Brennan and Clare reply

Abstract

Mortality and two indicators of morbidity SIR-The article by Brennan and Clare' seeks to demonstrate the relationship between two indicators of self-reported morbidity derived from the 1971 census and age-band specific mortality data for all causes of death at a level of aggregation below county level. They argue from the strength and nature of this relationship that it is reasonable to use mortality data in the RAWP formula applied to area or health district level.The conclusions of the paper are challenged on three grounds: (1) The data available is unsuited to the hypothesis being discussed; (2) The interpretation of the results is of doubtful statistical validity; and (3) The logic of the argument is open to question.The authors acknowledge that the morbidity data available are restricted to persons expected to be economically active and relate only to self-reported conditions which prevent economic activity. The population is thus composed mainly of males of working age, a group which is a relatively low user of health service resources. The authors concede that the morbidity data for the group '65 years and over' are unreliable in that these represent only a small part of the total morbidity spectrum, but proceed to make an extrapolation of doubtful validity from the morbidity experience of a younger age group. They further acknowledge that the relationship of the indicators to the total spectrum of morbidity for all age groups did not appear to be constant at each geographical data point. The absence of data for children aged 0-15 years precludes the use of standardised mortality ratios; the width of age-banding required to match the two sets of data for age-specific rates is, however, such that standardisation for age is minimal. In view of these major reservations concerning the data, it is surprising that the authors felt justified in proceeding.The interpretation of the results is contentious and in our view is of doubtful statistical validity.We note with interest that the scattergram supplied for the age-group 45-64 years is 'of the most highly correlated data'. The overall impression is misleading in that it does not reveal the strength of the relationship for the lower values of the morbidity range in which most of the data points lie. Neither can it indicate the considerable variation which may occur among narrow age bands with the age-group as a whole. The authors have used throughout their paper and calculations age-specific mortality rates in broad age bands. The RAWP formula uses SMRs which are disease-group-specific. At no point can we find an explanation in the paper for the assumption that the effect of these will be the same.In view of the above problems, the relationship claimed for morbidity and mortality must be regarded as not proven and the evidence insufficient to support the use of mortality data for resource allocation below regional levels. We welcome, however, the final conclusion that the RAWP formula should be tempered by planning aspirations.