Predicting quality of life in nursing home residents using facial expression analysis: a feasibility study of SOLO well-being scores

Background Quality of life (QOL) is a comprehensive evaluation of an individual's living conditions and is a crucial outcome measure in nursing homes for older adults to improve care quality. However, evidence on the accuracy of objective well-being assessment using facial expressions remains limited. SOLO (Solo Wellbeing Ltd., Israel) is a system that estimates emotional well-being scores based on facial expressions. In this study, we aimed to evaluate whether SOLO-derived well-being scores could effectively predict caregiver-rated QOL in nursing home residents. Methods This study was conducted from March to July 2025 in nursing homes in Japan. Residents who provided informed consent and were able to undergo facial recording were eligible; those without consent or with health risks related to participation were excluded. Participants were classified according to the Independence Degree of Daily Living for Older Adults with Dementia into Group 1 (normal level or Level I) and Group 2 (Levels II–IV). Objective QOL (QOL-D) was evaluated by caregivers using questionnaires. In addition, SOLO well-being scores were collected repeatedly over approximately 4 weeks preceding the caregiver-rated objective QOL assessment using facial expression analysis. Results Fifty-nine residents were included (mean age 89.49 years, SD = 6.07; 46 women and 13 men), with 14 in Group 1 and 45 in Group 2. Hierarchical multiple regression analyses demonstrated that SOLO well-being scores were a significant positive predictor of QOL-D scores. No significant interaction was observed between SOLO well-being scores and dementia-related level of independence. These associations remained after adjustment for age and sex. Conclusions Despite limitations, facial expression analysis may have potential as a low-burden supplementary approach to caregiver-rated QOL assessment in nursing home residents. Further validation in larger and more diverse samples and across different recording conditions is needed.