Group de Análisis para el Desarrollo Measuring and Monitoring Health Equity Martín Valdivia Dakha, Bangladesh May 2005
Basic ideas for monitoring health equity: What do we need? In operational terms, we need a relevant stratifier, a healthrelated outcome, and an inequality measure to follow a stratifier: income/expenditures, wealth, gender, education, race/ethnicity, age, health condition health-related outcome types: health status, access and utilization of health care, health financing, allocation of public resources inequality measures: poor/rich ratio, concentration index We also need a way to connect actions to health inequity outcomes assign responsibility promote changes
Basic ideas for monitoring health equity: What do we need? In practice, we first and mostly need a deep understanding of the society under consideration knowledge of mechanisms that generate and perpetuate social inequalities functioning of health system Also, we need to know about most relevant stratifiers, knowledge of disadvantaged, excluded groups epidemiological situation, for instance, related to the demographic transition and, some knowledge of key advantages behind most common inequality measures
Which are the most common stratifiers? The list includes Socio-economic status (SES) education race/ethnicity gender age health condition, e.g., TBC, HIV, etc. They are all related but each may reveal specific information regarding the nature of social inequalities in a particular society
Which are the most common stratifiers? They can be measured at the different levels individual, household local communities regions, countries What are the data sources? Surveys censuses facility-based data
Measuring SES Usual indicators are income, consumption/expenditure, wealth income: includes labor and non-labor income consumption/expenditures: includes food and non-food, selfconsumption wealth: monetary and non-monetary Complicated to capture it with precision in a survey require several, detailed questions that recall activities of several days, even months Standard household surveys would take more than half an hour for expenditures and similar for income measurement error is a relevant element They are usually measured at the household level
Measuring SES (II) Are they any different? Income fluctuations are based on risk and life-cycle patterns Consumption/expenditure is less volatile than income because individual/households tend to smooth it from income fluctuations Wealth may be even more volatile than income because it is the mechanism through which consumption is smoothed (precautionary savings It is also understood that income and wealth have larger reporting error bias That is why it is pretty standard to base SES on consumption or expenditures, (G) whenever possible Nevertheless, it is not so simple yet
Demography and household SES The idea is to connect the monetary value of consumption with welfare (unobserved) that connection varies with household size. Same level of expenditure allows more welfare for smaller households Differences (inequalities) are indeed underestimated because poor households tend to be larger use per capita household expenditures, denote it g Still, g is sometimes questioned because of two effects i. differences in the needs of members of different ages and gender, and ii. the existence of economies of scale in the household s consumption (TV, radio, refrigerator) iii. Sometimes, they are not empirically relevant
How do we characterize the distribution of SES? Poverty rate (head count index): proportion of the population living below the poverty line Poverty line is often times obtained by estimating the level of expenditure an average household requires to satisfy a basic level of consumption. Poverty line is always arbitrary Poverty rate is not sensitive to differences in the distribution below the poverty line (how poor are the poor?) Poverty gap: is based on the average difference between the poverty line and the level of expenditures of each household for those above the poverty line, the gap is assumed to be zero This index is sensitive to the distance from the poverty line, getting a higher index when we have more households far below the poverty line
How do we characterize the distribution of SES? (II) None of the previous indicators care about the top portion of the distribution. There are other ways to compare the inequality in the distribution of SES, like the rich to poor ratio: First, need to define rich and poor. A popular way is to order the population by a SES indicator, and then divide the population in 5 groups of equal size (quintiles) with the first one having the poorest 20% and the fifth one having the richest 20% Then, we compare average per capita expenditures of each group A limitation is that such indicator is insensitive to changes in the situation of the middle groups
The distribution of per capita expenditures in Peru 12000 12000 10000 10000 8000 8000 S/. 6000 S/. 6000 4000 4000 2000 2000 0 Poorest Q2 Q3 Q4 Richest pc expendture quintiles 0 Poorest D2 D3 D4 D5 D6 D7 D8 D9 Richest pc expenditure deciles The rich to poor ratio will depend on the number of groups in which we divide the populations It still matters for comparing two societies, or the evolution over time
How do we characterize the distribution of SES? (III) There are other more complex inequality indicators: Robin Hood index Gini coefficient Atkinson index Theil s entropy measure
Problems with survey data First, keep in mind that surveys often have less than complete coverage they cover neither the poorest nor the richest Estimates of household expenditures are sensitive to length of the recall period level of aggregation - number of items for which expenditures are asked The collection of expenditure or income data is not only complex but also very expensive Often times, surveys cannot afford to measure them
Alternative ways to approximate SES Most health surveys include information on the tenancy of household assets Type of materials used in the construction of the house (floors, walls, etc) Type of sanitary infrastructure (water, sewage) Appliances (refrigerator, radio, TV, car) Advantages of using assets as a measure of wealth Assets are fewer and easier to measure. The standardization of questionnaires is less a problem. Assets used in the index are less likely to face reporting bias - most of them are observable to the interviewer Do not have to worry about price differences
The Asset Index (Filmer & Pritchett, 1998) Problem is we want one summary indicator coming out of the several relevant assets for which we may have information With k assets (a 1, a k ), we want an index of the following type: A ˆ ˆ i = γ 1ai1 +... + γ kaik (1) ( ) Where do we get the weights γ k from? Often times, assets and weights are agreed upon in focus groups with experts An alternative is to let the data decide through, for instance, a principal component analysis Application is available in most statistical packages (SAS, SPSS, STATA, etc.)
