Table 4.

Multiple-Group Confirmatory Factor Analysis Models and Fit Indices Evaluating Measurement Invariance of the Patient Health Questionnaire-9 (PHQ-9) across Sex, Race/Ethnicity, and Education Level in U.S. Adults

		χ2	df	Δχ2	p-value	RMSEA	RMSEA 90% CI	TLI	CFI	ΔCFI	ΔRMSEA
Sex n = 31,366	Configural	1048.15	52	---	<0.001	0.035	0.033–0.037	0.985	0.989	---	---
	Scalar	877.68	84	117.93	<0.001	0.025	0.023–0.026	0.992	0.991	0.002	−0.010
	Strict	907.43	75	54.46	<0.001	0.027	0.025–0.028	0.991	0.991	0.000	−0.008
Race/Ethnicity n = 30,179	Configural	1204.45	166	---	<0.001	0.032	0.030–0.034	0.988	0.989	---	---
	Scalar	1629.07	262	411.65	<0.001	0.029	0.028–0.031	0.990	0.986	−0.003	−0.003
	Strict	1419.80	235	287.26	<0.001	0.029	0.027–0.030	0.990	0.987	0.001	−0.003
Education Level n = 31,344	Configural	1092.39	130	---	<0.001	0.034	0.032–0.036	0.985	0.989	---	---
	Scalar	1507.24	258	700.51	<0.001	0.028	0.026–0.029	0.990	0.986	−0.003	−0.006
	Strict	1088.25	222	486.91	<0.001	0.025	0.023–0.026	0.992	0.990	0.004	−0.009

Note. We used the two-factor Model 2 shown in Figure 1. For absolute fit indices (RMSEA), exact fit = 0.00, close fit = 0.01–0.05, acceptable fit = 0.05–0.08, mediocre fit = 0.08–0.10, and poor fit = greater than 0.10. For relative fit indices (TLI and CFI), exact fit = 1.00, close fit = 0.95–0.99, acceptable fit = 0.90–0.95, mediocre fit = 0.85–0.90, and poor fit = less than 0.85. For nested model testing, a change in CFI ≥ −0.004 or change in RMSEA ≥ 0.010 signifies that measurement invariance was not established for that step. RMSEA = root mean square error of approximation. CI = confidence interval. TLI = Tucker-Lewis index. CFI = comparative fit index.