Predicting admission for fall-related injuries in older adults using artificial intelligence: A proof-of-concept...

Area of Science:

• Geriatric Medicine
• Artificial Intelligence in Healthcare
• Trauma Research

Background:

• Pre-injury frailty is a known predictor of outcomes in older trauma patients.
• Identifying older adults at risk for fall-related hospital admissions is crucial for preventative care.

Purpose of the Study:

• To utilize machine learning to identify a "signature" or combination of clinical variables predicting fall-related hospital admissions in older adults.
• To hypothesize that the 5-item modified Frailty Index, combined with other factors, can predict these admissions.

Main Methods:

• Analysis of the National Readmission Database (2010-2014) for trauma admissions in older adults.
• Inclusion of variables such as age, sex, chronic conditions, prior fall admissions, comorbidities, 5-item modified Frailty Index, and insurance status.
• Application of logistic regression and random forest machine learning models, with decision trees used to extract high-risk factor combinations.

Main Results:

• Development of 18 distinct predictive models.
• Being female was frequently associated with fall-related admissions.
• A combination of being female, aged 65-74 years, and having a 5-item modified Frailty Index score >3 predicted admission for fall-related injuries in 80.3% of the study population.

Conclusions:

• Machine learning successfully generated 18 "signatures" to identify older adults at risk for fall-related hospital admissions.
• Validation of these high-risk combination models in other databases, such as TQIP, is recommended for future research.