Skip Navigation


Journal of Public Health Advance Access originally published online on October 25, 2006
Journal of Public Health 2006 28(4):379-383; doi:10.1093/pubmed/fdl061
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
28/4/379    most recent
fdl061v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (5)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Adams, J.
Right arrow Articles by White, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Adams, J.
Right arrow Articles by White, M.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2006, Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved.

Removing the health domain from the Index of Multiple Deprivation 2004—effect on measured inequalities in census measure of health



J. Adams, MRC Special Training Fellow in Health Services and Health of the Public Research

M. White, Professor of Public Health
Institute of Health and Society, William Leech Building, Newcastle University, Newcastle upon Tyne, NE2 4HH, UK

Address correspondence to J. Adams, E-mail: j.m.adams{at}ncl.ac.uk

The Index of Multiple Deprivation (IMD) 2004 is a summary measure of area-level deprivation in England that combines weighted scores in seven deprivation domains. IMD 2004 is used extensively by local public health departments and researchers to describe and monitor socioeconomic inequalities in health. However, the inclusion of a health domain in IMD 2004 leads to the possibility of ‘mathematical coupling’ where a relationship between IMD 2004 and markers of health is predicated by the inclusion of health in IMD 2004—effectively placing measures of health on both sides of the correlation equation. We explored the effect of removing the health domain from IMD 2004 on assignment of small areas to deprivation groups and measured inequalities in health. There was excellent agreement between the deprivation quintiles that small areas were assigned to by IMD 2004 and IMD 2004-minus-health ({kappa} = 0.895). Removing the health domain had little, practical, effect on measured socioeconomic inequalities in census measures of health. These findings may not hold for other measures of health, and in the context of socioeconomic inequalities in health, removing the health domain from IMD 2004 probably represents best practice.

Keywords: census, IMD 2004, socioeconomic factors


   Introduction
TOP
 Introduction
Methods
 Results
 Discussion
 Competing interests
 Acknowledgements
 References
 
The UK government recently published explicit targets for reducing socioeconomic inequalities in two key markers of population health—infant mortality and life expectancy at birth.1 Socioeconomic deprivation is a difficult concept to define, and numerous different individual, household and area-based measures of socioeconomic position and deprivation exist,2,3 leading to confusion and a lack of comparability between studies. This is reflected in the UK government’s current targets that aim to reduce individual social-class inequalities in infant mortality, but area-level inequalities in life expectancy at birth.1

The current version of the Index of Multiple Deprivation (IMD 2004) is a summary measure of area-level deprivation that combines weighted scores in seven deprivation domains (Table 1).4 It is extensively used by local public health departments in England and by researchers to describe and monitor socioeconomic inequalities in health. However, unlike many previous markers of area-level deprivation—including the Towsend Deprivation Score,5 Carstairs Deprivation Index6 and the ward and enumeration-district level 1998 Index of Local Deprivation—IMD 2004, along with its immediate predecessor the Index of Deprivation 2000, includes health-related data. In the context of health inequalities analyses, where IMD 2004 is compared with markers of health, the inclusion of a health domain in IMD 2004 leads to the possibility of ‘mathematical coupling’7—‘the phenomenon whereby two variables will inevitably correlate if one contains or shares, directly or indirectly, all or part of the other’8 (see Archie7 for worked examples showing how even randomly generated numbers can be strongly correlated if mathematically coupled). Although the possibility of mathematical coupling when using IMD 2004 in health inequality analyses has been previously noted,9 no attempt to overcome it has yet been documented.


View this table:
[in this window]
[in a new window]

  		Table 1 Domain composition of the Index of Multiple Deprivation (IMD) 2004 and weightings before and after exclusion of health domain

 

Using census measures of health, we explored the effect of excluding the health domain from IMD 2004.


   Methods
TOP
 Introduction
 Methods
Results
 Discussion
 Competing interests
 Acknowledgements
 References
 
Data on IMD 2004 ranks in all seven domains, at the lower super output area (LSOA) level in England, were downloaded from the UK Department for Communities and Local Government website (http://www.communities.gov.uk). Although weighted scores from each domain are combined to produce IMD 2004, unweighted scores are not published and have to be generated from published ranks.4 New weights for the remaining six domains, after the exclusion of the health domain, were generated by redistributing the weight originally allocated to the health domain across the other domains, in proportion to the original weights (Table 1). These weights and calculated domain scores were used to generate IMD 2004-minus-health scores. Owing to the difficulty in interpreting IMD 2004 and IMD 2004-minus-health scores (e.g. a score of 20 does not represent twice as much ‘deprivation’ as a score of 10), scores for both indices were divided into quintiles for analyses.

Agreement between quintiles of deprivation assigned by IMD 2004 and IMD 2004-minus-health was assessed using the {kappa}-statistic. This was interpreted according to the schema proposed by Landis and Koch (1977), which proposes that {kappa} > 0.75 represents excellent agreement, {kappa} < 0.4 represents moderate or poor agreement and {kappa} values between 0.4 and 0.75 represent fair or good agreement.10

Data from the 2001 census on the presence of self-reported limiting long-term illness (LLTI) and less-than-good general health, for all LSOAs in England, were downloaded from http://www.nomisweb.co.uk. Sex-specific age-standardized LLTI and less-than-good health ratios for each LSOA in England were calculated using the indirect method and rates in England as a whole.

