JMIR Publications

ABSTRACT

Background:

The frailty syndrome is one of the most common non-communicable diseases, and it is associated to lower physical and mental capacities in older adults. The frailty diagnosis is mostly focused on biological variables; however, it is very likely that this diagnosis could fail, due to the high biological variability in this syndrome. Therefore, artificial intelligence (AI) could be a potential strategy to identify and diagnose this complex and multifactorial geriatric syndrome.

Objective:

To analyse the existing scientific evidence on the use of AI for the diagnosis or detection of the frailty syndrome in older adults, as well as identifying which model provides an enhanced accuracy, sensitivity, specificity, and area under the curve.

Methods:

A search was conducted using PRISMA protocol on different databases: PubMed, Web of Science, Scopus and Google Scholar. The search strategy was conducted following PICO criteria. The studies selected met the defined inclusion and exclusion criteria, and those contained information on diagnosis or detection of the frailty syndrome in older adults through any type of AI.

Results:

A total of 926 studies were identified from the 4 databases. After pre-processing and screening, only 26 studies were included in this review. All studies were aimed at diagnosing and detecting frailty syndrome. Machine learning was the most widely used type of AI, included in 18 studies. Moreover, of the 26 included studies, 9 studies used clinical data, being the clinical histories the most used type of data in this section. The remaining 17 studies used non-clinical data, with activity monitoring using an inertial sensor in a clinical and non-clinical context were the most used data in this section. Regarding the performance of each AI model, 10 studies achieved a value of precision, sensitivity, specificity or area under the curve ≥90.

Conclusions:

The findings of this exploratory review reports the overall status of recent studies using AI to diagnose frailty syndrome. Moreover, findings show that the combined use through AI of clinical data, along with non-clinical information, such as the kinematics of inertial sensors that monitor activities in a non-clinical context, could be a correct tool for the diagnosis of frailty syndrome.