Towards person-centered pain management in dementia: usability of a digital medical device in Portuguese residential care facilities

This is the authors' abstract. We don't add key points for this paper.

Introduction: Pain is common in people living with dementia (PLWD), yet frequently remains unrecognized and undertreated. Traditional observational tools are limited by subjectivity and suboptimal psychometric properties. The PainChek® mobile application (app) is a medical device that enables objective assessment in non-verbal people, such as those with more advanced dementia, using artificial intelligence (AI)-driven recognition of facial muscle movements. This study evaluated the usability of the Portuguese (PT) version of the PainChek® app among formal caregivers in residential care. Methods: Usability was evaluated in a controlled environment, guided by ISO 9241-11:2018 and ISO 25062:2025. After training, 19 participants performed nine tasks using the PainChek®-PT app while undertaking concurrent think-aloud; performance was screen and audio-recorded and supplemented with observation notes. Participants completed the Post-Study System Usability Questionnaire (PSSUQ-PT). Metrics included effectiveness (success rates, errors), efficiency (task completion time, attempts), satisfaction, usability problems and likely defects. Descriptive statistics were performed and confidence intervals (CI) computed, using 95% CIs for success rates and task times. PSSUQ scores were interpreted using 99% CIs against normative datasets. Results: Eight tasks achieved over 95% success rate. Task 5 (interrupted assessment of an existing resident) represented an outlier at 42% success (95% CI: 23.1-63.7). Seventeen distinct error types were recorded across tasks. Completion times varied considerably; Task 3 (assessing an existing resident's pain via video) had the longest geometric mean (335 s, 95% CI: 269.6-417.2), while log-out (Task 9) had the shortest (11 s, 95% CI: 6.6-18.4). Resident archiving (Task 7) recorded the highest number of attempts (17). The mean overall PSSUQ-PT score was 1.9 (SD: 1.06; 99% CI: 1.20-2.60), outperforming the normative mean of 2.82, with no overlap in 99% CIs. Seven usability problems were identified and linked with likely underlying usability defects. Conclusion: PainChek®-PT demonstrated effectiveness, efficiency and satisfaction in this context of use, supporting its intended use for conducting pain assessments in non-verbal PLWD. The robustness of the overall PSSUQ score indicates high perceived usability. The errors and usability problems identified in the simulated environment may be addressed through training and minor design refinements.