Rasch analysis of the Patient Rated Elbow Evaluation questionnaire

Table 1 Summary fit statistics for individual subscales of the PREE

Analysis	Item fit residual		Person Fit residual		Item-trait interaction		Unidimensionality	PSI
Analysis	Mean	SD	Mean	SD	Chi square (df)	P	Per C < 5 % (95 % C.I)	PSI
Pain subscale
Initial	−0.01	1.02	−0.34	0.88	13.77 (15)	0.001*	5 % (2 % - 7 %)	0.87
Final	−0.09	0.94	−0.34	0.87	17.87 (18)	0.47	- (since items were split for DIF)	0.90
Specific activities subscale
Initial	0.08	1.27	−0.32	1.21	55.96 (33)	0.01*	1 % (7 % –13 %)*	0.83
Final	0.05	1.82	−0.39	1.08	10.87 (15)	0.76	3 % (1 % – 6 %)	0.91
Usual activities subscale
Initial	−0.72	0.56	−0.41	0.94	13.39 (12)	0.34	2 % (1 % - 5 %)	0.82
Final	−0.55	1.02	−0.44	1.02	6.01 (10)	0.82	-(since items were split for DIF)	0.86

*Source of misfit to the Rasch model; SD = Standard deviation; df = Degrees of freedom; per C < 5 % = proportion of t tests that were significant at level of significance of 0.05; 95 % CI = 95 % confidence interval; PSI = Person separation index; PREE – Patient Rated Elbow Evaluation
• For the data to satisfy Rasch model requirements
• Mean is expected to be approx. around zero (Can range between + 2.5 to −2.5)
• S.D. should be approx. 1
• Chi square value is expected to be small and statistically non-significant
• For a measure to display evidence of unidimensionality, less than 5 % of the t-tests should be significant at p = 0.05. If more than 5 % of the tests are significant, then the lower bound of the 95 % confidence interval should be less than 5 % to offer some support of unidimensionality
• PSI (Person separation index) PSI should be greater than 0.7 to obtain good power for the tests of fit

ISSN: 1477-7525