HEALTH ECONOMICS Health Econ. 24: 1523–1530 (2015) Published online 7 October 2014in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/hec.3102
HEALTH ECONOMICS LETTER
INDIVIDUAL INCOME, AREA DEPRIVATION, AND HEALTH: DO INCOME-RELATED HEALTH INEQUALITIES VARY BY SMALL AREA DEPRIVATION? MARTIN SIEGELa,b, , ANDREAS MIELCKc and WERNER MAIERc a Berlin
Centre of Health Economics Research (BerlinHECOR), Department of Health Care Management, Technische Universität Berlin, Berlin, Germany b Institute c Institute
of Health Economics and Clinical Epidemiology, University Hospital of Cologne, Germany
of Health Economics and Health Care Management, Helmholtz Zentrum München, Neuherberg, Germany
ABSTRACT This paper aims to explore potential associations between health inequalities related to socioeconomic deprivation at the individual and the small area level. We use German cross-sectional survey data for the years 2002 and 2006, and measure small area deprivation via the German Index of Multiple Deprivation. We test the differences between concentration indices of income-related and small area deprivation related inequalities in obesity, hypertension, and diabetes. Our results suggest that small area deprivation and individual income both yield inequalities in health favoring the better-off, where individual income-related inequalities are significantly more pronounced than those related to small area deprivation. We then apply a semiparametric extension of Wagstaff’s corrected concentration index to explore how individual-level health inequalities vary with the degree of regional deprivation. We find that the concentration of obesity, hypertension, and diabetes among lower income groups also exists at the small area level. The degree of deprivation-specific income-related inequalities in the three health outcomes exhibits only little variations across different levels of multiple deprivation for both sexes. Copyright © 2014 John Wiley & Sons, Ltd. Received 17 September 2013; Revised 21 July 2014; Accepted 8 August 2014 KEY WORDS:
regional deprivation; income; health inequalities; concentration index
1. INTRODUCTION The existence of income-related health inequalities favoring the better-off is well established in the literature (Wagstaff et al., 2003; Jones and López Nicolás, 2006; van Doorslaer et al., 2004; van Doorslaer and Koolman, 2004). Stafford and Marmot (2003), for instance, further argue that potential risks associated with living in deprived areas may reinforce the association between individual socioeconomic position and health, and more recent research suggests that regional deprivation may be an important determinant for the distribution of health (Schuurman et al., 2007; Walker et al., 2011; Connolly et al., 2000; Maier et al., 2012, 2013, 2014; Kuznetsov et al., 2011, 2012; Koller et al., 2013; Jansen et al., 2014). Following Maier et al. (2014) who argue that small area deprivation and individual-level socioeconomic status are independent determinants of health, one may expect inequalities with respect to individual incomes to exist at the small area level. Using a pooled sample from the 2002 and 2006 waves of the Taylor Nelson Sophres Health Care Access Panel (HCAP), this is the first paper measuring variations of income-related health inequalities across different levels of small area deprivation. We take obesity, hypertension, and diabetes as
Correspondence to: Department of Health Care Management, Technische Universität Berlin. E-mail: [email protected]
Copyright © 2014 John Wiley & Sons, Ltd.
1524
M. SIEGEL
an example and apply the German Index of Multiple Deprivation (GIMD) to measure small area deprivation at the commune level (Maier et al., 2012, 2013). Both household income related and small area deprivation related health inequalities are measured using concentration indices (Wagstaff et al., 1991). We first measure the differences between them following Wagstaff and Watanabe (2003) to explore to what extent inequalities with respect to the two indicators differ in our data. Using a semiparametric extension of Wagstaff’s corrected concentration index (Siegel and Mosler, 2014), the major aim of this paper is to reveal how individuallevel income-related health inequalities measured at the small area level vary with the degree of regional multiple deprivation. 2. METHODS 2.1. Data Data are drawn from the HCAP provided by Kantar Health.1 To obtain sufficiently large numbers of observations throughout all regions, we follow Siegel et al. (2013) and pool the 2002 and 2006 samples of the HCAP; further descriptions of the data can be found elsewhere (Siegel et al., 2013; Potthoff et al., 2004). The pooled sample comprises 117 167 individuals (48 574 households), where 75 122 individuals (29 421 households) participated in 2002 and 60 555 individuals (28 828 households) in 2006, each including the 18 510 individuals (8718 households) who participated in both surveys. Individuals younger than 20 years (28 390) were excluded because chronic conditions rarely affect younger individuals and a meaningful interpretation of the BMI is problematic for children and adolescents. We removed 1176 observations owing to missing data on income and another 7991 owing to unmatched regional codes. The sample eventually comprises 79 610 individuals (41 767 females and 37 843 males) in 43 652 households situated in 5603 communes. Communes are the smallest geographic unit for which data are available in Germany, ranging from small villages to large cities. For more information on German communes see the German Federal Statistical Office’s website (www.regionalstatistik.de) and Maier et al. (2013). 2.2. Variables The modified OECD equivalence scale is applied to compute net equivalent household income in Euro as a measure of individual socioeconomic status (van Doorslaer et al., 2004; Siegel and Mosler, 2014). The first health outcome is obesity defined as a BMI of 30 or higher (WHO, 2014), where the BMI is computed from selfreported anthropometric data as body weight in kilograms divided by the squared body height in meters. The second health outcome is hypertension, where individuals were asked to report whether they had hypertension within the preceding 12 months. The third health outcome is diabetes and was surveyed accordingly. As the surveys do not allow a unique distinction between type 1 and type 2 diabetes, we analyze diabetes regardless of its type or insulin dependency (Siegel et al., 2013). We use the indirect standardization approach (van Doorslaer et al., 2000; O’Donnell et al., 2008) to compensate biases potentially arising from the strong age-dependence of the prevalence of obesity, hypertension, and diabetes (Siegel et al., 2013). Data are stratified by sex and individuals were then grouped by age into five year intervals, each indicated by a dummy variable. The age-predicted risks of obesity, hypertension, and diabetes are obtained through logistic regressions using maximum likelihood estimation. 2.3. Measuring small area deprivation Maier et al. (2012) were the first to apply the notion of multiple deprivation to a German context. They ranked small areas by a set of different deprivation domains following the key principles described by Noble et al. (2006). They used the Bavarian Index of Multiple Deprivation derived for the Federal State of Bavaria 1
