Unbiased measurement of health-related quality-of-life
Bellinda King Kallimanis
Project: Project financed by Academic Medical Centre, University of Amsterdam
Project running from: 12 March 2008 – 12 March 2012
Promotores: Prof. dr. F.J. Oort, Prof. dr. M.A.G. Sprangers
Summary of project
Problem and objective
Health-related quality-of-life (HRQL) is generally measured through self-report. Selfassessment brings about the problem that patients may have different frames of reference when answering HRQL items. As a result, the measurement of HRQL may be biased. That is, observed differences in HRQL scores may reflect something else than true differences in HRQL. Measurement bias may not only be caused by differences in individual and environmental characteristics (e.g., gender, age, education, ethnicity, mother tongue), but also by differences in treatment and other clinical variables (e.g., diagnosis, disease severity). Therefore, even when patients are randomised, treatment effects on HRQL are biased when treated patients have another frame of reference than control patients. In the proposed research, the objectives are to identify predominant sources of bias in the measurement of HRQL, to account for these biases, and to determine the clinical significance of true effects on unbiased HRQL.
Method
Structural equation modelling with latent variables (LVM) provides a way to detect measurement bias, to account for apparent bias, and to measure true (i.e., unbiased) effects on HRQL. In the proposed research, LVM will be used in secondary analyses of existing data sets from randomised and non-randomised trials in clinical and psychosocial medicine. We will examine a range of clinical, individual, and environmental sources of bias in HRQL outcomes. Several suitable data sets are available for secondary analysis. The clinical significance of both measurement bias and true effects in HRQL will be evaluated with a generalisation of the “number needed to treat” and some other effect size indices used in medical and social science research.
Possible results
We will obtain insight into the size of measurement bias and its impact on observed differences, changes and effects in HRQL. With that, we will also obtain insight into the true effects of clinical, individual, and environmental variables on (unbiased) HRQL, and into the clinical significance of these true effects. Knowledge of true effects in HRQL will facilitate treatment decisions and patient care, and thus further evidence-based medicine.
Date of defence: 22 November 2011
Title of thesis: Unbiased measurement of health-related quality-of-life