Response shift, recall bias and their effect on measuring change in health-related quality of life amongst older hospital patients

Table 2 Mean change, intraclass correlation coefficient (ICC), and limits of agreement (LOA) between change scores calculated from conventional longitudinal assessments and the patients' perspective (with and without adjustment for recall bias).

Measure	Patient perspective adjusted for recall bias	Conventional change mean (95% CI)	Patient perspective change mean (95%CI)	ICC (95% CI)	Limits of agreement			p-value*
					Lower LOA (95% CI)	Mean difference (95% CI)	Upper LOA (95% CI)
EQ-5D Utility	No	0.287 (0.216,0.359)	0.441 (0.367,0.518)	0.34 (0.16,0.50)	-1.007 (-1.092,-0.922)	-0.150 (-0.239,-0.069)	0.700 (0.616,0.785)	< 0.001*
EQ-5D Utility	Yes	0.287 (0.216,0.359)	0.303 (0.232,0.375)	0.98 (0.97,0.99)	-0.150 (-0.163,-0.136)	-0.016 (-0.116,-0.084)	0.118 (0.105,0.131)	0.019*
EQ-5D VAS	No	9.3 (5.4,13.2)	26.7 (22.8,30.7)	0.40 (0.22,0.55)	-60.7 (-65.0,-56.4)	-17.4 (-21.7,-13.1)	25.8 (21.5,30.1)	< 0.001*
EQ-5D VAS	Yes	9.3 (5.4,13.2)	11.0 (6.7,15.3)	0.90 (0.86,0.93)	-19.9 (-21.8,-18.1)	-1.7 (-3.5,0.1)	16.5 (14.7,18.3)	0.060

* A p-value < 0.05 indicates that a systematic difference exists (i.e. change from patient perspective was either consistently higher or consistently lower than conventional change scores.

ISSN: 1477-7525