Kantar Health was formerly Taylor Nelson Sophres Infratest Healthcare
Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
INDIVIDUAL INCOME, AREA DEPRIVATION, AND HEALTH
1525
(Kuznetsov et al., 2011, 2012; Maier et al., 2012), and extended it to the GIMD for the whole country (Maier et al., 2013, 2014; Koller et al., 2013). The GIMD comprises seven domains, within which the area units are ranked by deprivation in ascending order: income deprivation, employment deprivation, educational deprivation, commune or district revenue deprivation, social capital deprivation, environment deprivation, and security deprivation. The within-domain rank j D j=M for commune j D 1; : : : ; M is exponentially transformed to a (0;100) scale, 100 ; (1) Dj D 23 ln 1 j 1 exp 23 where a higher j indicates higher deprivation. Equation (1) yields transferred domain ranks Dj > 50 only for the most deprived decile .j > 0:9/. This assures that high deprivation in one domain is not easily canceled out by low deprivation in another when computing the overall Index of Multiple Deprivation as a weighted sum of the transformed within-domain scores Dj (Maier et al., 2012; Noble et al., 2006). As higher incomes represent higher socioeconomic status, we require higher ranks to represent better-off communes when comparing income-related and small area deprivation related inequalities. We thus deviate from the existing literature where a higher GIMD indicates more deprivation and use an inverse ranking of areas instead. 2.4. Measuring inequalities Our measure of inequality is the concentration index C (Wagstaff et al., 1991). C stems from the concentration curve, where the cumulative share of some health variable y is plotted against the cumulative share of the population ranked by socioeconomic status. The curve lies below (above) the 45ı line of equality, if y concentrates among the better-off (worse-off). Measuring twice the area between the concentration curve and the 45ı line, C is bounded in the .1I 1/ interval and becomes positive (negative), if the concentration curve lies below (above) the line of equality. The concentration curve equals the line of equality and C is zero where no inequality is observed. Wagstaff and Watanabe (2003) have shown that the difference between two concentration indices C1 and C2 , with rank variables r1 and r2 derived from different socioeconomic indicators, can be written as n
C2 C1 D C D
2 X yi ri ; n
(2)
i D1
where n is the number of observations, is the mean of y and ri D r2i r1i . A modified convenient regression approach 2 2r y D ˇ0 C ˇ1 r C
(3)
2 with r being the variance of r allows one to approximate the standard error C of C D ˇ1 using the ı-method, and thereby to account for the sample variability of (Lindelow, 2006; O’Donnell et al., 2008; Wagstaff and Watanabe, 2003). Note that C D 0 does not necessarily imply that socioeconomic ranks do not differ between two socioeconomic status variables, that is, that ri D 0 for all i. However, C indicates how the choice of the socioeconomic status indicator influences the magnitude of the measured inequality (Wagstaff and Watanabe, 2003). To estimate small area deprivation specific income-related health inequalities, we follow Siegel and Mosler (2014) and estimate
2 Copyright © 2014 John Wiley & Sons, Ltd.
r2 .´/ y D ˇ0 .´/ C ˇ1 .´/r.´/ C .´/
(4) Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
1526
M. SIEGEL
to obtain the concentration index C.´/ D ˇ1 .´/ as a smooth function of ´ 2 Z R (´ denotes the deprivation rank here). With individuals i sorted in ascending order by income, the fractional rank must be computed locally as i X kh .ui / ; (5) ri .´/ D kh´ .uj / ´ 2 j D1 hP i1 P n where ui D ´i ´ and kh´ .ui / D Kh´ .ui / are the kernel weights with niD1 kh´ .ui / D 1. j1 Kh´ .uj / This assures that its local mean and variance are (asymptotically) 1=2 and 1=12, respectively, for any given ´ 2 Z (Siegel and Mosler, 2014). Equation (4) and the ´-specific mean .´/ of y are estimated nonparametrically using a Nadaraya-Watson estimator. The quartic kernel function assigns higher weights kh´ .ui / to observations closer to ´, lower weights to observations further away from ´ and zero weights if an observation is outside the bandwidth. To find an optimal balance between bias and uncertainty at any given ´ 2 Z and avoid boundary effects, the bandwidth parameter h´ is chosen inversely to the local density f´ (Fan and Gijbels, 1992), that is, h´ D 1:06 O ´ n0:1 fO´0:3 , with O ´ being the standard error of ´ obtained from the data. The bounds of C.´/ for binary variables depend inversely on .´/, and different approaches to correct the concentration index have been proposed (Erreygers, 2009; Wagstaff, 2005; , 2011). Allanson and Petrie (2013) and Kjellsson and Gerdtham (2013) argue that choosing a correction method is mainly a normative decision depending on one’s notion of the most unequal society. Alike Siegel and Mosler (2014), we here employ C.´/ Wagstaff’s (2005, 2011) correction formula, W .´/ D 1 , adopting a definition where the most unequal y .´/ society is observed if the poorest (richest) n individuals of the society exhibit the characteristic of interest (e.g. if the ndiabetes D ndiabetes diabetics are the poorest (richest) ndiabetes individuals). We apply a bootstrap approach with 1000 draws to compute pointwise confidence intervals for the corrected varying inequality indices, where the complete estimation process (i.e. including the computation of f´ ; h´ ; r.´/; kh´ .u/; .´/) is performed for each draw. 3. RESULTS Table I presents the age-standardized estimates for the income-related and the multiple deprivation related inequalities of obesity, hypertension, and diabetes computed separately for males and females. The results indicate a highly significant concentration of the health outcomes among the worse-off for both sexes and both socioeconomic indicators. The concentration indices in column (3) suggest that males exhibit a weaker incomerelated concentration of health outcomes among the worse-off than females while no significant sex-specific differences are found for inequalities with respect to the GIMD in column (5). Column (7) in Table I demonstrates that health inequalities with respect to household income are considerably stronger than those with respect to small area deprivation. All differences are highly significant .p < 0:01/ in the female sample. In the male sample, the differences are statistically significant at the 99% level for income and obesity and at the 95% level for diabetes. The choice of the social status variable does not significantly affect the estimated