Gradient between north and south remainsBMJ 1994; 308 doi: https://doi.org/10.1136/bmj.308.6922.204 (Published 15 January 1994) Cite this as: BMJ 1994;308:204
EDITOR, - It is surprising that Margaret Eames and colleagues occasionally mentioned, but did not incorporate in their abstract or their section on the “implications for the Health of the Nation,” the overall geographical finding that premature mortality increases steeply from the south to the north of England.1 The differences in regional mortality have been fairly constant for at least the past 30 years2,3 and the search for reasons for and solutions to the north-south divide is surely one of the greatest challenges to medical and public health research. The omission was particularly disappointing in view of the new finding of the paper that variations in deprivation account for some, but not all, regional variations in premature mortality (and hence the north-south gradient). The paper implies the need to target not only improvements in health care but also improvements in social and material circumstances in the north more than the south, but these implications were omitted.
Two additional points arise from the paper. Firstly, the authors seem to have attributed to individual people an overall average characteristic of an area (the ecological fallacy). Thus the authors state: “Although, intuitively, increasing affluence may not be expected to have a continual effect on mortality, this was not seen in the dataset.” However, the measures of affluence (or, rather, its inverse, deprivation) are composed of percentages - for example, the percentage unemployment - in a given area and therefore a continuous relation would be expected. The authors thus seem to imagine, wrongly, that in electoral wards (which are highly heterogenous with respect to deprivation of individual people) the deprivation measures used in some sense describe the deprivation of all the residents.
Secondly, the attempt to show that the prevalence of smoking explains north-south variation in mortality from diseases related to smoking was inadequate and should have been omitted. If two regions have values at opposite extremes and the 12 others show little variation this cannot support, even crudely, a hypothesis about general north- south variation.