Abstract:

Objective To systematically review the fall prediction and monitor value of machine learning （ML） for older adults. Methods Databases including CNKI，SinoMed，VIP，Wanfang Data，Web of Science，PubMed，Embase and CINAHL were searched to retrieve all studies that focused on ML in predicting fall of older adults. The searching time was set from the date of database’s establishment to July 2022. After that，two reviewers independently screened literature，extracted data and assessed the risk of bias of included studies by the standard of QUADAS-2. Results A total of 15 studies were included. In most studies，neural network and random forest models were mainly being used. In addition，though a few studies focused on specific populations，such as older women with osteoporosis and older patients in acute care，the majority of studies included older adults. Based on the quality evaluation，the included studies were highly applicable and had a low risk of bias. Conclusion ML has high predictive value and application value in fall identification and fall risk prediction for older adults. Meanwhile，there are still some problems. In the future，big data analysis should still base on clinical experience，as well as pay attention to specific populations，in order to develop a deeper combination and development of artificial intelligence and nursing.

Key words: Machine learning, Falls, Older adults, Systematic review, Risk prediction