The ability of each of the measures of deprivation (quintiles of IMD 2004 and IMD 2004-minus-health—the independent or explanatory variables) to predict each of the measures of health (sex-specific age-standardized LLTI and less-than-good health ratios—the dependent or outcome variables) was investigated using linear regression. As residuals were not normally distributed in these models, bias-corrected bootstrapping with 100 repetitions was used to generate 95% confidence intervals (CIs).

To determine whether the relationship between measures of deprivation and measures of health varied according to how measures of deprivation were assigned (i.e. IMD 2004 or IMD 2004-minus-health), we generated interaction terms. These took the form of ‘assigned quintile*deprivation type’ where ‘deprivation type’ equalled 1 for IMD 2004 and 2 for IMD 2004-minus-health. Likelihood ratio tests comparing models with and without interaction terms were used to assess whether there was evidence of interaction.

All analyses were conducted in Stata v8.0. Ethical approval was not required for these analyses of publicly available data.


   Results
TOP
 Introduction
 Methods
 Results
Discussion
 Competing interests
 Acknowledgements
 References
 
Data from all 32 482 LSOAs in England were included in the analyses. There was agreement between deprivation quintile assigned according to IMD 2004 and IMD 2004-minus-health in 29 758 (91.6%) LSOAs ({kappa} = 0.895), indicating an excellent level of agreement.

Table 2 summarizes mean age-standardized LLTI and less-than-good health ratios in deprivation quintiles assigned according to IMD 2004 and IMD 2004-minus-health in men and women. Also summarized are linear regression coefficients for the ability of deprivation quintile to predict markers of health and the results of likelihood ratio tests comparing the linear relationship between health and quintiles of deprivation assigned according to IMD 2004 and IMD 2004-minus-health. There was evidence of strong and statistically significant trends in markers of health according to quintiles of deprivation score in all cases. Although there was evidence that the relationship between health and IMD 2004 was different from the relationship between health and IMD 2004-minus-health (likelihood ratio tests statistically significant), the actual difference in regression coefficients was very small in all cases.


View this table:
[in this window]
[in a new window]

  		Table 2 Age-standardized limiting long-term illness and less-than-good general health ratios by quintile of the Index of Multiple Deprivation (IMD) 2004 and IMD 2004-minus-health score, with regression coefficients and likelihood ratio tests for interaction, lower super output areas, England

 


   Discussion
TOP
 Introduction
 Methods
 Results
 Discussion
Competing interests
 Acknowledgements
 References
 
Main finding of this study
There was strong agreement between quintiles of deprivation generated from the new and original IMD 2004 scores. Although there was evidence that the relationship between census measures of health and quintiles of deprivation based on the new and original IMD 2004 scores differed, these differences were small and of limited practical importance. These conclusions were not altered by dividing the data into deciles rather than quintiles (data not shown).

What is already known on this topic
The IMD 2004 is generally accepted as the governments’ preferred measure of area-level deprivation. It is widely available and easy to use in relation to many other centrally produced statistics. IMD 2004 is extensively used by public health practitioners and researchers to describe and monitor socioeconomic inequalities in health.

The concept of ‘multiple deprivation’ endorsed by IMD 2004 proposes that deprivation can, and does, occur in many different areas and, specifically, that deprivation is not confined to income deprivation or poverty. The inclusion of the health domain in IMD 2004 highlights both that poor health in the absence of, for instance, income deprivation is still a form of deprivation and that poor health in the presence of income, and other forms of deprivation, deepens the deprivation experienced. There is, thus, strong theoretical justification for the inclusion of the health domain in IMD 2004.

However, in the context of socioeconomic inequalities in health, the inclusion of a health domain in IMD 2004 may lead to mathematical coupling where a relationship between IMD 2004 and markers of health is predicated by the inclusion of health in IMD 2004 —effectively leading to measures of health being on both sides of the correlation equation. Although the potential problem of mathematical coupling when using IMD 2004 to explore inequalities in health has been previously identified,9 this is the first study of the effects of removing the health domain from IMD 2004.

What this study adds
We have found that removing the health domain from IMD 2004 has little effect on either the deprivation quintile to which LSOAs in England are assigned or the relationship between these and census markers of health. There is, therefore, little evidence that any mathematical coupling occurring leads to important influences on measured socioeconomic inequalities in health in this case.

Although the health domain has a weighting of 13.5% in IMD 2004, we found that removing it did not lead to substantial changes in the degree of deprivation-related health inequalities measured. This is probably explained by the strong inter-correlations between many of the domains of IMD 2004.11 The statistically significant likelihood ratio tests comparing models with and without interaction terms, despite little actual difference in regression coefficients, are probably due to the very large ‘sample’ size (n = 32 482) which allows very small differences, of little practical importance, to be identified.

