Friends and family test results only moderately associated with conventional measures of hospital qualityBMJ 2013; 347 doi: https://doi.org/10.1136/bmj.f4986 (Published 20 August 2013) Cite this as: BMJ 2013;347:f4986
- Felix Greaves, public health registrar1,
- Anthony A Laverty, research associate1,
- Christopher Millett, reader1
- 1Department of Primary Care and Public Health, Imperial College London, Reynolds Building, London W6 8RF, UK
NHS England recently released the first national data for the friends and family test.1 2 Proponents of the test argue that it injects the consumer voice into the NHS, provides patients with understandable information to make choices, and gives information much faster than conventional national patient surveys.3 4 Opponents suggest that the test is not sufficiently validated, has inadequate methodological consistency (hospitals can choose how to collect responses), and asks questions that are inappropriate in some patient groups (who would ever recommend a cancer service?).5 6
In an early analysis, we compared the test’s first three months’ findings for inpatients with those of the most recent (2012) NHS inpatient survey (using the overall score question) at the hospital level, obtained from the UK Data Service.7 We used Spearman’s rank correlation coefficient to compare unadjusted scores for the 156 acute hospitals with results from friends and family tests. The correlation coefficient was 0.46 (P<0.001).
We also compared friends and family test ratings with the latest results for the summary hospital mortality indicator for January to December 2012, obtained from the Health and Social Care Information Centre.8 For the 142 hospitals with test ratings, the correlation coefficient was 0.21 (P=0.01).
These results suggest that the friends and family test is associated with other measures of hospital quality, but that these associations are only mild to moderate, possibly because of sampling and response bias. Caution is needed when using them as a comparative measure of hospital performance.
Cite this as: BMJ 2013;347:f4986
Competing interests: None declared.