Rapid Responses to:

PAPERS: Robert S Kahn, Paul H Wise, Bruce P Kennedy, and Ichiro Kawachi State income inequality, household income, and maternal mental and physical health: cross sectional national survey BMJ 2000; 321: 1311-1315 [Abstract] [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Error in Table 1

Paul Duberstein   (27 November 2000)

Re: Error in Table 1

Robert S Kahn   (29 November 2000)

Relationship between area income inequality and individual health may still be statistical artefact

Matt Sutton   (6 December 2000)

Re: statistical artefact

Robert Kahn, Paul Wise, Bruce Kennedy, Ichiro Kawachi   (12 December 2000)

Error in Table 1 27 November 2000
Paul Duberstein, Associate Professor of Psychiatry University of Rochester School of Medicine and Dentistry Send response to journal: Re: Error in Table 1	This is a fascinating report. As I was studying the Gini scores posted on the web, I noticed that the Gini for Arizona (.437) is higher than the Gini for Connecticut, yet Arizona's disparity is categorized as medium and Connecticut's is categorized as high. I hope this is merely a typographical error with no implications for the statisitical modelling.
Re: Error in Table 1 29 November 2000
Robert S Kahn Send response to journal: Re: Re: Error in Table 1	Arizona was incorrectly placed in the middle column of the table and should be in the high income inequality column. In all analyses it was appropriately treated as a high income inequality state. We appreciate your careful reading of the paper.
Relationship between area income inequality and individual health may still be statistical artefact 6 December 2000
Matt Sutton, Senior Research Fellow Department of General Practice, University of Glasgow Send response to journal: Re: Relationship between area income inequality and individual health may still be statistical artefact	EDITOR - Gravelle demonstrated that the relationship between population health and income inequality may be a statistical artefact, properly explained by the non-linear relationship between income and health at the individual level. Kahn et al use individual level data on income and health, and area level estimates of income inequality, to try to identify the separate effect of area income inequality on individual health. However, their analysis could fall foul of Gravelle's critique. Kahn et al attempt to control for the effect of household income on health by categorising individuals into five income groups. They find that women in the lowest income group have a higher probability of fair/poor health if they live in areas with high income inequality. Conversely, women in the highest income group have a lower probability of fair/poor health if they live in high income inequality areas. These results are consistent with insufficient controlling for the effect of low income on health outcome, through the categorisation of individual income data into groups. High area income inequality does not just imply that there is a larger proportion of the population in the lowest and highest income groups. In addition, individuals in the lowest income group tend to have extremely low incomes, and individuals in the highest income group tend to have extremely high incomes. A comparison of individuals with less than $10,000 annual household income in high and low income inequality areas could involve a comparison of individuals with close to zero incomes and individuals with close to $10,000 incomes. Therefore, Kahn et al's findings are consistent with the non-linear relationship between income and health at the individual level. Kahn et al should report the distribution of incomes within their income groups in the different areas, or control for the effects of household income by modelling the non-linear relationship between income and health using a continuous income variable. Research efforts would be better devoted to highlighting the health damage caused by differences in income and evaluating potential remedies, rather than to chasing an effect of area income inequality on individual health. If area income inequality is found to impact on individual health, an uncomfortable conclusion will be that population health can be improved whilst retaining the current level of national income inequality. This could be achieved by maximising inequality between areas and minimising inequality within areas through the redistribution of people between areas rather than income between individuals. Matt Sutton Senior Research Fellow Department of General Practice, University of Glasgow, 4 Lancaster Crescent, GLASGOW G12 0RR 1. Gravelle H. How much of the relation between population mortality and unequal distribution of income is a statistical artefact? BMJ 1998; 316: 382-5. 2. Kahn RS, Wise PH, Kennedy BP, Kawachi I. State income inequality, household income, and maternal and physical health: cross sectional national survey. BMJ 2000; 321: 1311-5.
Re: statistical artefact 12 December 2000
Robert Kahn Children's Hosptial Medical Center, Paul Wise, Bruce Kennedy, Ichiro Kawachi Send response to journal: Re: Re: statistical artefact	Sutton highlights an important analytic challenge for the field. The potential for statistical artefact led us to examine the association between income inequality and health among the poorest tenth women as well. As reported in the paper, the association was perhaps even stronger, though the confidence intervals were wide. We conducted one further test for statistical artefact not reported in the text. In the stratified analysis of income inequality and health among women in the lowest income quintile (Table 3), we also ran a model that included a continuous income term. The continuous income term was not signficant and the effects of income inequality did not change. While not definitive, these analyses help diminish the probability that statistical artefact undermines our findings.