Table 3.

Metrics for each index version and the obstacle-crossing test from the model derivation sample (n = 45) for predicting fall status at 3-months post discharge. Cut off scores were determined from ROC analysis. Values above the cut off indicate high fall risk. Somer's d is for the binary variable using the indicated cut-off scores.

Test	Cut off	Sensitivity (95% CI)	Specificity (95% CI)	LR+ (95% CI)	LR– (95% CI)	AUC (95% CI)	Somers’ d
Obstacle-crossing test	Pass/fail	0.67 (0.43–0.85)	0.83 (0.63–0.95)	4.00 (1.56–10.28)	0.40 (0.21–0.75)	0.75 (0.60–0.90)	0.519, p < .001
Index A (0–20 points)	≥9.5	0.71 (0.48–0.89)	0.75 (0.53–0.90)	2.86 (1.36–6.01)	0.38 (0.19–0.78)	0.83 (0.71–0.95)	0.464, p < .001
Index B (0–15 points)	≥5.5	0.90 (0.70–0.99)	0.67 (0.45–0.84)	2.71 (1.52–4.86)	0.14 (0.04–0.55)	0.82 (0.70–0.95)	0.593, p < .001
Index C (0–16 points)	≥6.5	0.86 (0.64–0.97)	0.67 (0.45–0.84)	2.57 (1.42–4.65)	0.21 (0.07–0.63)	0.85 (0.74–0.96)	0.534, p < .001
Index D3 (0–8 points)	≥3.5	0.71 (0.48–0.89)	0.83 (0.63–0.95)	4.29 (1.68–10.91)	0.34 (0.17–0.69)	0.85 (0.73–0.96)	0.559, p < .001

AUC denotes area under the ROC curve, LR+/LR– denotes positive/negative likelihood ratio.

Higher LR+ values are better, lower LR– values are better.