Systematic Review of Computer Vision Performance for Age-Friendly Healthcare Facility Auditing: A Task-Based Evidence Synthesis

Abstract

Population aging increases the need for healthcare environments that are safe, accessible, and supportive of mobility for older adults. Facility auditing remains an important part of quality improvement, but manual inspection depends on staff time, checklist interpretation, and the moment of observation. This systematic review synthesizes task-based evidence on computer vision for age-friendly healthcare facility auditing. The review followed PRISMA 2020 and PRISMA-S principles. Eligible studies used image, video, RGB-D, depth, infrared, or thermal computer vision in hospitals, intensive care units, wards, nursing homes, long-term care, assisted living, or closely related care environments. Studies were included if they reported task-level performance or demonstrated clear operational relevance to environmental safety, accessibility, hygiene, mobility, posture, falls, or indoor object recognition. The evidence was synthesized narratively because the studies differed widely in their settings, sensor types, target labels, datasets, and metrics. The evidence clustered into environment understanding, hygiene and contact monitoring, mobility and activity monitoring, and posture or fall-related analysis. Technical performance was often promising within local datasets, but few studies translated model outputs into audit indicators, facility dashboards, or administrative decisions. Future work should align computer vision outputs with age-friendly audit tools, validate models across facilities, report privacy, workflow fit, and operational usefulness.