. 2020 May 30;15(2):221–233. doi: 10.1177/2049463720920760

Table 5.

ROC analysis to detect the failure of DPMS according to the index of propensity score of each step of the hierarchical multiple logistic regression model (n = 117).

	AUC	95% CI	Cutoffs	Sensitivity	Specificity
Step I – (model I: analgesic use (four or more than 4 days per week) and HPT)
	0.681	(0.58–0.78)	0.2185269	0.982	0.983
			0.2283567	0.965	0.967
			0.2412153	0.965	0.950
			0.2471011	0.965	0.933
			0.2535917	0.947	0.933
			0.2612195	0.947	0.917
Step II – (model II: model I along with catastrophizing pain (cutoff-Q₂₅ = 33), BDNF (cutoff-Q₇₅ = 58.11) and PSQI (cutoff-Q₇₅ ⩾ 17))
	0.785	(0.70–0.87)	0.0564277	1.000	0.983
			0.0736317	1.000	0.967
			0.0814282	1.000	0.950
			0.0838870	0.982	0.950
			0.0892560	0.982	0.933
			0.0942823	0.982	0.917
			0.0962838	0.982	0.900
Step III – (model III: model II along with number of psychiatric diagnosis according to the MINI (major depressive episode with dysthymia, maniac-depressive disorder and generalized anxiety disorder)
	0.818	(0.74–0.89)	0.0153715	1.000	0.983
			0.0305837	1.000	0.967
			0.0388753	1.000	0.950
			0.0442709	1.000	0.933
			0.0537171	1.000	0.917
			0.0593224	1.000	0.900
			.0610924	0.982	0.900
Step IV – global index (model 4: model III along with FIQ (cutoff-Q₇₅ = ⩾78)
	0.832	(0.75–0.90)	0.0089371	1.000	0.983
			0.0139259	1.000	0.967
			0.0193599	1.000	0.950
			0.0263755	1.000	0.933
			0.0318990	1.000	0.917
			0.0349366	1.000	0.900

AUC: area under the curve; BDNF: brain-derived neurotrophic factor; CI: confidence interval; ROCs: receiver operator characteristics; FIQ: fibromyalgia impact questionnaire; PSQI: Pittsburgh sleep quality index; HTP: heat pain threshold.