Adaptation and validation of the Hungarian version of Thyroid-Related Patient-Reported Outcome-39 (ThyPro-39) questionnaire: testing factor structure, known-group validity with the comparison of quality of life in Hashimoto’s thyroiditis and Graves’ disease

Journal of Patient-Reported Outcomes

Table 4 Model fit coefficients of the CFA models

Models	χ²(df) [Scaling correction]	RMSEA [90% CI]	CFI	TLI	SRMR	ssaBIC
Full scale:
Model 1: Twelve-factor model	879.1 (599) [1.0440]	0.040 [0.034-0.046]	0.944	0.934	0.052	31,118
Model 2a: Bifactor model: one general and twelve specific factors#	1009.1 (628) [1.0536]	0.046 [0.040-0.051]	0.924	0.915	0.069	31,190
Model 2b Bifactor model: one general and twelve specific factors# with one non-specific item	1109.1 (665) [1.0520]	0.048 [0.043-0.053]	0.916	0.906	0.068	31,918
Model 3a: Bifactor model: two general and twelve specific factors##	923.8 (629) [1.0556]	0.040 [0.034-0.046]	0.941	0.934	0.072	31,100
Model 3b: Bifactor model: two general and twelve specific factors## with one non-specific item	1038.5 (666) [1.0536]	0.044 [0.039-0.049]	0.929	0.921	0.074	31,843
Psychosocial factor:
Model 4: Seven-factor model	336.6 (168) [1.0488]	0.059 [0.050-0.068]	0.948	0.935	0.047	15,716
Model 5a: Bifactor model: one general factor and seven specific factors	327.8 (169) [1.0746]	0.057 [0.048-0.066]	0.951	0.939	0.055	15,713
Model 5b: Bifactor model: one general factor and seven specific factors with one non-specific item	377.1 (189) [1.0720]	0.058 [0.050-0.067]	0.945	0.933	0.055	16,470
Somatic symptom factor:
Model 6: 5-factor model	182.6 (109) [1.0835]	0.048 [0.036-0.060]	0.947	0.934	0.056	15,497
Model 7: Bifactor model: one general factor and five specific factors	129.3 (103) [1.0957]	0.030 [0.007-0.045]	0.981	0.975	0.042	15,455

Note: RMSEA = root mean square error of approximation. CFI = comparative fit index. TLI = Tucker and Lewis index of fit. SRMR = standardized root mean squared residual. #: The residual variance of TQ7c item is constrained in zero to avoid a negative residual. ##: The residual variances of TQ7c and TQ1c items are constrained in zero to avoid negative residuals. The negative residuals may indicate identification problems with these models
Comparison of Models 4 and 5a resulted Δχ2 = 0.14, Δdf = 1, p = .706, it indicates that there is no statistical difference in the degree of fits, however sample-size adjusted BIC value is lower in the bifactor model favoring the bifactor measurement model
Comparison of Models 6 and 7 resulted Δχ2 = 64.3, Δdf = 6, p = < 0.001 in favor for Model 7, and in addition the sample-size adjusted BIC value is also lower in this latter model