Evaluating the clinical utility of an easily applicable prediction model of suicide attempts, newly developed and validated with a general community sample of adults

Table 3 Net benefit and delta net benefit results, as shown in Fig. 1

\( {p}_{\text{t}}\) %	Net benefit		Median net benefit across 100 cross-validations
	Net benefit		Logistic regression		CART
	Treat all	Treat none	Total	Delta	Total	Delta
0.00	0.0123	0	0.0123	0	0.0123	0
0.50	0.0074	0	0.0087	0.0013	0.0054	-0.0020
0.75	0.0049	0	0.0078	0.0029	0.0056	0.0007
1.00	0.0024	0	0.0060	0.0037	0.0049	0.0026
1.25	< 0	0	0.0046	–	0.0041	–
1.50	< 0	0	0.0037	–	0.0034	–
1.75	< 0	0	0.0036	–	0.0032	–
2.00	< 0	0	0.0040	–	0.0028	–

Note. CART = Classification and regression tree; the column \( {p}_{\text{t}}\) % contains the seven selected threshold probabilities; 0% was added to satisfy reporting guidelines for decision curve analysis. The \( {p}_{\text{t}}\) of 1.25% is larger than the outcome rate in the test data (0.0123 = 1.23%); yielding a negative net benefit (< 0) for treat all, as do all \( {p}_{\text{t}}\) > 1.25%. Empty cells are empty because delta net benefit is equal to net benefit, according to our delta net benefit definition. Lower net benefit results were observed 18 times (logistic regression) and 119 times (CART) across 700 cross-validations, compared to either treat all or treat none

ISSN: 1471-244X