Establishing the HLS-Q12 short version of the European Health Literacy Survey Questionnaire: latent trait analyses applying Rasch modelling and confirmatory factor analysis

Table 3 Unidimensionality, data-model fit and reliability applying Rasch modelling of the various short versions

	HLS-Q12^a		HL-SF12 [26]		HLS-EU-Q16 [11]
	One-dimensional	Three-dimensional	One-dimensional	Three-dimensional	One-dimensional	Three-dimensional
Unidimensionality
Proportion (%) of significant t-tests (lower CI-prop)	8.78% (0.07)		8.01% (0.07)		10.44% (0.09)
PSI	0.767		0.759		0.830
PSI^b	0.687		0.668		0.703
Fractal indexes		na		na		na
c	0.50		0.39		0.46
A	0.90		0.91		0.85
r	0.80		0.87		0.82
Total item chi square (df), probability	112.61 (108), 0.361		130.21 (108), 0.072		208.90 (144), 0.00034
Log-likelihoods
Deviance (ep)	18,590 (37)	18,556 (42)	18,707 (37)	18,696 (42)	22,964 (49)	22,815 (54)
AIC (ep)	18,664 (37)	18,640 (42)	18,781 (37)	18,780 (42)	23,062 (49)	22,923 (54)
Reliability
PSR (MLE)	0.762	0.537/0.517/0.575	0.758	0.501/0.497/0.564	0.825	0.713/0.619/0.545

Note. The table shows the results of the tests of unidimensionality based on paired t-tests of person-location estimates for subsets of items. It also reports fractal indexes, c, A and r. The person separation index (PSI) and fractal indexes were estimated for the complete dataset (HLS-Q12 n = 696, HL-SF12 n = 680, HLS-EU-Q47 n = 670). Significant t-tests ≤5% (or lower confidence interval [CI] proportion ≤ 5%), small drops in PSI (^b after adjusting for violations of local independence due to subtest structure), high values of A and r, and low values of c could indicate unidimensionality. Analyses were performed by using RUMM2030 software
Log-likelihoods and person separation reliability (PSR) were estimated for the one- and three-dimensional approaches to the short versions by using ConQuest4 software. Lower values of deviance and AIC indicate a better fit
AIC: Akaike’s information criterion, ep: number of estimated parameters, na: not applicable, PSR (MLE): person separation reliability based on a marginal maximum likelihood estimate
All the short versions are developed on the basis of the HLS-EU-Q47
^aHLS-Q12 developed through the present study

ISSN: 1472-6963