Principal variable is not what it seems in league tables

Editor—Sir Brian Jarman's analysis of hospital death rates with “Dr Foster's guide to better health” (Sunday Times) may serve to improve the quality of hospital care—indirectly.¹ The principal dependent variable is, however, not what it seems, even after adjustment for age, sex, diagnosis, emergency admission, and length of stay, so that like is not compared with like. Rates derived from hospital episode statistics, deaths per 1000 finished consultant episodes, almost defy interpretation, because the denominators are episodes, not patients. Although this analysis selects a subset of episodes that end in discharge or death, the denominators represent admissions, not people. Fairer measures of hospital performance are based on 30 day deaths per 100 000 population.²,³

The first conclusion of the study should read that the number of hospital episodes (or admissions) has increased by approximately 2.6% annually. The numbers of deaths have remained nearly constant. It is only a consequence of increased activity that “episode fatality rates” seems to have fallen.

The second main observation compares episode fatality rates with the ratio of doctors to beds, a ratio of two provision measures: (hospital) doctors and (acute hospital) beds per 100 000 population. It would be preferable to examine relations with these two measures of provision independently. High ratios of doctors to beds are found in tertiary centres, and low episode fatality rates in such hospitals could be an artefact of denominator inflation: more doctors in more specialties so that one patient and one illness appears as more than one episode in more than one specialty.

The third main observation, association with provision in general practice, may be true yet have little to do with the quality of hospital care, if districts and communities of high provision have appropriate alternatives for care of the dying, at home or in hospices, in the final days, after curative treatment has been abandoned. There are many factors outside hospital that affect hospital death rates even after adjustments as in this analysis, including admission and discharge policies and care in the community.⁴

Dr Foster's guide, effectively a full league table, may help health professionals and managers to identify weaknesses, where weaknesses occur, and improve services more than they alarm patients.⁵ It is to be hoped that poor ratings may not be improved by the simple expedient of denominator adjustment. It should, however, not be forgotten that, across most of the country, patients do not have choice; when ill we go to “our” local hospital.