Who let the dogs in?

The Effect of Therapy Dogs versus Art Therapy on Stress Reduction in ED Staff

Kline JA, VanRyzin K, Davis JC, Parra JA, Todd ML, Shaw LL, et al. Randomized trial of therapy dogs versus deliberative coloring (art therapy) to reduce stress in emergency medicine providers: Stress reduction by dogs. Acad Emerg Med [Internet]. 2020;27(4):266–75.

Why is this paper important?

Burnout amongst clinical staff in Emergency Medicine is common and appears to have a higher incidence than in other specialities.

55-70% of emergency physicians, nurses, and residents are at risk of leaving their respective professions due to high levels of burnout, while the rate of burnout in other specialities is around 45-55%.

Research suggests that the happiness and well-being of healthcare providers are closely linked to higher levels of patient satisfaction. In turn, this highlights how the mental and emotional health of providers can directly influence the quality of care delivered.

Therefore, strategies that promote well-being and help reduce burnout deserve careful consideration.

Evidence suggests that interacting with animals can help reduce stress levels. Similarly, mindfulness activities, such as deliberate colouring, may also have the potential to lower stress by promoting relaxation and focus.

So, which is more effective in reducing stress in emergency care providers: art therapy or dog therapy? Let’s find out.

This was a single-centre, prospective, randomised controlled trial, testing the effectiveness of therapy dogs versus colouring a mandala versus no intervention on provider stress levels.

The authors wondered if a 5-minute interaction with a therapy dog while on shift in the ED would lower provider-perceived and manifested stress levels compared with voluntary, deliberate colouring of mandalas.

What was the population?

127 emergency care providers consisting of nurses and doctors working in a single centre ED (Lois and Sydney Eskenazi Hospital, which had 108,000 annual attendances in 2018). The study ran from May 2018 to August 2019.

5 of the 127 withdrew (due to feeling the shift was too busy to participate). Of the 122 subjects, 48% were resident physicians, 30% were nurses, and 23% were attending physicians.

Exclusion criteria were dislike, allergy, fear, or other reasons to not interact with a therapy dog, all as stated by the provider.

What was the intervention?

Two interventions were tested, dog therapy and colouring of a mandala. These occurred approximately at the midpoint of the provider’s shift.

The dog therapy sessions involved interaction with a therapy dog, which participants could pet or touch if they wished. In the colouring group, participants were provided with a choice of three mandalas and a full set of colouring pencils. Both activities were conducted in a quiet room, physically separated from the clinical environment, with no electronic devices, telephones, windows, or speakers present

What was the control/comparison?

A convenience sample of providers who were not offered either of the above activities was used as a control or comparison group.

What outcomes were assessed?

There were two (!) primary outcomes or ‘co-primary’ outcomes, as the authors put it. The first was self-assessment of stress using a VAS. The second was a 10-item validated perceived stress score, modified to reflect ‘the past several hours’ rather than months, as the scale was initially intended. Salivary cortisol concentration was also measured as a marker of ‘stress’.

Measurements were taken at the beginning of a shift (T1), approximately 30 minutes post-intervention (T2), and near the end of a shift (T3).

The secondary outcome was the comparison of patient-reported scores on 10 empathic behaviours.

What did the authors find?

60% of participants were enrolled on an evening shift (commencing between 1500 and 1700). Enrolment was relatively evenly distributed throughout the week, with the fewest on Fridays (n=12, 10%) and the most on Tuesdays (n=26, 21%).

In the dog group, providers spent a median of 5 minutes 49 seconds with the dog.

In the colouring group, providers spent a median of 5 minutes 26 seconds with their pencils.

Self-reported stress, as measured by the VAS, appeared to decrease in the dog therapy group, while it increased in the colouring and control groups.

Stress as measured by the modified Perceived Stress Scale (mPSS) or ‘Emergency care stress scale’ as the authors have labelled it, appeared to increase in the colouring and control groups throughout the shift. In the dog therapy group, there was a slight decrease in stress, overall. Figure 2 (below) illustrates this.

The only statistically significant finding regarding this ‘co-primary outcome’, however, was that resident mean VAS was higher than nursing VAS at T1 (p=0.02)

Salivary cortisol concentrations showed a greater decrease in both intervention groups compared with the control group (p = 0.02). By T3, the mean salivary cortisol levels were statistically significantly different between the groups (p < 0.001).

Colouring therapy was more effective than the control (p < 0.001), and dog therapy was more effective than the control (p = 0.003) in terms of reduced salivary cortisol concentration levels.

Interestingly, the difference at T3 for colouring versus dogs was non-significant (p>0.9).

There was no significant difference between groups for the secondary outcome.

So, the results could be described as mixed, but how good was the paper? Let’s consider the CASP checklist for RCTs…

Did the study address a clearly formulated research question?

Yes. The authors employed a clear PICO structure, as outlined earlier in this blog post.

Was the assignment of participants to interventions randomised?

Yes, all participants in the interventions were randomly assigned by a pre-printed random sequence to receive either exposure to a therapy dog or to colour a mandala. So, in theory, the allocation sequence was concealed from investigators and participants. This is also borne out by participants being surprised or disappointed at their intervention allocation.

Were all participants who entered the study accounted for at its conclusion?

Can’t tell. Examining Table 1 from the paper, the total number of participants (121) does not match the number listed (122). It is unclear whether this difference would have affected the significance of any results; nonetheless, it cannot be accounted for.

Was there adequate blinding in the study?

No. In a study of this nature, achieving blinding is virtually impossible. This limitation makes interpreting subjective stress scores more challenging. For example, the lower stress scores observed in the dog therapy group may reflect participants’ innate preferences for dogs rather than a true reduction in stress levels.