inequality of hypertension among males. The difference between household income related and small area deprivation related health inequalities are significantly smaller in the male sample. Figure 1 presents the deprivation-specific income-related Wagstaff indices for obesity, hypertension, and diabetes (see Figure A1 in the Web Appendix for the corresponding deprivation-specific prevalences). Overall, the figures suggest only small and insignificant variations of income-related health inequalities with varying deprivation ranks. The estimates for the female sample hardly change with the deprivation rank. The results for the male sample in Figure 1 can be summarized as follows: no income-related gradient in the distribution of obesity is observed within the most deprived 10% of the communes among males, income-related inequalities of hypertension among males hardly exist in the worse-off 50% of communes, and the income-related inequality of diabetes among males varies around :1. Computing deprivation-specific concentration and Erreygers indices yielded similar results (see Figure A2 and Figure A3 in the Web Appendix). Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
1527
INDIVIDUAL INCOME, AREA DEPRIVATION, AND HEALTH
Table I. Rank sensitivity: household income and small area deprivation inequalities concerning
(1)
(2)
s.e.b
(3)
(5)
(6)
(7)
(8)
0.0020 0.0071 0.0056 0.0117
–0.2588 0.0245 0.0029 0.0355
0.0020 0.0094 0.0074 0.0153
0.0018 0.0064 0.0056 0.0135
–0.2558 0.0814 0.0309 0.0953
0.0019 0.0084 0.0074 0.0177
(4)
n D 37; 843 1,463.51ec 14.98%e 19.76%e 5.71%e
0.2996d –0.0592 –0.0238 –0.0858
0.0018 0.0069 0.0056 0.0118
0.0408 –0.0347 –0.0267 –0.0503 n D 41; 767
female income obesity hypertension diabetes
sensitivity C2 C1 s.e.b
C2
male income obesity hypertension diabetes
GIMDa
individual income s.e.b C1
1,396.60ec 16.12%e 17.89%e 4.05%e
0.2969d –0.1209 –0.0675 –0.1550
0.0017 0.0062 0.0055 0.0133
0.0411 –0.0395 –0.0366 –0.0597
significant at the 95% level significant at the 99% level a German Index of Multiple Deprivation b standard error c Mean of net equivalent household income d the income-related concentration index C1 of income corresponds to a Gini index of income e prevalence
4. DISCUSSION This paper explored the associations between income-related health inequalities and small area level deprivation. The results suggest that socioeconomic rankings based on both individual socioeconomic status (assessed by household income) and small area deprivation (assessed by GIMD) yield inequalities in obesity, hypertension, and diabetes, which significantly favor the better off. This is clearly in line with the existing literature (van Doorslaer et al., 2004; van Doorslaer and Koolman, 2004; Wagstaff et al., 2003; Jones and López Nicolás, 2006; Schuurman et al., 2007; Walker et al., 2011; Connolly et al., 2000; Maier et al., 2012, 2013; Kuznetsov et al., 2011, 2012; Koller et al., 2013). The significant differences between the income-related and the deprivation-related concentration indices in Table I indicate that using individual-level socioeconomic status yields considerably stronger health inequalities compared with the communes’ socioeconomic status. Maier et al. (2014) argue that regional deprivation and individual socioeconomic status are both important and independent determinant of health. The deprivation-specific Wagstaff indices observed at the commune level suggest that income-related health inequalities also exist within areas of similar multiple deprivation. Although some income-related health inequalities seem somewhat stronger in communes characterized by specific levels of area deprivation, we do not observe clear patterns in the variations with the degree of regional deprivation. The only exemption is obesity among males, where income-related inequalities in the worst-off 10% of communes are considerably smaller than in most better-off communes. For the comparisons of income and GIMD-related inequalities, one should note that the income deprivation domain and net equivalent household income are not fully independent. Although the deprivation index was not computed from the HCAP, household incomes are still included in the GIMD as they are part of the average per tax payer income used as the income deprivation domain (Maier et al., 2012). Chen and Roy (2009) and Clarke and van Ourti (2010) both propose correction formulas for C to account for potential biases involved by repetitive socioeconomic values. Chen and Roy (2009), however, found the empirical bias to be comparably small, and Clarke and van Ourti (2010) found almost no bias for data with more Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
1528
M. SIEGEL
Figure 1. Deprivation-specific Wagstaff indices of obesity, hypertension, and diabetes; deprivation-specific Wagstaff indices W .´/ of income-related inequalities (solid lines) in obesity (top), hypertension (middle) and diabetes (bottom) with quantile-based 95% confidence intervals (dashed lines); commune rank is 0 for the worst-off and 1 for the best-off
than 10 to 15 socioeconomic categories. As we distinguish 326 income values after equivalizing the household income and each commune received a unique deprivation rank, we believe potential biases owing to repetitive values for the socioeconomic indicator to be negligibly small. One may argue that self-reported specific health outcomes would have to be diagnosed by a physician and may thus suffer from a certain bias. As a consequence, one could expect reporting biases where distinct inequalities in health care utilization are observed. However, approximately 90% of the German population contact a physician within a year and Germany is known for its equitable access to health care (van Doorslaer et al., 2004, 2006). Biases owing to inequalities in health care utilization thus seem rather unlikely. The potential of biases arising from social distances between physicians and less educated or lower income patients, however, remains. Considering the results found by Kelly-Irving et al. (2011), one may speculate that this would most likely lead to an underestimation of the concentration among the worse-off. Concerning the results for obesity, it should be mentioned that self-reported anthropometric data may involve some measurement or reporting errors, which are again likely to lead to an underestimation of the prevalence of obesity. 5. CONCLUSIONS We found that individual-level and commune-level socioeconomic deprivation both yield a concentration of health disadvantages among the worse-off. Individual-level socioeconomic status related inequalities were observed within communes regardless of their respective degree of regional multiple deprivation. One may Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
INDIVIDUAL INCOME, AREA DEPRIVATION, AND HEALTH
1529
consider this as support for the notion that area-level socioeconomic status represents a different dimension of socioeconomic status measured at a different level.
