Canada-UK AI Study

PURPOSE:

The study team aims to develop and evaluate an artificial intelligence-enhanced socially intelligent robot designed to alleviate children's distress and pain in a clinical setting.

HYPOTHESIS:

The study team believes that interaction with a robust, adaptive, socially intelligent robot can effectively distract children during painful clinical procedures, thereby reducing pain and distress.

JUSTIFICATION:

A recent review examined eight studies where a robot was used to reduce children's pain and distress: overall, while the results seem promising, there is also a need for improved methodology and measures to draw conclusions. In particular, the authors suggest more effective interventions could be created by ensuring that healthcare experts and engineers collaborate from the start, and that user and family partners contribute to a user-centred design process. Our proposed work includes such family input and research team collaboration. Previous reviews have also identified an urgent need to increase the autonomy of the robots to improve their robustness and adaptability. Our project will aim to address this by developing and evaluating a responsive AI-enhanced social robot.

In addition, as AI systems such as robots grow more pervasive in daily life, understanding the impact of such systems on society has become ever more crucial. For social robots, in particular, an important consideration is determining the social role that the robot should play, as well as an ethical and appropriate means of making clear the capabilities of the robot.

METHODS:

• Months 24-36; University of Alberta, University of Toronto
• The goal is to conduct a clinical trial of the robot system developed in Study Design: This randomized controlled superiority trial will consist of two arms: (1) The Standard of Care (Control) Arm will include topical anesthetic cream as well as a combination of parental support, child life services and/or healthcare provider support, at the discretion of the treating team and family. (2) The Robot (Intervention) Arm will include the addition of robot distraction to current departmental standard of care. Distraction will include interaction with a socially intelligent, autonomous humanoid robot, the development of which will be informed by the earlier stages of this study (co-design and usability phases).

Primary Objective: Our primary objective of is to compare the reduction of pain and distress with the use of distraction (via the robot intervention) versus current standard of care in children aged 5 to 11 years who are undergoing intravenous insertion (IVI).

Secondary Objectives: The secondary objectives are (a) to compare the reduction of fear in the robot intervention versus standard of care arms; (b) to compare the reduction of parental/caregiver anxiety with their child's use of robot intervention versus standard of care; (c) to compare child, parental/caregiver and nurse satisfaction with the procedure in the intervention versus standard of care arms; (d) to examine the association between parental/caregiver anxiety and child outcomes (ie. pain, distress, fear); (e) to assess children's degree of engagement with the socially intelligent, autonomous humanoid robot; and (f) to assess the ethical needs and issues related to interaction with a socially intelligent, autonomous humanoid robot.

Data Collection: Research Assistants (RAs) will screen the electronic emergency department trackboard and communicate with on-site clinical staff to identify potentially eligible patients. For children flagged as requiring an IVI the RA will further assess eligibility, and acquire informed consent/ assent as appropriate. The RA will then gather baseline demographic/ history information and access a secure Research Electronic Data Capture (REDCap) randomization tool to determine the child's group assignment (ie. either Robot intervention or Control group).

Approximately five minutes prior to the start of the procedure (defined as cleaning of the IV site), the RA will begin the video recording and collect pre-procedure pain, fear and parental anxiety measures. For children randomized to the intervention group, the RA will bring in the robot and explain how to interact with it. The clinical nurse will then insert the IV following standard institutional protocols. For children randomized to the intervention group, the RA will bring in the robot and explain how to interact with it. The clinical nurse will then insert the IV following standard institutional protocols. Immediately following the first attempt at IVI (regardless of success), the RA will repeat the pain, fear and parental anxiety measures. Within 5-10 minutes of the procedure, satisfaction and acceptability questionnaires will be completed with the child, parent/ caregiver and the clinical nurse. Video recording will conclude 5 minutes after the procedure and only one attempt will be recorded, regardless of success of the IVI. Video recording will allow for coding of OSBD-R distress scores at a later time. Visit details will be collected from the child's medical chart.

Setting and Sample: The study team plans to recruit 85-90 patients in total. Patients will be recruited from the emergency departments of the Stollery Children's Hospital (Edmonton) and SickKids (Toronto).

PLAN FOR DATA ANALYSIS (clinical data):

Statistical analyses will be conducted using statistical software Statisical Analysis Software (SAS) (version 9; SAS Institute, Cary, North Carolina). Baseline variables will be described using appropriate summary statistics. For pain during the procedure, the mean (or median, if appropriate) scores will be compared between the two groups using independent samples t-tests (normal distribution) or Mann-Whitney U-tests (skewed). For Observational Scale of Behavioral Distress-Revised (OSBD-R), a change score (during procedure minus pre-procedure) will be calculated for each child, and the mean change scores will be compared between study groups (Mann-Whitney U test). Multiple linear regression analyses will be conducted with distress as the response variable, baseline pain and group indicators as explanatory variables, and possible effect modifiers (i.e. age, sex). Intention-to-treat approach will be used. The significance level is set at 0.05.