Table 2.

Predictive accuracy to forecast major depression at 6, 12, and 18 months among primary care patients with diabetes recruited in the Diabetes–Depression Care-Management Adoption Trial study.

Predicted outcome	Algorithm to predict log odds risk score of major depression	AUROCa	Brier score	Cutoffb	Sensitivity (%)	Specificity (%)
6-month depression	100×exp(–4.58 + 0.55×PHQ2c + 0.13×SDSd + 0.80×PBDe + 0.80×DSSf + 0.09×DEBg)	0.80	0.07	22; 14; 8; 5	49; 70; 86; 90	88; 79; 64; 45
12-month depression	100×exp(–4.83 + 0.54×PHQ2 + 0.21×SDS + 1.26×PBD + 0.80×DSS + 0.05×DEB)	0.81	0.06	22; 14; 8; 5	56; 67; 75; 86	90; 84; 71; 56
18-month depression	100×exp(–4.53 + 0.46×PHQ2 + 0.16×SDS + 1.23×PBD + 0.62×DSS + 0.20×DEB)	0.83	0.07	22; 14; 8; 5	67; 76; 90; 91	87; 78; 64; 47

^aAUROC: area under receiver operating curve.

^bThe cutoff points 22, 14, 8, and 5 correspond to 100×exp(–1.5), 100×exp(–2.0), 100×exp(–2.5), and 100×exp(–3.0), respectively.

^cPHQ-2: Patient Health Questionnaire - 2-item.

^dSDS: Sheehan Disability Scale.

^ePBD: ever had a problem with depression.

^fDSS: Diabetes Symptoms Scale.

^gDEB: diabetes emotional burden.