Proxy respondents are frequently used in surveys, including those assessing health-related quality of life (HRQOL). In cancer, most research involving proxies has been undertaken with paired proxy-patient populations, where proxy responses are compared to patient responses for the same individual. In these populations, proxy-patient differences are small and suggest proxy underestimation of patient HRQOL. In practice, however, proxy responses will only be used when patient responses are not available. The difference between proxy and patient reports of patient HRQOL where patients are not able to report for themselves in cancer is not known. The objective of this study was to evaluate the difference between patient and proxy reports of patient HRQOL in a large national cancer survey, and determine if this difference could be mitigated by adjusting for clinical and sociodemographic information about patients.
Data were from the Cancer Care Outcomes Research and Surveillance (CanCORS) study. Patients or their proxies were recruited within 3–6 months of diagnosis with lung or colorectal cancer. HRQOL was measured using the SF-12 mental and physical composite scales. Differences of ½ SD (=5 points) were considered clinically significant. The primary independent variable was proxy status. Linear regression models were used to adjust for patient sociodemographic and clinical covariates, including cancer stage, patient age and education, and patient co-morbidities.
Of 6471 respondents, 1011 (16%) were proxies. Before adjustment, average proxy-reported scores were lower for both physical (−6.7 points, 95% CI -7.4 to −5.9) and mental (−6 points, 95% CI -6.7 to −5.2) health. Proxy-reported scores remained lower after adjustment (physical: −5.8 points, −6.6 to −5; mental: −5.8 points, −6.6 to 5). Proxy-patient score differences remained clinically and statistically significant, even after adjustment for sociodemographic and clinical variables.
Proxy-reported outcome scores for both physical and mental health were clinically and significantly lower than patient-reported scores for these outcomes. The size of the proxy-patient score differences was not affected by the health domain, and adjustment for sociodemographic and clinical variables had minimal impact.