Were the study groups similar at the start of the RCT?

It is difficult to say. As shown in Table 1, the demographics of the study groups were broadly similar in number but not precisely matched. Only role, gender, and race were considered, which could have introduced selection bias by failing to account for other confounding variables that might influence outcomes. For example, childhood socio-economic status could affect prior exposure to or ownership of dogs, and therefore influence participants’ preferences and perceived stress reduction after contact with therapy dogs.

Similarly, there may have been inherent baseline differences between groups that affected outcomes. For instance, if attending physicians were more likely to be dog owners, their stress responses to therapy dog interaction might differ from those of other staff groups.

Apart from the experimental intervention, did each study group receive the same level of care (that is, were they treated equally)?

No. As these interventions occurred during a clinical shift, there was no way to control the exposures of each study group, which could have also affected the outcome. Some shifts are just more stressful than others.

The Emergency Department can often be chaotic by its very nature, and scheduling an intervention may cause undue stress. This possibility was not accounted for.  Known variables that could have affected outcomes, such as caffeine consumption, were not controlled or recorded.

Were the effects of the intervention reported comprehensively?

No. While several statistical tests are mentioned, no power calculation is provided. Sample sizes were chosen based on assumption rather than calculation. P-values are listed throughout, however.

Outcomes were clearly specified, but the approach is unusual, with two ‘co-primary’ outcomes. This can increase the likelihood of a significant finding occurring due to chance, so it is generally not considered best practice.

Was the precision of the estimate of the intervention or treatment effect reported?

No. Confidence intervals are not reported for the co-primary outcomes, only for the secondary outcome.

Do the benefits of the experimental intervention outweigh the harms and costs?

Can’t tell. Although no cost-effectiveness analysis was undertaken, providing colouring pencils and mandalas as a form of stress relief is a low-cost intervention. The host hospital for the study had an existing animal therapy department, and all dogs and handlers were therapy certified and registered with a credentialed organisation. No volunteers were paid (human nor animal!). Most hospitals would not have such a resource available to them.

Conversely, considering harm, while self-reported allergy was an exclusion criterion, there remains the possibility that this was deliberately under-reported by providers in the hope of having an opportunity to interact with a therapy dog. This could have resulted in increased stress secondary to allergic symptoms experienced by the provider during the shift.

Can the results be applied to your local population/in your context?

No. External validity is suboptimal. Whilst the ED in the study has a similar number of attendances to mine and is part of a developed Western healthcare system, the study participants are pretty different from those who work in my area, especially in terms of the ‘race’ of the nursing group. This may lead to different results if the study were to be conducted locally, as the effects of race on stress levels in response to the interventions have not been explored.

The internal validity of the study (as alluded to when discussing methodology in the points above) is limited, making me even more sceptical about the accuracy of the results and their applicability to my workforce or workplace.

Furthermore, the co-primary outcomes yielded different results. However, due to flaws in methodology, I feel that one cannot infer too much from them!

Would the experimental intervention provide greater value to the people in your care than any of the existing interventions?

Can’t tell. While dog therapy and art therapy may be promising avenues for stress reduction among ED healthcare providers, more robust research is needed before any conclusions can be drawn about their effectiveness.

I have recently introduced a ‘book-swap’ to my place of work, so perhaps designing a study to compare it against art +/- dog therapy for stress reduction could be an area to explore for the future!

References

Li H, Cheng B, Zhu XP. Quantification of burnout in emergency nurses: A systematic review and meta-analysis. Int Emerg Nurs [Internet]. 2018;39:46–54. Available from: http://dx.doi.org/10.1016/j.ienj.2017.12.005

Soltanifar A, Pishbin E, Attaran Mashhadi N, Najaf Najafi M, Siahtir M. Burnout among female emergency medicine physicians: A nationwide study: BURNOUT IN FEMALE EMERGENCY PHYSICIANS. Emerg Med Australas [Internet]. 2018;30(4):517–22. Available from: http://dx.doi.org/10.1111/1742-6723.12941

Takayesu K, Ramoska J, Clark EA. Factors associated with burnout during emergency medicine residency. Acad Emerg Med. 2014;21:1031–5.

Bragard I, Dupuis G, Fleet R. Quality of work life, burnout, and stress in emergency department physicians: a qualitative review: A qualitative review. Eur J Emerg Med [Internet]. 2015;22(4):227–34. Available from: http://dx.doi.org/10.1097/MEJ.0000000000000194

Shanafelt TD, Boone S, Tan L, Dyrbye LN, Sotile W, Satele D, et al. Burnout and satisfaction with work-life balance among US physicians relative to the general US population. Arch Intern Med [Internet]. 2012;172(18):1377–85. Available from: http://dx.doi.org/10.1001/archinternmed.2012.3199

Lu DW, Dresden S, McCloskey C, Branzetti J, Gisondi MA. Impact of burnout on self-reported patient care among emergency physicians. West J Emerg Med [Internet]. 2015;16(7):996–1001. Available from: http://dx.doi.org/10.5811/westjem.2015.9.27945

Bellolio MF, Cabrera D, Sadosty AT, Hess EP, Campbell RL, Lohse CM, et al. Compassion fatigue is similar in emergency medicine residents compared to other medical and surgical specialties. West J Emerg Med [Internet]. 2014;15(6):629–35. Available from: http://dx.doi.org/10.5811/westjem.2014.5.21624

Panagioti M, Geraghty K, Johnson J. Association between physician burnout and patient safety, professionalism, and patient satisfaction: a systematic review and metaanalysis. JAMA Intern Med. 2018;178:1317–30.