Retest reliability and validity of a surveillance measure of health-related quality of life

E. Andresen, T. Catlin, K. Wyrwich, Quality of Life Research 2001;10(3):199.

Measures of health-related quality of life (HRQoL) are increasingly common to clinical research, but uncommonly used in surveillance systems. For surveillance use, HRQoL measures need to be brief, generic, and broadly applicable to public health questions. HRQoL has been proposed as a method to monitor health in the American public health plan, "Healthy People 2010." We investigated the reliability and validity of the 4 HRQoL surveillance questions from the Behavioral Risk Factor Surveillance System--an ongoing nation-wide random-digit dialed telephone survey of Americans (aged greater than equal to 18 years of age). In 1999, randomly sampled adults from a midwestern state were re-interviewed for a retest of the HRQoL questions; 868 adults completed both interviews. The HRQoL questions include Self-Reported Health and 3 "days" questions measuring poor physical health, poor mental health, and activity limitation in the past 30 days. Poor mental and physical health days combine for a positive summary measure of Healthy Days. Mental and physical health days were also dichotomized using a cutoff of 14 days. Reliability is reported as kappa statistics for categorical questions and intraclass correlation coefficients for continuous questions. Validity of continuous "days" questions was examined by linear trend analysis across Self-Reported Health categories. Retest reliability was excellent (0.75 or higher) for Self-Reported Health and Healthy Days measures, and moderate (0.58 to 0.71) for other measures. Reliability was lower for older adults. Other demographic subgroups (e.g., gender, race/ethnic group) showed no regular pattern of differing reliability. Validity of the "days" questions showed strong linear trends across categories of Self-Reported Health, especially for physical health days. Retest reliability of the HRQoL surveillance measures is moderate to excellent, and internal validity is strong. Scaling options will require future attention, as will research into appropriate metrics for what constitutes important population differences.