To determine whether use of proxy respondents in a patient experience survey was related to patient characteristics, and to compare patient and proxy responses. Secondary analysis, using propensity score matching, of the NHS adult inpatient survey, a large cross-sectional survey. Hospitals (n = 161) providing inpatient services in England in 2011. The survey received 70 863 responses: 10 661 (15.6%) involved proxy respondents in some way. None. Prevalence of proxy response was explored by patient demographic characteristics. Responses were compared using seven composite domains and one overall rating. Cases involving proxy responses were matched to similar independent responses via propensity score matching and mean scores compared using t-tests. Use of proxy respondents was common, with 15.7% of responses involving a proxy in some way: higher than in other similar collections internationally. Proxy response was more common for some patient groups, such as older people and those from black and minority ethnic groups. Reports made by or with the assistance of proxy respondents were markedly less positive than those from patients completing the survey unaided. This pattern was consistent across all tested variables, although the biggest differences were observed for a subjective 'overall rating' question. The prevalence of proxy response varied according to patient characteristics, but proxies were consistently less positive than patients responding unaided. Possible explanations include genuine differences in care, differential health outcomes or differences in perceptions. Patient experience surveys should collect information on use of proxy respondents to enable more refined analysis.