CONFLICT OF INTEREST The author has no conflict of interest. ACKNOWLEDGEMENT
Funding by the German Federal Ministry of Education and Research (grant no. 01EH1202A) is gratefully acknowledged. REFERENCES
Allanson P, Petrie D. 2013. Understanding the vertical equity judgements underpinning health inequality measures. Health Economics. DOI: 10.1002/hec.2984. [Epub ahead of print]. Chen Z, Roy K. 2009. Calculating concentration index with repetetive values of indicators of socio-economic welfare. Journal of Health Economics 28(1): 169–175. Clarke P, van Ourti T. 2010. Calculating the concentration index when income is grouped. Journal of Health Economics 29(1): 151–157. Connolly V, Unwin N, Sherriff P, Bilous R, Kelly W. 2000. Diabetes prevalence and socioeconomic status: a population based study showing increased prevalence of type 2 diabetes mellitus in deprived areas. Journal of Epidemiology and Community Health 54(3): 173–177. Erreygers G. 2009. Correcting the concentration index. Journal of Health Economics 28(2): 504–515. Fan J, Gijbels I. 1992. Variable bandwidth and local linear regression smoothers. The Annals of Statistics 20(4): 2008–2036. Jansen L, Eberle A, Emrich K, Gondos A, Holleczek B, Kajüter H, Maier W, Nennecke A, Pritzkuleit R, Brenner H, for the G. E. K. I. D Cancer Survival Working Group. 2014. Socioeconomic deprivation and cancer survival in Germany: An ecological analysis in 200 districts in Germany. International Journal of Cancer 134(12): 2951–2960. Jones A, López Nicolás A. 2006. Allowing for heterogeneity in the decomposition of measures of inequality in health. Journal of Economic Inequality 4(3): 347–365. Kelly-Irving M, Delpierre C, Schieber A-C, Lepage B, Rolland C, Afrité A, Pascal J, Cases C, Lombrail P, Lang T. 2011. Do general practitioners overestimate the health of their patients with lower education? Social Science & Medicine 73(9): 1416–1421. Kjellsson G, Gerdtham U-G. 2013. On correcting the concentration index for binary variables. Joural of Health Economics 32(3): 659–670. Koller D, Hoffmann F, Maier W, Tholen K, Windt R, Glaeske G. 2013. Variation in antibiotic prescriptions: is area deprivation an explanation? Analysis of 1.2 million children in Germany. Infection 41(1): 121–127. Kuznetsov L, Maier W, Hunger M, Meyer M, Mielck A. 2011. Associations between regional socioeconomic deprivation and cancer risk: analysis of population-based cancer registry data from Bavaria, Germany. Preventive Medicine 53(4-5): 328–330. Kuznetsov L, Maier W, Hunger M, Meyer M, Mielck A. 2012. Regional deprivation in Bavaria, Germany: linking a new deprivation score with registry data for lung and colorectal cancer. International Journal of Public Health 57(5): 827–835. Lindelow M. 2006. Sometimes more equal than others: how health inequalities depend on the choice of welfare indicator. Health Economics 15(3): 263–279. Maier W, Fairburn J, Mielck A. 2012. Regional deprivation and mortality in Bavaria. Development of a community-based index of multiple deprivation. Gesundheitswesen 74: 416–425. Maier W, Holle R, Hunger M, Peters A, Meisinger C, Greiser KH, Kluttig A, Völzke H, Schipf S, Moebus S, Bokhof B, Berger K, Mueller G, Rathmann W, Tamayo T. 2013. The impact of regional deprivation and individual socio-economic status on the prevalence of type 2 diabetes in Germany. A pooled analysis of five population-based studies. Diabetic Medicine 30(3): e78–e86. Maier W, Scheidt-Nave C, Holle R, Kroll LE, Lampert T, Du Y, Heidemann C, Mielck A. 2014. Area Level Deprivation Is an Independent Determinant of Prevalent Type 2 Diabetes and Obesity at the National Level in Germany. Results from the National Telephone Health Interview Surveys ’German Health Update’ GEDA 2009 and 2010. PLoS One 9(2): e89661. Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
1530
M. SIEGEL
Noble M, Wright G, Smith G, Dibben C. 2006. Measuring multiple deprivation at the small-area level. Environment and Planning 38(1): 169–185. O’Donnell O, van Doorslaer E, Wagstaff A, Lindelow M. 2008. Analyzing Health Equity using Household Survey Data, WBI Learning Resources Series. World Bank Institute: Washington D.C. Potthoff P, Heinemann LAJ, Güther B. 2004. A household panel as a tool for cost-effective health-related population surveys: Validity of the “Healthcare Access Panel”. German Medical Science 2: 1–8. Schuurman N, Bell N, Dunn JR, Oliver L. 2007. Deprivation indices, population health and geography: an evaluation of the spatial effectiveness of indices at multiple scales. Journal of Urban Health 84(4): 591–603. Siegel M, Lüngen M, Stock S. 2013. On age-specific variations in income-related inequalities in diabetes, hypertension and obesity. International Journal of Public Health 58(1): 33–41. Siegel M, Mosler K. 2014. Semiparametric modeling of age-specific variations in income related health inequalities. Health Economics 23(7): 870–878. Stafford M, Marmot M. 2003. Neighbourhood deprivation and health: does it affect us all equally? International Journal of Epidemiology 32: 357–366. van Doorslaer E, Koolman X. 2004. Explaining the differences in income-related health inequalities across European countries. Health Economics 13(7): 609–628. van Doorslaer E, Koolman X, Jones AM. 2004. Explaining income-related inequalities in doctor utilisation in Europe. Health Economics 13(7): 629–647. van Doorslaer E, Masseria C, Koolman, X for the OECD Health Equity Research Group. 2006. Inequalities in access to medical care by income in developed countries. Canadian Medical Association Journal 174(2): 177–183. van Doorslaer E, Wagstaff A, van der Burg H, Christiansen T, Graeve DD, Duchesne I, Gerdtham U-G, Gerfin M, Geurts J, Gross L, Häkkinen U, John J, Klavus J, Leu RE, Nolan B, O’Donnell O, Propper C, Puffer F, Schellhorn M, Sundberg G, Winkelhake O. 2000. Equity in the delivery of health care in Europe and the US. Journal of Health Economics 19(5): 553–583. Wagstaff A, Paci P, van Doorslaer E. 1991. On the measurement of inequalities in health. Social Science & Medicine 33(5): 545–557. Wagstaff A, van Doorslaer E, Watanabe N. 2003. On decomposing the causes of health sector inequalities with an application to malnutrition inequalities in Vietnam. Journal of Econometrics 112(1): 207–223. Wagstaff A. 2005. The bounds of the concentration index when the variable of interest is binary, with an application to immunization inequality. Health Economics 14(4): 429–432. Wagstaff A. 2011. The concentration index of a binary outcome revisited. Health Economics 20(10): 1155–1160. Wagstaff A, Watanabe N. 2003. What difference does the choice of ses make in health inequality measurement? Health Economics 12(10): 885–890. Walker JJ, Livingstone SJ, Colhoun HM, Lindsay RS, McKnight JA, Morris AD, Petrie JR, Philip S, Sattar N, Wild SH, on behalf on the Scottish Diabetes Research Network Epidemiology Group. 2011. Effect of socioeconomic status on mortality among people with type 2 diabetes: a study from the Scottish Diabetes Research Network Epidemiology Group. Diabetes Care 34(5): 1127–1132. WHO. 2014. Global database on body mass index: BMI classification. The World Health Organization. (Available from: http://apps.who.int/bmi/index.jsp?introPage=intro_3.html) [Accessed on 21 April 2014].
SUPPORTING INFORMATION Additional supporting information may be found in the online version of this article at the publisher’s web-site.
Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
HEALTH ECONOMICS LETTER
INDIVIDUAL INCOME, AREA DEPRIVATION, AND HEALTH: DO INCOME-RELATED HEALTH INEQUALITIES VARY BY SMALL AREA DEPRIVATION? MARTIN SIEGELa,b, , ANDREAS MIELCKc and WERNER MAIERc a Berlin
Centre of Health Economics Research (BerlinHECOR), Department of Health Care Management, Technische Universität Berlin, Berlin, Germany b Institute c Institute
of Health Economics and Clinical Epidemiology, University Hospital of Cologne, Germany
of Health Economics and Health Care Management, Helmholtz Zentrum München, Neuherberg, Germany
ABSTRACT This paper aims to explore potential associations between health inequalities related to socioeconomic deprivation at the individual and the small area level. We use German cross-sectional survey data for the years 2002 and 2006, and measure small area deprivation via the German Index of Multiple Deprivation. We test the differences between concentration indices of income-related and small area deprivation related inequalities in obesity, hypertension, and diabetes. Our results suggest that small area deprivation and individual income both yield inequalities in health favoring the better-off, where individual income-related inequalities are significantly more pronounced than those related to small area deprivation. We then apply a semiparametric extension of Wagstaff’s corrected concentration index to explore how individual-level health inequalities vary with the degree of regional deprivation. We find that the concentration of obesity, hypertension, and diabetes among lower income groups also exists at the small area level. The degree of deprivation-specific income-related inequalities in the three health outcomes exhibits only little variations across different levels of multiple deprivation for both sexes. Copyright © 2014 John Wiley & Sons, Ltd. Received 17 September 2013; Revised 21 July 2014; Accepted 8 August 2014 KEY WORDS:
regional deprivation; income; health inequalities; concentration index
1. INTRODUCTION The existence of income-related health inequalities favoring the better-off is well established in the literature (Wagstaff et al., 2003; Jones and López Nicolás, 2006; van Doorslaer et al., 2004; van Doorslaer and Koolman, 2004). Stafford and Marmot (2003), for instance, further argue that potential risks associated with living in deprived areas may reinforce the association between individual socioeconomic position and health, and more recent research suggests that regional deprivation may be an important determinant for the distribution of health (Schuurman et al., 2007; Walker et al., 2011; Connolly et al., 2000; Maier et al., 2012, 2013, 2014; Kuznetsov et al., 2011, 2012; Koller et al., 2013; Jansen et al., 2014). Following Maier et al. (2014) who argue that small area deprivation and individual-level socioeconomic status are independent determinants of health, one may expect inequalities with respect to individual incomes to exist at the small area level. Using a pooled sample from the 2002 and 2006 waves of the Taylor Nelson Sophres Health Care Access Panel (HCAP), this is the first paper measuring variations of income-related health inequalities across different levels of small area deprivation. We take obesity, hypertension, and diabetes as
Correspondence to: Department of Health Care Management, Technische Universität Berlin. E-mail: [email protected]
Copyright © 2014 John Wiley & Sons, Ltd.
1524
M. SIEGEL
an example and apply the German Index of Multiple Deprivation (GIMD) to measure small area deprivation at the commune level (Maier et al., 2012, 2013). Both household income related and small area deprivation related health inequalities are measured using concentration indices (Wagstaff et al., 1991). We first measure the differences between them following Wagstaff and Watanabe (2003) to explore to what extent inequalities with respect to the two indicators differ in our data. Using a semiparametric extension of Wagstaff’s corrected concentration index (Siegel and Mosler, 2014), the major aim of this paper is to reveal how individuallevel income-related health inequalities measured at the small area level vary with the degree of regional multiple deprivation. 2. METHODS 2.1. Data Data are drawn from the HCAP provided by Kantar Health.1 To obtain sufficiently large numbers of observations throughout all regions, we follow Siegel et al. (2013) and pool the 2002 and 2006 samples of the HCAP; further descriptions of the data can be found elsewhere (Siegel et al., 2013; Potthoff et al., 2004). The pooled sample comprises 117 167 individuals (48 574 households), where 75 122 individuals (29 421 households) participated in 2002 and 60 555 individuals (28 828 households) in 2006, each including the 18 510 individuals (8718 households) who participated in both surveys. Individuals younger than 20 years (28 390) were excluded because chronic conditions rarely affect younger individuals and a meaningful interpretation of the BMI is problematic for children and adolescents. We removed 1176 observations owing to missing data on income and another 7991 owing to unmatched regional codes. The sample eventually comprises 79 610 individuals (41 767 females and 37 843 males) in 43 652 households situated in 5603 communes. Communes are the smallest geographic unit for which data are available in Germany, ranging from small villages to large cities. For more information on German communes see the German Federal Statistical Office’s website (www.regionalstatistik.de) and Maier et al. (2013). 2.2. Variables The modified OECD equivalence scale is applied to compute net equivalent household income in Euro as a measure of individual socioeconomic status (van Doorslaer et al., 2004; Siegel and Mosler, 2014). The first health outcome is obesity defined as a BMI of 30 or higher (WHO, 2014), where the BMI is computed from selfreported anthropometric data as body weight in kilograms divided by the squared body height in meters. The second health outcome is hypertension, where individuals were asked to report whether they had hypertension within the preceding 12 months. The third health outcome is diabetes and was surveyed accordingly. As the surveys do not allow a unique distinction between type 1 and type 2 diabetes, we analyze diabetes regardless of its type or insulin dependency (Siegel et al., 2013). We use the indirect standardization approach (van Doorslaer et al., 2000; O’Donnell et al., 2008) to compensate biases potentially arising from the strong age-dependence of the prevalence of obesity, hypertension, and diabetes (Siegel et al., 2013). Data are stratified by sex and individuals were then grouped by age into five year intervals, each indicated by a dummy variable. The age-predicted risks of obesity, hypertension, and diabetes are obtained through logistic regressions using maximum likelihood estimation. 2.3. Measuring small area deprivation Maier et al. (2012) were the first to apply the notion of multiple deprivation to a German context. They ranked small areas by a set of different deprivation domains following the key principles described by Noble et al. (2006). They used the Bavarian Index of Multiple Deprivation derived for the Federal State of Bavaria 1
