Development and validation of a machine learning-based fall-related injury risk prediction model using nationwide claims database in Korean community-dwelling older population

Table 2 Performance comparison of each model on validation cohort

Models	No. (%) of high-risk patients	No. of FRI^a	Accuracy	Sensitivity	Specificity	PPV	NPV	AUROC	p-value^b
Catboost	189,445 (34.3)	3543	0.652	0.647	0.652	0.019	0.995	0.700	< 0.001
LightGBM	202,745 (36.7)	3679	0.636	0.662	0.636	0.018	0.994	0.700	0.004
XGBoost	193,956 (35.1)	3587	0.650	0.648	0.650	0.018	0.995	0.699	0.005
Random Forest	175,115 (31.2)	3383	0.660	0.637	0.660	0.019	0.994	0.699	0.380
Logistic regression	219,624 (39.7)	3839	0.603	0.691	0.602	0.017	0.995	0.698	-

FRIs fall-related injury, PPV positive predictive value, NPV negative predictive value, AUROC area under the receiver operating characteristic curve
^aCounted based on fall occurring within 3 months from the entry date. The total number of patients who experienced fall within 3 months was 5555
^bP-value for comparison of area under the receiver operating characteristic curve with logistic regression

ISSN: 1471-2318