Differences between Proxy and Patient Assessments of Cancer Care Experiences and Quality Ratings

To assess the impact of proxy survey responses on cancer care experience reports and quality ratings.

Secondary analysis of data from Cancer Care Outcomes Research and Surveillance (CanCORS). Recruitment occurred from 2003 to 2005.

The study was a cross-sectional observational study. The respondents were patients with incident colorectal or lung cancer or their proxies.

Analyses used linear regression models with an independent variable for proxy versus patient responses as well as study site and clinical covariates. The outcomes were experiences with medical care, nursing care, care coordination, and care quality rating. Multiple imputation was used for missing data.

Among 6,471 respondents, 1,011 (16 percent) were proxies. The proportion of proxy respondents varied from 6 percent to 28 percent across study sites. Adjusted proxy scores were modestly higher for medical care experiences (+1.28 points [95 percent CI:+ 0.05 to +2.51]), but lower for nursing care (-2.81 [95 percent CI: -4.11 to -1.50]) and care coordination experiences (-2.98 [95 percent CI: -4.15 to -1.81]). There were no significant differences between adjusted patient and proxy ratings of quality.

Proxy responses have small but statistically significant differences from patient responses. However, if ratings of care are used for financial incentives, such differences could be exaggerated across practices or areas if proxy use varies.