JMIR Publications

ABSTRACT

Background:

The multidomain structure of Intrinsic capacity (IC) is currently assessed using multiple scales across five domains (cognition, locomotion, psychological capacity, sensory capacity, and vitality), the process is resource-intensive. However how to measure IC objectively and efficiently remains an open question. The rapidly developing artificial intelligence (AI) technologies can play a new role in the measurement of IC in older adults, promoting research in this field.

Objective:

We aimed to identify and synthesize AI-based measurement tools for assessing IC in older adults, summarize the digital biomarkers involved, and highlight methodological gaps and future directions.

Methods:

Databases searched encompassed PubMed, Embase, CINAHL, PsycInfo, Cochrane Library, SinoMed, and CNKI, covering publications up to July2025 following PRISMA-ScR guidelines. MeSH terms and keywords mainly included "aged," "intrinsic capacity," "artificial intelligence", and etc.

Results:

Based on the inclusion criteria, 157 studies were included. AI-based measurement tools have expanded rapidly since 2016, with studies conducted in 28 countries, primarily the United States and China. The vast majority of studies focused on the measurement of IC along a single dimension, more than one third of which assessing cognition. AI-based measurement tools can be categorized by frequency of use, from highest to lowest, as follows: multimodal data acquisition devices, computer vision systems, wearable devices, smart homes, speech analysis systems, AI audiometers, digital pens, virtual reality systems, robots, and Fixed sensor devices. A total of 21 types of digital biomarkers were extracted, grouped into three major categories. Several key digital biomarkers were identified, including gait parameters, digital task performance, physical activity features, speech and language features, and facial features. Overlapping digital biomarkers have been identified across the cognitive, locomotor, psychological, and vitality domains. Machine learning and deep learning were the predominant AI techniques used.

Conclusions:

AI-based measurement holds great promise for scalable IC assessment, but advancing standardized, explainable, and ethical frameworks remains crucial to support early and equitable ageing care.