Video Abstract
Social robots (SRs) are increasingly present in medical and educational contexts, but their use in inpatient pediatric settings has not been demonstrated in studies. In this study, we aimed to (1) describe the introduction of SR technology into the pediatric inpatient setting through an innovative partnership among a pediatric teaching hospital, robotics development, and computational behavioral science laboratories and (2) present feasibility and acceptability data.
Fifty-four children ages 3 to 10 years were randomly exposed to 1 of 3 interventions: (1) interactive SR teddy bear; (2) tablet-based avatar version of the bear; or (3) plush teddy bear with human presence. We monitored intervention enrollment and completion patterns, obtained qualitative feedback on acceptability of SR use from child life–specialist stakeholders, and assessed children’s positive and negative affect, anxiety, and pain intensity pre- and postintervention.
The intervention was well received and appeared feasible, with 93% of those enrolled completing the study (with 80% complete parent data). Children exposed to the SR reported more positive affect relative to those who received a plush animal. SR interactions were characterized by greater levels of joyfulness and agreeableness than comparison interventions. Child life specialist stakeholders reported numerous potential benefits of SR technology in the pediatric setting.
The SR appears to be an engaging tool that may provide new ways to address the emotional needs of hospitalized children, potentially increasing access to emotionally targeted interventions. Rigorous development and validation of SR technology in pediatrics could ultimately lead to scalable and cost-effective tools to improve the patient care experience.
Comments
Analysis questions on Social Robots for Hospitalized Children
Dear Editors,
I have read with interest the study, Social Robots for Hospitalized Children, by Logan et al. published June 26, 2019. The team present an interesting feasibility study, and the findings suggest that fully automated, social robots may ultimately hold promise as one way to ease the distress chronically ill children face in the in-patient setting. The authors are appropriately transparent about the limitations of the technology and about the execution of large-scale effectiveness studies in this area. I do have concerns, however, about the execution and presentation of the statistical analyses.
In figure 2 to compare total intervention time by the three intervention groups (i.e., social robot (SR), avatar, or plush toy) and in table 4 to compare self-reported outcomes by intervention group, the authors appropriately use ANOVA to compare the between group differences. The results of those analyses stated that only two significant between group differences were found: one for the SR versus plush condition for total intervention time, and one for the SR versus plush condition for child-reported, positive affect. The authors clearly state “No other statistically significant group differences were found on self-report.” In the discussion the authors state “Overall, preliminary findings suggest that hospitalized children benefit from SR technology, as evidenced by increases in reported positive affect after exposure to the robot relative to the other interventions,” yet that was not how the results were presented.
Additionally and more importantly, in all of the analyses of each outcome assessed from the video footage, between-group comparisons were not presented. Instead, the authors present a test for linear trend across the group means. The authors do not state why a linear trend in the hypothesized direction would be appropriate to assume, and when looking at the figures for those analyses (figures 3-5), it appears that the significant linear trend is driven by the group mean for the plush condition being different from the other two conditions. Specifically, it appears on visual inspection of those figures that the group means for the SR versus avatar condition are equivalent. It is unclear to me why the authors did not present the results of the between-group contrasts.
The authors’ work is exciting and promising. However, I would like to see better clarity regarding if there were any significant differences in the study outcomes between the SR and avatar condition. Else, the paper as written runs the risk of overstating the study findings.
Regards,
Jennifer A. Emond, PhD, MS
Assistant Professor of Biomedical Data Science and Pediatrics
Geisel School of Medicine at Dartmouth College
[email protected]
REF: Logan DE, Breazeal C, Goodwin MS, et al. Social Robots for Hospitalized Children. Pediatrics. 2019.