On Sunday 9th April we discussed this paper:
Richard G. Wilkinson and Kate E. Pickett (2007) Income Inequality and Socioeconomic Gradients in Mortality, American Journal of Public Health, April 2008, Vol 98, No. 4. Accessible online via: http://ajph.aphapublications.org/doi/pdf/10.2105/AJPH.2007.109637
This paper is part of a wider project, the findings of which are accessible by the book The Spirit Level, which expands upon the theme of ecological comparisons and indicates that inequalities are problematic, even to those well-off in society. These ideas are still largely debated.
A full transcript can be obtained here and will be stored in our Archive, also.
Summary of #PHTwitJC Discussion
1 – Were the aims of the study clear?
Overall it was felt the aims were ‘hard to digest’ and ‘complex’ – dispite some participants acknowledging familiarity with the topic and major concepts. It was suggested that as this paper is part of the larger ‘Spirit Level’ project, a level of prior knowledge was assumed.
2 – Were the parameters used clear, relevant and valid?
It was noted that the study used US geographical parameters. These are conveniently well collected, with good coverage and high inter-region equality differences. The general consensus was that the population parameters were sensible, although more depth and detail would have been useful.
3 – Are the findings justifiable considering the information inputted ?
Some interesting critiques resulted from this question, for example one participant felt the findings were justified but noted:
“The authors were careful not to suggest the causal process – wouldve made it more debatable.” @duncautumnstore
Another participant queried why the statistical significance level differed from convention:
“Not sure why interaction sig was set at 0.1” @riyadhonline
Furthermore, it was noted that the identification and elimination of outliers was not made apparent by the authors. One participant suggested that graphs with data points would be a useful addition to the paper in order for readers to consider the effect of outliers, although it was acknowledged that frequently in scientific journals space is limited.
Further critique included no reported attempt model for other confounders that may vary across counties/states (such as ethnicity, crime rate and smoking prevalence).
4 – What implications do the findings have for public health practice & policy?
” it’s the classic question of how much evidence is enough evidence.” @duncautumnstore
In general it was felt that possible confounding variables were not explored, and participants felt it would be useful to compare the strength of correlation with other relevant factors.
It was also queried whether an ecological study design is innappropriate to understand the processes underlying population health inequalities – as ecological study designs can only infer association, not causal processes.
@michaelgrayer suggested that a cohort design tracking theorised causal variables, with a view to compare outcomes & covariates in individuals could help develop these ideas into evidence-based policy.
Overall there was no clear policy or practical implication as participants found it difficult to translate the findings as many queries remained, and a wider debate on the political nature of public health.