Kantar Health was formerly Taylor Nelson Sophres Infratest Healthcare
Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
INDIVIDUAL INCOME, AREA DEPRIVATION, AND HEALTH
1525
(Kuznetsov et al., 2011, 2012; Maier et al., 2012), and extended it to the GIMD for the whole country (Maier et al., 2013, 2014; Koller et al., 2013). The GIMD comprises seven domains, within which the area units are ranked by deprivation in ascending order: income deprivation, employment deprivation, educational deprivation, commune or district revenue deprivation, social capital deprivation, environment deprivation, and security deprivation. The within-domain rank j D j=M for commune j D 1; : : : ; M is exponentially transformed to a (0;100) scale, 100 ; (1) Dj D 23 ln 1 j 1 exp 23 where a higher j indicates higher deprivation. Equation (1) yields transferred domain ranks Dj > 50 only for the most deprived decile .j > 0:9/. This assures that high deprivation in one domain is not easily canceled out by low deprivation in another when computing the overall Index of Multiple Deprivation as a weighted sum of the transformed within-domain scores Dj (Maier et al., 2012; Noble et al., 2006). As higher incomes represent higher socioeconomic status, we require higher ranks to represent better-off communes when comparing income-related and small area deprivation related inequalities. We thus deviate from the existing literature where a higher GIMD indicates more deprivation and use an inverse ranking of areas instead. 2.4. Measuring inequalities Our measure of inequality is the concentration index C (Wagstaff et al., 1991). C stems from the concentration curve, where the cumulative share of some health variable y is plotted against the cumulative share of the population ranked by socioeconomic status. The curve lies below (above) the 45ı line of equality, if y concentrates among the better-off (worse-off). Measuring twice the area between the concentration curve and the 45ı line, C is bounded in the .1I 1/ interval and becomes positive (negative), if the concentration curve lies below (above) the line of equality. The concentration curve equals the line of equality and C is zero where no inequality is observed. Wagstaff and Watanabe (2003) have shown that the difference between two concentration indices C1 and C2 , with rank variables r1 and r2 derived from different socioeconomic indicators, can be written as n
C2 C1 D C D
2 X yi ri ; n
(2)
i D1
where n is the number of observations, is the mean of y and ri D r2i r1i . A modified convenient regression approach 2 2r y D ˇ0 C ˇ1 r C
(3)
2 with r being the variance of r allows one to approximate the standard error C of C D ˇ1 using the ı-method, and thereby to account for the sample variability of (Lindelow, 2006; O’Donnell et al., 2008; Wagstaff and Watanabe, 2003). Note that C D 0 does not necessarily imply that socioeconomic ranks do not differ between two socioeconomic status variables, that is, that ri D 0 for all i. However, C indicates how the choice of the socioeconomic status indicator influences the magnitude of the measured inequality (Wagstaff and Watanabe, 2003). To estimate small area deprivation specific income-related health inequalities, we follow Siegel and Mosler (2014) and estimate
2 Copyright © 2014 John Wiley & Sons, Ltd.
r2 .´/ y D ˇ0 .´/ C ˇ1 .´/r.´/ C .´/
(4) Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
1526
M. SIEGEL
to obtain the concentration index C.´/ D ˇ1 .´/ as a smooth function of ´ 2 Z R (´ denotes the deprivation rank here). With individuals i sorted in ascending order by income, the fractional rank must be computed locally as i X kh .ui / ; (5) ri .´/ D kh´ .uj / ´ 2 j D1 hP i1 P n where ui D ´i ´ and kh´ .ui / D Kh´ .ui / are the kernel weights with niD1 kh´ .ui / D 1. j1 Kh´ .uj / This assures that its local mean and variance are (asymptotically) 1=2 and 1=12, respectively, for any given ´ 2 Z (Siegel and Mosler, 2014). Equation (4) and the ´-specific mean .´/ of y are estimated nonparametrically using a Nadaraya-Watson estimator. The quartic kernel function assigns higher weights kh´ .ui / to observations closer to ´, lower weights to observations further away from ´ and zero weights if an observation is outside the bandwidth. To find an optimal balance between bias and uncertainty at any given ´ 2 Z and avoid boundary effects, the bandwidth parameter h´ is chosen inversely to the local density f´ (Fan and Gijbels, 1992), that is, h´ D 1:06 O ´ n0:1 fO´0:3 , with O ´ being the standard error of ´ obtained from the data. The bounds of C.´/ for binary variables depend inversely on .´/, and different approaches to correct the concentration index have been proposed (Erreygers, 2009; Wagstaff, 2005; , 2011). Allanson and Petrie (2013) and Kjellsson and Gerdtham (2013) argue that choosing a correction method is mainly a normative decision depending on one’s notion of the most unequal society. Alike Siegel and Mosler (2014), we here employ C.´/ Wagstaff’s (2005, 2011) correction formula, W .´/ D 1 , adopting a definition where the most unequal y .´/ society is observed if the poorest (richest) n individuals of the society exhibit the characteristic of interest (e.g. if the ndiabetes D ndiabetes diabetics are the poorest (richest) ndiabetes individuals). We apply a bootstrap approach with 1000 draws to compute pointwise confidence intervals for the corrected varying inequality indices, where the complete estimation process (i.e. including the computation of f´ ; h´ ; r.´/; kh´ .u/; .´/) is performed for each draw. 3. RESULTS Table I presents the age-standardized estimates for the income-related and the multiple deprivation related inequalities of obesity, hypertension, and diabetes computed separately for males and females. The results indicate a highly significant concentration of the health outcomes among the worse-off for both sexes and both socioeconomic indicators. The concentration indices in column (3) suggest that males exhibit a weaker incomerelated concentration of health outcomes among the worse-off than females while no significant sex-specific differences are found for inequalities with respect to the GIMD in column (5). Column (7) in Table I demonstrates that health inequalities with respect to household income are considerably stronger than those with respect to small area deprivation. All differences are highly significant .p < 0:01/ in the female sample. In the male sample, the differences are statistically significant at the 99% level for income and obesity and at the 95% level for diabetes. The choice of the social status variable does not significantly affect the estimated inequality of hypertension