Explaining differences in English hospital death rates using routinely collected data
BMJ 1999; 318 doi: https://doi.org/10.1136/bmj.318.7197.1515 (Published 05 June 1999) Cite this as: BMJ 1999;318:1515
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
EDITOR: Jarman et al. present an intriguing analysis (1) - but I do
not yet think we can state with confidence, 'More doctors means fewer
deaths'. As the authors make clear, there has been a lively discussion on
comparative hospital death rates in the United States. From this debate
two points emerge clearly; for any given population the standardised
admissions rate is positively correlated with the standardised mortality
rate - but is inversely correlated with the standardised hospital death
rate (defined as any death within 30 days of a hospital admission)(2,3).
Where a population is admitted to hospital more frequently, then a higher
proportion of admissions will not be associated with subsequent death, and
hence there will be a lower apparent rate of hospital mortality.
It was therefore most informative that the authors included with
their explanatory variables a calculation of the standardised admission
ratio (though not apparently separated into 'all cases' and 'emergencies',
for the two versions of their model). Unfortunately, the figures used
were the aggregates for the Health Authority of hospital location, rather
than individually for each hospital's emergency catchment area (often very
different). Even in this half-baked form, standardised admissions entered
significantly into both the 'all cases' and 'emergency' multiple
regression models. It is not impossible that a fully specified version of
this variable might not displace some or both the 'doctor' variables
One way to resolve this issue would be to exploit a generally
observed feature of the NHS in England, that there tends to be relatively
little overlap in the geographical catchment areas of acute hospitals for
adult emergency admissions. Hence it should be possible to repeat the
authors' analysis exactly - but aggregated by electoral ward of residence
rather than by hospital of admission. The aggregate discharge data would
be the same (except that true standardised admission rates could be used).
The community data could then be used directly (with GP numbers per head
calculated using FH registers) - but the hospital data would need to be
cross-attributed. It would also then be possible to match the hospital
mortality data set with community data on general mortality and life
expectancy.
Tom Hennell
Strategic Analyst
NHS Executive North West
1. Jarman B, Gault S, Alves B, Hider A, Dolan S, Cook A, Hurwitx B,
Iezzoni LI. Explaining differences in English hospital death rates using
routinely collected data. BMJ 1999; 318:1515-20.
2. Welch WP, Miller ME, Welch HG, Fisher ES, Wennberg JE. Geographic
variation in expenditures for physicians' services in the United States. N
Engl J Med 1993;328:621-7.
3. Health Care Financing Administration. Medicare hospital
information report. 1992 Technical supplement. Section D. Hospitalization
and mortality data for states. Washington, D.C.: Department of Health and
Human Services, 1992.
4. Manheim LM, Feinglass J, Shortell SM, Hughes EFX. Regional
variation in Medicare hospital mortality. Inquiry 1992;29:55-66.
Competing interests: No competing interests
Data won't support authors' conclusions
Editor
Efforts to compare the quality of care among hospitals have defeated
many investigators in the past. To the list of failures the Journal has
added the paper by Jarman and his associates that attempts to explain
differences in English hospital death rates using routinely collected data
(1). Their major finding is an inverse association between hospital
mortality rates and the number of hospital doctors and general
practitioners, from which they conclude that the ratio of doctors to
population "seem to be critical determinants of standardised hospital
death rates; the higher these ratios, the lower the death rates". By
implication, and as widely interpreted by the press, hospitals with fewer
doctors have, as a result, higher death rates. This may well be true, but
the data are an inadequate basis for drawing that conclusion.
The strength or weakness of such investigations hinges on the
accuracy of the measurement of prognostic or risk factors in patients
treated by the hospitals being compared. In the present study, severity of
the primary illness could not be estimated and comorbidity was limited to
a count of sub-diagnoses for each patient in Hospital Episode Statistics.
The inadequacy of such routine data has been well documented by the NHS
Executive (2), by Iezzoni (3, 4), who is listed as a coauthor of the
current paper, and by Jencks (5), while the impact of comorbidity, when
measured carefully, on mortality has been demonstrated (6). The dangers of
using inadequate information has been shown by Iezonni and her colleagues
who reported the counter-intuitive finding that many sub-diagnoses "(eg.,
adult-onset diabetes mellitus, essential hypertension, previous myocardial
infarction, angina, and ventricular premature beats)" were associated with
lower death rates (3). Much the same finding had previously been reported
by Jencks and his associates (5).
Severity of illness and comorbidity are best judged prospectively,
preferably in "consultation with physicians" (3). Second best is
retrospective individual patient record review comparing condition
specific diagnoses and procedures case by case. The least satisfactory
approach is to rely on routinely collected undifferentiated data, as was
done in the study by Jarman and his associates. The authors are aware of
possible shortcomings of their approach, writing that "a matched pair
study of patients admitted to hospitals with high and low standardised
mortality ratios could help to elucidate [their] findings." Short of such
an investigation, the public, as well as the profession, is left to draw
conclusions that may very well be incorrect.
John P Bunker MD FRCP
Visiting Professor
Department of Epidemiology and Public Health
University College London School of Medicine
Nick Black
Professor
Department of Public Health and Policy
London School of Hygiene and Tropical Medicine
1 Jarman B, Gault S, Alves B, Hider A, Dolan S, Cook A, Hurwitz B,
Iezzoni LI. Explaining differences in English hospital death rates using
routinely collected data. BMJ 1999;318:1515-20.
2 NHS Executive. Quality and performance in the NHS: clinical
indicators. NHSE June 1999.
3 Iezonni LI, Foley SM, Daley J, Hughes J, Fisher ES, Heeren T.
Comorbidities, complications, and coding bias: does the number of
diagnosis codes matter in predicting in-hospital mortality? JAMA
1992;267:2197-2203/
4 Iezzoni LI, Schwartz M, Ash AS, Mackiernan YD. Does severity
explain differences in hospital length of stay for pneumonia patients? J
Health Serv Res Policy 1996;1:65-76.
5 Jencks SF, Williams DK, Kay TL. Assessing hospital-associated
deaths from discharge data: the role of length of stay and comorbidities
JAMA 1988;260:2240-46
6 Imamura K, Black N. Does comorbidity affect the outcome of
surgery? Total hip replacement in the UK and Japan . Int J Qual Health
Care 1998;10:113-123.
Competing interests: No competing interests