Module 1a: Inequalities and inequities in health and health care utilization Concentration curve and concentration index This presentation was prepared by Adam Wagstaff, Caryn Bredenkamp and Sarah Bales 1
The basic idea
The basic idea The poor typically lag behind the better off in terms of health outcomes and utilization of health services Policymakers would like to track progress is the gap narrowing? and see how their country compares to other countries Data are often presented in terms of economic or socioeconomic groups With several groups, it s not easy to see how inequalities compare or have changed
U5MR per 1000 live births In which country are child deaths distributed most unequally? 300 250 200 150 Poorest "quintile" 2nd poorest "quintile Middle "quintile" 100 50 0 India Mali 2nd richest "quintile" Richest "quintile" Rate ratios can be used: but don t consider how skewed the distribution is in the middle quintiles Comparison made difficult by differences in average levels
Let s get measuring!
Cumulative % of illness Illness concentration curve Here inequality disfavors the poor: they bear a greater share of illness than their share in the population 75% of disease burden The further the CC is from the line of equality, the greater the inequality! Poorest 50% of population Cumulative % population, ranked in ascending order of income, wealth, etc.
cumul % under-5 deaths 100% Comparing too many concentration curves is bad for your eyes! 80% 60% 40% Equality Brazil Cote d'ivoire Ghana Nepal Nicaragua Pakistan Cebu S Africa Vietnam 20% 0% 0% 20% 40% 60% 80% 100% cumul % live births, ranked by equiv consumption Brazil is most unequal, but how do the rest compare?
Cumulative proportion of illness The concentration index is a useful tie-breaker 75% of disease burden Concentration index (CI) = 2 x shaded area CI lies in range (-1,1) CI < 0 because variable is concentrated among the poor Poorest 50% of population
Cumulative proportion of illness The case where inequalities in illness favor the poor Concentration index (=2 x shaded area, as before) is positive in this case because variable is concentrated among the better off 25% of disease burden Here inequality favors the poor: they bear a smaller share of illness than their share in the population Poorest 50% of population
Beware! A negative CI doesn t necessarily imply bad outcomes for the poor. It depends on whether the health variable being analyzed is a good outcome or a bad outcome.
How to do it in ADePT?
What ADePT does ADePT uses the living standards variable to rank individuals and create population quintiles (using household weights) It then produces a table showing the distribution of the outcomes across quintiles and calculates the concentration indices It also charts the concentration curves
What ADePT asks for Health outcomes, for example: Under 1 mortality rate Height-for-age z-score Stunting prevalence Diagnosed with diabetes Self assessed health (dichotomous, not categorical) Living standards measure - continuous variable (e.g. Consumption, expenditure, income, asset index or score) Weights and survey settings relate to sample design information (sampling weight, cluster, strata) Household ID 13
14
ZAMBIA (WHS) 15
2) Select the dataset you want to use. NOTE: It is assumed the observation is for individuals not households. CAMBODIAN DHS Data can be in STATA or SPSS format 16
If you have STATA installed, you can open your dataset in STATA to explore and understand it better 3) Enter label here 17
4) Select Household ID from menu. The household ID can be more than one variable (e.g. provcode districtcode hhcode) 18
5) Select living standards measure (continuous variable) NOTE: should be in per capita terms if income, expenditure or consumption. 19
7) Click on Sampling weight variable then select weight variable from menu. NOTE: Weights will be further discussed below. 6) Click to add sample design information (cluster, strata, weight, ) 20
8) Select health outcome variable(s) from menu. Pay attention: Only one space allowed between variables. 21
9) If necessary, select only the observations relevant to the analysis (e.g. people age 18 and older) 22
11 G to ta 10) Select results tables desired including Original Data Report to check on summary statistics of inputs to analysis 23
Results of checking data, any errors Summary statistics of all variables Health inequality results 24
364 c were d Only ca olde 25
