M McKee1, D Hunter. 1. Department of Public Health and Policy, London School of Hygiene and Tropical Medicine, London.
Abstract
OBJECTIVE: To examine certain methodological issues related to the publication of mortality league tables, with particular reference to severity adjustment and sample size. DESIGN: Retrospective analysis of inpatient hospital records. SETTING: 22 hospitals in North West Thames health region for the fiscal year 1992-3. SUBJECTS: All admissions with a principal diagnosis of aortic aneurysm, carcinoma of the colon, cervical cancer, cholecystectomy, fractured neck of femur, head injury, ischaemic heart disease, and peptic ulcer. MAIN MEASURES: In hospital mortality rates adjusted by disease severity and calculated on the basis of both admissions and episodes. RESULTS: The numbers of deaths from specific conditions were often small and the corresponding confidence intervals wide. Rankings of hospitals by death rate are sensitive to adjustment for severity of disease. There are some differences that cannot be explained using routine data. CONCLUSIONS: Comparison of crude death rates may be misleading. Some adjustment for differences in severity is possible, but current systems are unsatisfactory. Differences in death rates should be studied, but because of the scope for manipulating data, this should be undertaken in a collaborative rather than a confrontational way. Any decision to publish league tables of death rates will be on political rather than scientific grounds.
OBJECTIVE: To examine certain methodological issues related to the publication of mortality league tables, with particular reference to severity adjustment and sample size. DESIGN: Retrospective analysis of inpatient hospital records. SETTING: 22 hospitals in North West Thames health region for the fiscal year 1992-3. SUBJECTS: All admissions with a principal diagnosis of aortic aneurysm, carcinoma of the colon, cervical cancer, cholecystectomy, fractured neck of femur, head injury, ischaemic heart disease, and peptic ulcer. MAIN MEASURES: In hospital mortality rates adjusted by disease severity and calculated on the basis of both admissions and episodes. RESULTS: The numbers of deaths from specific conditions were often small and the corresponding confidence intervals wide. Rankings of hospitals by death rate are sensitive to adjustment for severity of disease. There are some differences that cannot be explained using routine data. CONCLUSIONS: Comparison of crude death rates may be misleading. Some adjustment for differences in severity is possible, but current systems are unsatisfactory. Differences in death rates should be studied, but because of the scope for manipulating data, this should be undertaken in a collaborative rather than a confrontational way. Any decision to publish league tables of death rates will be on political rather than scientific grounds.
Authors: R E Park; R H Brook; J Kosecoff; J Keesey; L Rubenstein; E Keeler; K L Kahn; W H Rogers; M R Chassin Journal: JAMA Date: 1990-07-25 Impact factor: 56.272