Removing the health domain from IMD 2004 did require some technical expertise and time. Although our findings suggest that removing the health domain from IMD 2004 is unnecessary for routine public health analyses using census measures of health, mathematical coupling and the associated problems in interpreting results remain a potential issue when using IMD 2004 to investigate patterns in other measures of health. In the context of socioeconomic inequalities in health, removing the health domain from IMD 2004 probably represents best practice.

Limitations
We used only census measures of health—the presence of long-term limiting illness and less-than-good self-reported health. These measures of health do not closely approximate to those included in the health domain of IMD 2004, and it is possible that a similar analysis using different measures of health would lead to greater evidence of important mathematical coupling. In addition, although it has been suggested that LLTI rates may not be an accurate marker of morbidity burden,12 there is good evidence that self-reported health is an accurate predictor of long-term morbidity and mortality.13

The area-level nature of our analyses means that our findings cannot necessarily be applied at the individual level. However, IMD 2004 and its predecessors have routinely been used to describe and monitor variations in individual and area-level health data.9,14 –17 Despite the inherent limitations of area-based analyses, these findings remain highly relevant to public health practitioners and routine public health practice.


   Competing interests
TOP
 Introduction
 Methods
 Results
 Discussion
 Competing interests
Acknowledgements
 References
 
Neither author has any competing interests to declare.


   Acknowledgements
TOP
 Introduction
 Methods
 Results
 Discussion
 Competing interests
 Acknowledgements
References
 
J.A. is supported by a UK Medical Research Council Special Training Fellowship in Health Services and Health of the Public Research.


   References
TOP
 Introduction
 Methods
 Results
 Discussion
 Competing interests
 Acknowledgements
 References
 

  1. Department of Health. Tackling Health Inequalities: A Programme for Action. London: The Stationary Office, 2003.
  2. Galobardes B, Shaw M, Lawlor DA et al. Indicators of socioeconomic position (part 2). J Epidemiol Community Health 2006;60:95–101.[Abstract/Free Full Text]
  3. Galobardes B, Shaw M, Lawlor DA et al. Indicators of socioeconomic position (part 1). J Epidemiol Community Health 2006;60:7–12.[Abstract/Free Full Text]
  4. Noble M, Wright G, Dibben C et al. Indices of Deprivation 2004: Report to the Office of the Deputy Prime Minister. London: Neighbourhood Renewal Unit, 2004.
  5. Townsend P, Phillimore P, Beattie A. Health and Deprivation: Inequality and the North. Bristol: Croom-Helm, 1988.
  6. Carstairs V, Morris R. Deprivation and Health in Scotland. Newcastle upon Tyne: Aberdeen University Press, 1991.
  7. Archie J. Mathematical coupling of data: a common source of error. Ann Surg 1981;193:296–303.[Web of Science][Medline]
  8. Blake P. The problem of mathematical coupling: how can statistical artefact and biological causation be separated when relating protein intake to clearance in ‘predialysis’ and dialysis patients? Perit Dial Int 1997;17:431–4.[Free Full Text]
  9. Jordan H, Roderick P, Martin D. The index of multiple deprivation 2000 and accessibility effects on health. J Epidemiol Community Health 2004;58:250–7.[Abstract/Free Full Text]
  10. Landis J, Koch G. The measurement of observer agreement for categorical data. Biometrics 1977;33(1):159–74.[CrossRef][Web of Science][Medline]
  11. Adams J, White M. Socio-economic deprivation is associated with increased proximity to general practices in England: an ecological analysis. J Public Health 2005;27:80–1.
  12. Scrivener G, Lloyd D. Two proxies for morbidity in the 1991 UK census: permanent sickness and limiting long term illness. J Epidemiol Community Health 1996;50(2):223.
  13. Idler E, Benyammi Y. Self-rated health and mortality: a review of twenty-seven community studies. J Health Soc Behav 1997;38(1):21–37.[CrossRef][Web of Science][Medline]
  14. Sood MR, Joshi S, Akobeng AK et al. Growth in children with Helicobacter pylori infection and dyspepsia. Arch Dis Childhood 2005;90(10):1025–8.[Abstract/Free Full Text]
  15. Morris E, Landes D. The equity of access to orthodontic dental care for children in the North East of England. Public Health 2006;120(4):359–63.[CrossRef][Web of Science][Medline]
  16. Woolley E, Magennis P, Shokar P et al. The correlation between indices of deprivation and health-related quality of life in patients with oral and oropharyngeal squamous cell carcinoma. Br J Oral Maxillofac Surg 2006;44(3):177–86.[CrossRef][Web of Science][Medline]
  17. Woods LM, Rachet B, Riga M et al. Geographical variation in life expectancy at birth in England and Wales is largely explained by deprivation. J Epidemiol Community Health 2005;59(2):115–20.[Abstract/Free Full Text]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
28/4/379    most recent
fdl061v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (5)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Adams, J.
Right arrow Articles by White, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Adams, J.
Right arrow Articles by White, M.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?