State of the psychometric methods: patient-reported outcome measure development and refinement using item response theory

Journal of Patient-Reported Outcomes

Table 4 Model fit changes for short form selection

Model	Cronbach’s alpha	AIC	BIC	-2log likelihood	Δ in -2log likelihood	RMSEA	M₂ (df)
51 items	0.983	65,230.18	66,432.60	64,720.18	–	0.43	163,378.86 (1071)***
Short Form 1 Prioritizing Content (10 items)	0.946	13,762.99	14,234.40	13,562.99	51,157.19	0.01	1469.53 (1420)
Short Form 2 Prioritizing Precision (10 items)	0.945	13,825.12	14,296.54	13,625.12	51,095.06	0.01	1513.90 (1420)

Cronbach’s alpha = measure of internal consistency/reliability from Classical Test Theory (criterion: ≥.90).
AIC Akaike information criterion (criterion: the lower the number, the better the fit)
BIC Bayesian information criterion (criterion: the lower the number, the better the fit)
-2log likelihood = if models are nested, subtract at each step to see if step is significant
RMSEA Root mean square error of approximation (criterion: ≤ .05).
M₂ = model fit.
***p < .001 (Note: a significant value for model fit indicates that the model does NOT fit well)
df Degrees of freedom.