Predicting resistance as indicator for need to switch from first-line antiretroviral therapy among patients with elevated viral loads: development of a risk score algorithm

Table 3 Performance of models and derived risk scores

Predictor	Model with baseline VL (n = 290)			Model without baseline VL (n = 290)			Model without baseline VL or CD4 (n = 260)
Predictor	Cutoff	Sensitivity	Specificity	Cutoff	Sensitivity	Specificity	Cutoff	Sensitivity	Specificity
Unrestricted (RLS & non-RLS)
Model^a	0.657	22.7 %	98.1 %	0.640	22.7 %	97.2 %	0.741	13.4 %	98.4 %
Weighted risk score	≥9	28.0 %	96.7 %	≥9	16.0 %	98.6 %	≥9	14.7 %	98.1 %
Restricted (RLS only)
Model^a	0.653	28.0 %	97.2 %	0.697	18.7 %	98.1 %	0.691	14.9 %	98.4 %
Weighted risk score	≥9	26.0 %	97.4 %	≥9	10.0 %	99.5 %	≥9	14.0 %	99.5 %

^aCutoff values for the models are thresholds derived by summing the beta coefficients and converting to a probability RLS, resource-limited setting; VL, viral load

ISSN: 1471-2334