Measuring change in patient quality of life over time : an evaluation of scale responsiveness and patient response shift.

Measuring change in quality of life is increasingly central to health services and clinical research evaluation. This requires instruments that are responsive to change, and that the construct being assessed is stable. I have, therefore, addressed two methodological themes: scale responsiveness and instability of the underlying quality of life construct - response shift. Responsiveness theme: I evaluated performance characteristics of a commonly reported effect size statistic, the standardised response mean (SRM). Computer simulations modelled the impact of varying computational method and distributional characteristics upon bias of estimated effect size compared to underlying true value. The studies provide evidence and reassurance that the SRM exhibits little bias when sample size, mean underlying effect size and shape of underlying distribution are varied. However, alternate approaches to handling negative values can produce markedly different effect sizes, making comparison across studies that use different methods problematic. Furthermore, parametric SRMs calculated from lognormal data may provide a greatly inflated estimate of effect size. Response shift theme: I interviewed patients at different stages of clinical management for knee injury twice over six months. A multi-method approach incorporating the individualised SEIQoL-DW measure and a retrospective pretest-posttest using EQ-5D identified evidence of re-calibration, re-prioritisation and re-conceptualisation response shift. An empirically based typology of changes was developed drawn from existing response shift theory, but which further distinguishes subtler forms of change. The studies provide evidence that re-prioritisation and re-conceptualisation may be different levels of the same process. Furthermore, mechanisms producing response shift were identified, in particular, the interaction between level of satisfaction with quality of life domain and its perceived importance. Additional approaches to studying response shift using group level comparison of SEIQoL data were critically evaluated. The thesis extends the methods for identifying, assessing and conceptualising response shift changes whilst also exploring mechanisms which may explain these changes