Determination of cut-off and correlates of delay in treatment-seeking of febrile illness: a retrospective analysis

Table 2 Goodness-of-fit measures for various cut-offs for the delay in reporting among fever cases using hospital-based survey data (N = 350)

Goodness-of-fit parameters	Logistic models with various cut-offs for delay
Goodness-of-fit parameters	> 2 days	> 3 days	> 4 days	> 5 days
Overall model fit
Log likelihood	− 195.8208	− 230.0087	− 207.0900	−171.8360
AIC	417.6415	486.0173	440.1800	369.6721
BIC	467.7946	536.1704	490.3331	419.8252
Hosmer-Lemeshow test p-value	0.5965	0.5136	0.4436	0.2497
Model discrimination
ROC area	0.6748	0.6474	0.6626	0.6430
Somers’ D statistic	0.3495	0.2947	0.3253	0.2860
Model classification
Sensitivity	0.9639	0.6222	0.1339	0.0000
Specificity	0.1485	0.6118	0.9370	100.00
Positive predictive value (PPV)	0.7362	0.6292	0.5000	0.0000
Negative predictive value (NPV)	0.6250	0.6047	0.6969	0.7914
Correctly classified	0.7286	0.6171	0.6800	0.7914
Model calibration
Brier score	0.1876	0.2327	0.2031	0.1585
CITL	−2.63E-10	4.90E-08	4.95E-08	−7.06E-08
E/O ratio	1	1	1	1
C-slope	−1.79E-08	−9.34E-10	−4.49E-09	1.26E-08

AIC Alike information criterion, BIC Bayesian information criterion, ROC receiver operation characteristic, CITL calibration-in-the-large, E/O expected/observed, C-slope: calibration-slope

ISSN: 1471-2458