among males. The difference between household income related and small area deprivation related health inequalities are significantly smaller in the male sample. Figure 1 presents the deprivation-specific income-related Wagstaff indices for obesity, hypertension, and diabetes (see Figure A1 in the Web Appendix for the corresponding deprivation-specific prevalences). Overall, the figures suggest only small and insignificant variations of income-related health inequalities with varying deprivation ranks. The estimates for the female sample hardly change with the deprivation rank. The results for the male sample in Figure 1 can be summarized as follows: no income-related gradient in the distribution of obesity is observed within the most deprived 10% of the communes among males, income-related inequalities of hypertension among males hardly exist in the worse-off 50% of communes, and the income-related inequality of diabetes among males varies around :1. Computing deprivation-specific concentration and Erreygers indices yielded similar results (see Figure A2 and Figure A3 in the Web Appendix). Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
1527
INDIVIDUAL INCOME, AREA DEPRIVATION, AND HEALTH
Table I. Rank sensitivity: household income and small area deprivation inequalities concerning
(1)
(2)
s.e.b
(3)
(5)
(6)
(7)
(8)
0.0020 0.0071 0.0056 0.0117
–0.2588 0.0245 0.0029 0.0355
0.0020 0.0094 0.0074 0.0153
0.0018 0.0064 0.0056 0.0135
–0.2558 0.0814 0.0309 0.0953
0.0019 0.0084 0.0074 0.0177
(4)
n D 37; 843 1,463.51ec 14.98%e 19.76%e 5.71%e
0.2996d –0.0592 –0.0238 –0.0858
0.0018 0.0069 0.0056 0.0118
0.0408 –0.0347 –0.0267 –0.0503 n D 41; 767
female income obesity hypertension diabetes
sensitivity C2 C1 s.e.b
C2
male income obesity hypertension diabetes
GIMDa
individual income s.e.b C1
1,396.60ec 16.12%e 17.89%e 4.05%e
0.2969d –0.1209 –0.0675 –0.1550
0.0017 0.0062 0.0055 0.0133
0.0411 –0.0395 –0.0366 –0.0597
significant at the 95% level significant at the 99% level a German Index of Multiple Deprivation b standard error c Mean of net equivalent household income d the income-related concentration index C1 of income corresponds to a Gini index of income e prevalence
4. DISCUSSION This paper explored the associations between income-related health inequalities and small area level deprivation. The results suggest that socioeconomic rankings based on both individual socioeconomic status (assessed by household income) and small area deprivation (assessed by GIMD) yield inequalities in obesity, hypertension, and diabetes, which significantly favor the better off. This is clearly in line with the existing literature (van Doorslaer et al., 2004; van Doorslaer and Koolman, 2004; Wagstaff et al., 2003; Jones and López Nicolás, 2006; Schuurman et al., 2007; Walker et al., 2011; Connolly et al., 2000; Maier et al., 2012, 2013; Kuznetsov et al., 2011, 2012; Koller et al., 2013). The significant differences between the income-related and the deprivation-related concentration indices in Table I indicate that using individual-level socioeconomic status yields considerably stronger health inequalities compared with the communes’ socioeconomic status. Maier et al. (2014) argue that regional deprivation and individual socioeconomic status are both important and independent determinant of health. The deprivation-specific Wagstaff indices observed at the commune level suggest that income-related health inequalities also exist within areas of similar multiple deprivation. Although some income-related health inequalities seem somewhat stronger in communes characterized by specific levels of area deprivation, we do not observe clear patterns in the variations with the degree of regional deprivation. The only exemption is obesity among males, where income-related inequalities in the worst-off 10% of communes are considerably smaller than in most better-off communes. For the comparisons of income and GIMD-related inequalities, one should note that the income deprivation domain and net equivalent household income are not fully independent. Although the deprivation index was not computed from the HCAP, household incomes are still included in the GIMD as they are part of the average per tax payer income used as the income deprivation domain (Maier et al., 2012). Chen and Roy (2009) and Clarke and van Ourti (2010) both propose correction formulas for C to account for potential biases involved by repetitive socioeconomic values. Chen and Roy (2009), however, found the empirical bias to be comparably small, and Clarke and van Ourti (2010) found almost no bias for data with more Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
1528
M. SIEGEL
Figure 1. Deprivation-specific Wagstaff indices of obesity, hypertension, and diabetes; deprivation-specific Wagstaff indices W .´/ of income-related inequalities (solid lines) in obesity (top), hypertension (middle) and diabetes (bottom) with quantile-based 95% confidence intervals (dashed lines); commune rank is 0 for the worst-off and 1 for the best-off
than 10 to 15 socioeconomic categories. As we distinguish 326 income values after equivalizing the household income and each commune received a unique deprivation rank, we believe potential biases owing to repetitive values for the socioeconomic indicator to be negligibly small. One may argue that self-reported specific health outcomes would have to be diagnosed by a physician and may thus suffer from a certain bias. As a consequence, one could expect reporting biases where distinct inequalities in health care utilization are observed. However, approximately 90% of the German population contact a physician within a year and Germany is known for its equitable access to health care (van Doorslaer et al., 2004, 2006). Biases owing to inequalities in health care utilization thus seem rather unlikely. The potential of biases arising from social distances between physicians and less educated or lower income patients, however, remains. Considering the results found by Kelly-Irving et al. (2011), one may speculate that this would most likely lead to an underestimation of the concentration among the worse-off. Concerning the results for obesity, it should be mentioned that self-reported anthropometric data may involve some measurement or reporting errors, which are again likely to lead to an underestimation of the prevalence of obesity. 5. CONCLUSIONS We found that individual-level and commune-level socioeconomic deprivation both yield a concentration of health disadvantages among the worse-off. Individual-level socioeconomic status related inequalities were observed within communes regardless of their respective degree of regional multiple deprivation. One may Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
INDIVIDUAL INCOME, AREA DEPRIVATION, AND HEALTH
1529
consider this as support for the notion that area-level socioeconomic status represents a different dimension of socioeconomic status measured at a different level.
