Using mortality data to describe geographic variations in health status at sub-district level

To describe sub-district variations in health status, using mortality data that are processed locally.

A descriptive study of routinely collected death registration data, using multicause coding.

The London Borough of Croydon, with a population of 319,200 divided into 27 electoral wards.

Deaths of Croydon residents, registered with the Registrar of Births and Deaths, which occurred between January 1990 and December 1992 inclusive.

Variations in life expectancy, all-cause standardised mortality ratios (SMRs), and disease-specific mortality ratios between selected wards. Deaths in nursing homes were excluded to avoid bias.

Data from 8,930 death registrations, of which 852 occurred in nursing homes, were analysed by electoral ward. The range for all-cause SMRs, including nursing home deaths, was 153 (139-168) to 66 (58-75). When nursing home deaths were excluded, the SMRs for two wards that were significantly higher than the Croydon average fell into the average range. The range, excluding nursing home deaths, was 133 (113-153) to 71 (62-80). Life expectancy at birth varied from 79.8 years to 74.4 years, and life expectancy at age 65 by three years between wards at the two ends of the spectrum. The geographic distribution of ischaemic heart disease and diabetes showed significant differences.

We contend that death registration data are a useful tool for describing sub-district variations in health status. Deaths of nursing home residents are a source of bias and should be excluded from the analysis. Multicause coding allows a more accurate description of geographic variations in specific diseases, such as ischaemic heart disease and diabetes.