The clinical suitability of an Artificial Intelligence-enabled pain assessment tool for use in infants: Feasibility and usability evaluation study
Hughes, J. D.,
The clinical suitability of an Artificial Intelligence-enabled pain assessment tool for use in infants: Feasibility and usability evaluation study.
Journal of Medical Internet Research, 25.
Infants are unable to self-report their pain, which, therefore, often goes underrecognized and undertreated. Adequate assessment of pain, including procedural pain, which has short- and long-term consequences, is critical for its management. The introduction of mobile health–based (mHealth) pain assessment tools could address current challenges and is an area requiring further research.
The purpose of this study is to evaluate the accuracy and feasibility aspects of PainChek Infant and, therefore, assess its applicability in the intended setting.
By observing infants just before, during, and after immunization, we evaluated the accuracy and precision at different cutoff scores of PainChek Infant, which is a point-of-care mHealth–based solution that uses artificial intelligence to detect pain and intensity based solely on facial expression. We used receiver operator characteristic analysis to assess interpretability and establish a cutoff score. Clinician comprehensibility was evaluated using a standardized questionnaire. Other feasibility aspects were evaluated based on comparison with currently available observational pain assessment tools for use in infants with procedural pain.
Both PainChek Infant Standard and Adaptive modes demonstrated high accuracy (area under the curve 0.964 and 0.966, respectively). At a cutoff score of ≥2, accuracy and precision were 0.908 and 0.912 for Standard and 0.912 and 0.897 for Adaptive modes, respectively. Currently available data allowed evaluation of 16 of the 17 feasibility aspects, with only the cost of the outcome measurement instrument unable to be evaluated since it is yet to be determined. PainChek Infant performed well across feasibility aspects, including interpretability (cutoff score defined), ease of administration, completion time (3 seconds), and clinician comprehensibility.
This work provides information on the feasibility of using PainChek Infant in clinical practice for procedural pain assessment and monitoring, and demonstrates the accuracy and precision of the tool at the defined cutoff score.
Pain assessment, Clinical Utility, PainChek Infant, Artificial Intelligence, Machine learning, Facial assessment