauthenticated | Downloaded 04/11/23 01:06 PM UTC