Original Data Report ZAMBIA N mean min max hhid (Household ID) 3,802 35588002 11603424 35660907 hhexpcap (Welfare aggregate) 3,797 95,727.0 0.0 3,276,952.0 pweight (Household weights) 3,801 1,018.4 63 32,137 asthma (Outcome 1) 3,802 0.106 0 1 TB (Outcome 2) 3,791 0.142 0 1 badsah (Outcome 3) 3,801 0.070 0 1 woverweight (Outcome 4) 1,289 0.319 0 1 traffic_accident (Outcome 5) 3,787 0.013 0 1 generated (Household size) 3,802 1 1 1 Number of observations Note: only 1 observation per household 26
Table H3: Health inequality, unstandardized Selfassessed bad health Asthma TB symptoms Woman with BMI over 25 (overweight) Road traffic accident in past 12 months Quintiles of total household expenditures per capita in 4wks preceding survey Lowest quintile 0.0916 0.1169 0.1737 0.2634 0.0100 2 0.0714 0.1330 0.1423 0.2903 0.0064 3 0.0718 0.1240 0.1373 0.2528 0.0119 4 0.0676 0.0742 0.1314 0.2948 0.0158 Highest quintile 0.0308 0.0834 0.1175 0.4842 0.0332 Total 0.0666 0.1063 0.1405 0.3188 0.0155 Standard concentration index -0.1722-0.1029-0.0763 0.1110 0.3104 Examine the differences in concentration indices and distributions across quintiles 27
Cumulative % of outcome variable 100 Concentration curves Line of equality 80 60 40 20 adults self assessed health is bad or very bad (1/0) What can we say about the health status of adults in the poorest quintile? woman with BMI above 25 (1/0) What can we say about the BMI of women in the wealthiest quintile? 0 0 20 40 60 80 100 Cumulative % of population, ranked from poorest to richest 28
Presenting your results to policymakers
cumul % under-five deaths In which country are child deaths distributed most unequally? 100% 80% Child deaths are disproportionately concentrated among the poor in both countries 60% 40% 20% Line of equality India Mali But the disproportionate concentration (inequality) appears greater in India 0% 0% 20% 40% 60% 80% 100% cumul % births, ranked by wealth
Vietnam 1982-93 Pakistan 1981-90 Ghana 1978-89 Cote d Ivoire 1978-89 Nepal 1985-96 South Africa 1985-89 Phillipines (Cebu) 1981-91 Nicaragua 1983-88 Brazil (NE & SE) 1987-92 C and 95% conf interval Concentration indices for U5MR 0.1 0.0-0.1-0.2-0.3-0.4-0.5 31
Policy levers-i The distribution of health or utilization across income or wealth groups depends on: 1. How unequally their determinants are distributed, and 2. The importance (i.e. size of effect) of different determinants on health or utilization. A determinant that s very unequally distributed and matters a lot for health or utilization will account for a large share of the inequality in health or utilization
Policy levers-ii Take the example of distance Policymakers could: 1. Reduce inequality in distance, e.g. by spreading facilities more evenly, and/or 2. Make distance matter less, e.g. by paying a transportation allowance for people coming from far-flung villages, and compensating them for lost earnings
Policy levers-iii Examples of programs that reduce inequalities in key determinants: Brazil s Family Health program takes health care to poor communities through aggressive outreach, reducing inequalities in distance to provider Examples of programs that make key determinants matter less: Cambodia Health Equity Funds and Vietnam s Health Care for the Poor scheme eliminate user fees for the poor (but make sure providers still get paid!), making income matter less An example of a campaign that did both: Kenya s immunization campaign held immunization camps in poor areas, reducing distance to provider; and Raised awareness about vaccination, making formal education matter less for health
Where to go from here?
Data sources for CC and CI analysis Demographic and Health Surveys (DHS) World Health Survey (WHS) Living standards measurement survey (LSMS) Multi-indicator cluster survey (MICS) Other household surveys 36
Know your data It is important to read the questionnaire and codebook and examine your data before doing your analysis!!! 37
Know your data Datasets may not contain ready-to-use measures of living standards. Data sets cover different populations of interest and merging of files may be required to create meaningful distributions, i.e. population quintiles The data may require sampling weights (pweight) to be nationally representative. 38
Related materials Guide to methods: Analyzing Health Equity Using Household Survey Data ADePT Health Manual: Health Equity and Financial Protection Online video tutorials Health Equity and Financial Protection reports (ongoing) Health Equity and Financial Protection datasheets (ongoing) Book Attacking Inequality in the Health Sector Training events www.worldbank.org/povertyandhealth and www.worldbank.org/adept