Evidence for measurement bias of the short form health survey based on sex and metropolitan influence zone in a secondary care population

Table 2 Comparison of model fit for baseline model

	χ2 (df)	p value	RMSEA [95% CI]	CFI	TLI
Model 1: Ware et al. with correlated factors	311.71 (53)	<.001	.107 [.095–.118]^a	.790^a	.739
Model 2: model 1 + cross-loading items 1, 10, 12	198.31 (50)	<.001	.083 [.071–.096]^a	.880^a	.841
Model 3: model 2 + residual covariances for items from same SF-36 scale	161.93 (46)	<.001	.077 [.064–.090]^a	.906^a	.865
Model 4: model 3 + cross-loading items 4, 5 and residual covariance for items 9, 10	83.24 (43)	<.001	.047 [.031–.062]	.967	.950

Model 1: items 1,2,3,4,5,8 on physical health factor and items 6,7,9,10,11,12 on mental health factor
All models estimated using WLSMV
^aPoor model fit indicated by RMSEA > .05 and CFI < .95

ISSN: 1477-7525