Accuracy of risk scales for predicting repeat self-harm and suicide: a multicentre, population-level cohort study using routine clinical data

Table 3 Measures of diagnostic accuracy^a for repeat self-harm with 95% confidence intervals, N = 4000, m (missing data imputation) =1–5

Scale	Thresholds	Sens %	Spec %	PPV %	NPV %	LR+	LR-	DOR
MSHR	Low risk (0) vs. moderate/high risk (1+)	98 (97, 99)	15 (14, 17)	31 (30, 33)	95 (93, 97)	1.155 (1.154, 1.156)	0.13 (0.12, 0.15)	8.6 (5.6, 13.2)
ReACT	Low risk (0) vs. moderate/high risk (1+)	94 (93, 96)	23 (22, 25)	33 (31, 34)	91 (89, 93)	1.23 (1.228, 1.231)	0.240 (0.230, 0.250)	5.1 (3.9, 6.7)
SPS	Low risk (0–4) vs. moderate risk (5–6)	24 (22, 27)	76 (74, 78)	28 (25, 31)	72 (71, 74)	1.01 (0.98, 1.04)	0.996 (0.993, 0.999	1.0 (0.9, 1.2)
SPS	Moderate risk (5–6) vs. high risk (7–10)	29 (24, 34)	77 (74, 79)	34 (29, 40)	72 (69, 75)	1.23 (1.16, 1.30)	0.931 (0.923, 0.939)	1.3 (1.0, 1.7)
MSPS	Low risk (0–5) vs. moderate risk (6–8)	9 (7, 11)	90 (89, 91)	26 (22, 31)	72 (70, 73)	0.9 (0.7, 1.1)	1.01 (1.008, 1.012)	0.9 (0.7, 1.1)
MSPS	Moderate risk (6–8) vs. high risk (> 8)	12 (7, 20)	90 (86, 93)	32 (19, 48)	74 (69, 78)	1.3 (0.4, 3.6)	0.97 (0.95, 0.99)	1.3 (0.7, 2.6)

^aSens sensitivity, Spec specificity, PPV positive predictive value, NPV negative predictive value, LR+ positive likelihood ratio, LR- negative likelihood ratio, DOR diagnostic odds ratio

ISSN: 1471-244X