CONFLICT OF INTEREST The author has no conflict of interest. ACKNOWLEDGEMENT
Funding by the German Federal Ministry of Education and Research (grant no. 01EH1202A) is gratefully acknowledged. REFERENCES
Allanson P, Petrie D. 2013. Understanding the vertical equity judgements underpinning health inequality measures. Health Economics. DOI: 10.1002/hec.2984. [Epub ahead of print]. Chen Z, Roy K. 2009. Calculating concentration index with repetetive values of indicators of socio-economic welfare. Journal of Health Economics 28(1): 169–175. Clarke P, van Ourti T. 2010. Calculating the concentration index when income is grouped. Journal of Health Economics 29(1): 151–157. Connolly V, Unwin N, Sherriff P, Bilous R, Kelly W. 2000. Diabetes prevalence and socioeconomic status: a population based study showing increased prevalence of type 2 diabetes mellitus in deprived areas. Journal of Epidemiology and Community Health 54(3): 173–177. Erreygers G. 2009. Correcting the concentration index. Journal of Health Economics 28(2): 504–515. Fan J, Gijbels I. 1992. Variable bandwidth and local linear regression smoothers. The Annals of Statistics 20(4): 2008–2036. Jansen L, Eberle A, Emrich K, Gondos A, Holleczek B, Kajüter H, Maier W, Nennecke A, Pritzkuleit R, Brenner H, for the G. E. K. I. D Cancer Survival Working Group. 2014. Socioeconomic deprivation and cancer survival in Germany: An ecological analysis in 200 districts in Germany. International Journal of Cancer 134(12): 2951–2960. Jones A, López Nicolás A. 2006. Allowing for heterogeneity in the decomposition of measures of inequality in health. Journal of Economic Inequality 4(3): 347–365. Kelly-Irving M, Delpierre C, Schieber A-C, Lepage B, Rolland C, Afrité A, Pascal J, Cases C, Lombrail P, Lang T. 2011. Do general practitioners overestimate the health of their patients with lower education? Social Science & Medicine 73(9): 1416–1421. Kjellsson G, Gerdtham U-G. 2013. On correcting the concentration index for binary variables. Joural of Health Economics 32(3): 659–670. Koller D, Hoffmann F, Maier W, Tholen K, Windt R, Glaeske G. 2013. Variation in antibiotic prescriptions: is area deprivation an explanation? Analysis of 1.2 million children in Germany. Infection 41(1): 121–127. Kuznetsov L, Maier W, Hunger M, Meyer M, Mielck A. 2011. Associations between regional socioeconomic deprivation and cancer risk: analysis of population-based cancer registry data from Bavaria, Germany. Preventive Medicine 53(4-5): 328–330. Kuznetsov L, Maier W, Hunger M, Meyer M, Mielck A. 2012. Regional deprivation in Bavaria, Germany: linking a new deprivation score with registry data for lung and colorectal cancer. International Journal of Public Health 57(5): 827–835. Lindelow M. 2006. Sometimes more equal than others: how health inequalities depend on the choice of welfare indicator. Health Economics 15(3): 263–279. Maier W, Fairburn J, Mielck A. 2012. Regional deprivation and mortality in Bavaria. Development of a community-based index of multiple deprivation. Gesundheitswesen 74: 416–425. Maier W, Holle R, Hunger M, Peters A, Meisinger C, Greiser KH, Kluttig A, Völzke H, Schipf S, Moebus S, Bokhof B, Berger K, Mueller G, Rathmann W, Tamayo T. 2013. The impact of regional deprivation and individual socio-economic status on the prevalence of type 2 diabetes in Germany. A pooled analysis of five population-based studies. Diabetic Medicine 30(3): e78–e86. Maier W, Scheidt-Nave C, Holle R, Kroll LE, Lampert T, Du Y, Heidemann C, Mielck A. 2014. Area Level Deprivation Is an Independent Determinant of Prevalent Type 2 Diabetes and Obesity at the National Level in Germany. Results from the National Telephone Health Interview Surveys ’German Health Update’ GEDA 2009 and 2010. PLoS One 9(2): e89661. Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
1530
M. SIEGEL
Noble M, Wright G, Smith G, Dibben C. 2006. Measuring multiple deprivation at the small-area level. Environment and Planning 38(1): 169–185. O’Donnell O, van Doorslaer E, Wagstaff A, Lindelow M. 2008. Analyzing Health Equity using Household Survey Data, WBI Learning Resources Series. World Bank Institute: Washington D.C. Potthoff P, Heinemann LAJ, Güther B. 2004. A household panel as a tool for cost-effective health-related population surveys: Validity of the “Healthcare Access Panel”. German Medical Science 2: 1–8. Schuurman N, Bell N, Dunn JR, Oliver L. 2007. Deprivation indices, population health and geography: an evaluation of the spatial effectiveness of indices at multiple scales. Journal of Urban Health 84(4): 591–603. Siegel M, Lüngen M, Stock S. 2013. On age-specific variations in income-related inequalities in diabetes, hypertension and obesity. International Journal of Public Health 58(1): 33–41. Siegel M, Mosler K. 2014. Semiparametric modeling of age-specific variations in income related health inequalities. Health Economics 23(7): 870–878. Stafford M, Marmot M. 2003. Neighbourhood deprivation and health: does it affect us all equally? International Journal of Epidemiology 32: 357–366. van Doorslaer E, Koolman X. 2004. Explaining the differences in income-related health inequalities across European countries. Health Economics 13(7): 609–628. van Doorslaer E, Koolman X, Jones AM. 2004. Explaining income-related inequalities in doctor utilisation in Europe. Health Economics 13(7): 629–647. van Doorslaer E, Masseria C, Koolman, X for the OECD Health Equity Research Group. 2006. Inequalities in access to medical care by income in developed countries. Canadian Medical Association Journal 174(2): 177–183. van Doorslaer E, Wagstaff A, van der Burg H, Christiansen T, Graeve DD, Duchesne I, Gerdtham U-G, Gerfin M, Geurts J, Gross L, Häkkinen U, John J, Klavus J, Leu RE, Nolan B, O’Donnell O, Propper C, Puffer F, Schellhorn M, Sundberg G, Winkelhake O. 2000. Equity in the delivery of health care in Europe and the US. Journal of Health Economics 19(5): 553–583. Wagstaff A, Paci P, van Doorslaer E. 1991. On the measurement of inequalities in health. Social Science & Medicine 33(5): 545–557. Wagstaff A, van Doorslaer E, Watanabe N. 2003. On decomposing the causes of health sector inequalities with an application to malnutrition inequalities in Vietnam. Journal of Econometrics 112(1): 207–223. Wagstaff A. 2005. The bounds of the concentration index when the variable of interest is binary, with an application to immunization inequality. Health Economics 14(4): 429–432. Wagstaff A. 2011. The concentration index of a binary outcome revisited. Health Economics 20(10): 1155–1160. Wagstaff A, Watanabe N. 2003. What difference does the choice of ses make in health inequality measurement? Health Economics 12(10): 885–890. Walker JJ, Livingstone SJ, Colhoun HM, Lindsay RS, McKnight JA, Morris AD, Petrie JR, Philip S, Sattar N, Wild SH, on behalf on the Scottish Diabetes Research Network Epidemiology Group. 2011. Effect of socioeconomic status on mortality among people with type 2 diabetes: a study from the Scottish Diabetes Research Network Epidemiology Group. Diabetes Care 34(5): 1127–1132. WHO. 2014. Global database on body mass index: BMI classification. The World Health Organization. (Available from: http://apps.who.int/bmi/index.jsp?introPage=intro_3.html) [Accessed on 21 April 2014].
SUPPORTING INFORMATION Additional supporting information may be found in the online version of this article at the publisher’s web-site.
Copyright © 2014 John Wiley & Sons, Ltd.
Health Econ. 24: 1523–1530 (2015) DOI: 10.1002/hec