Generating SES Proxies in health facilities Assets can also be collected by specific health projects or health facilities to identify the SES of the people they serve It is feasible to include these assets in the entry records, especially if they are observable and verifiable. Let s call these variables X Use a household survey with asset and expenditure data to estimate the following regression HS HS y = X β + ε i i i Then, go to your project or facility data to calculate yˆ PFD = X PFD ˆ β i i HS
Alternatives to establish SES: Participatory approaches Previous approaches are common when household surveys are available Also, they are based on statistical properties and have a margin of error In that sense, they are less appropriate in the context of specific projects working in small localities Nevertheless, even in small poor areas, there are many inequalities and heterogeneities within the population and some prioritization is required
Alternatives to establish SES: Participatory approaches (II) A myriad of approaches and methods have emerged over the past two decades Rapid Rural Appraisal (RRA) Participatory Rural Appraisals Participatory Poverty Assessment (PPA) Shared principle: Poor or excluded people have extremely valuable information on the nature of poverty, exclusion, vulnerability, which are particularly relevant when working in local environments
Alternatives to establish SES: Participatory approaches (III) Key difference between RRA and PRA (Chambers, 1994) Nature of process RRA PARA Mode extractive elicitive sharing empowering Outsider's role investigator facilitator Information owned, analyzed and used by outsiders local people Methods similar Some of the Methods used Semi Structured interviews Focus groups Maps: Transect walks, participatory mappings, to gain knowledge on the way individuals see the territory/environment in which they live Time lines, to recover main events that affected the community Rankings and comparisons of wealth and well being
Participatory assessments of poverty and vulnerability You can use mapping methods to identify areas of the locality where the poorest, excluded and vulnerable are most concentrated, main environmental hazards semi-structured interviews and focus groups to identify the characteristics of the poor/excluded/vulnerable (locations, types of dwellings, occupational status, etc) rank individuals or blocks based on different criteria of wealth or well being (only in very small environments)
Limitations of participatory approaches It has often proven to be a cost-effective way to get information on the nature of poverty and to identify the poorest Nevertheless, it has some risks since quality of the information depends on who one interviews or works with individual or block rankings would be biased if its connection to some prioritization in the intervention become apparent Key recommendations get the help of experts in participatory approaches subsume the collection of information within a clear empowering process
Next step: Health indicators As with stratifiers, need to know the nature of epidemiology and functioning of health system in the society Relevant indicators may vary for each society and level of analysis (national, local) health status (IMR, undernutrition, prevalence of specific health conditions and diseases) access and utilization of health care health financing (access to health insurance, out of pocket expenditures) allocation of public resources
Health indicators and equity (II): Some issues to consider Access to health services and disease prevalence from health facilities Perception of need, tolerance and access to health care services Access to health insurance and out-of-pocket health expenditures
Needs, tolerance and use of health care 45 30 40 35 25 30 20 (%) 25 20 (%) 15 15 10 10 5 5 0 Poorest D2 D3 D4 D5 D6 D7 D8 D9 Richest Per capita expenditures deciles 0 Poorest D2 D3 D4 D5 D6 D7 D8 D9 Richest Per capita expenditures deciles Ill and disable ambulatory consultation Inequalities in the use of health care are affected by the perception of need and the tolerance of individuals, which is larger for poor and excluded populations Self-reported illness and disability is an imperfect notion of need as tolerance to pain varies across groups
Needs, tolerance and use of health care (II) (%) 35 30 25 20 15 10 5 0 120 100 80 Poorest D2 D3 D4 D5 D6 D7 D8 D9 Richest Asset index deciles Female consultations Two indicators of use of health services Use of ambulatory consultations by females Presence of professionals during birth Second is less affected for differences in need and tolerance Higher inequalities are found in the rate of institutional births (%) 60 40 20 0 Poorest D2 D3 D4 D5 D6 D7 D8 D9 Richest Asset indexs deciles Insitutional births
Access to health insurance and out of pocket health expenditures 60 18 (%) 50 40 30 20 10 hundreds of soles 16 14 12 10 8 6 4 2 nd 0 Poorest Q2 Q3 Q4 Richest Per capita expenditures quintiles Insured 0 Poorest Q2 Q3 Q4 Richest Per capita expenditures quintiles Out of pocket health expenditures When analyzing differences in health expenditures, need to consider differences in quality and access to health insurance as they tend to underestimate inequalities Also, consider that smaller monetary amounts represent larger shares of the budget for less favored groups (%) 7 6 5 4 3 2 1 0 Poorest Q2 Q3 Q4 Richest Per capita expenditures quintiles Share of out of pocket health expenditures
Measures for the magnitude of socioeconomic inequalities in health Simple measures have the advantage of easy calculation, not much data requirement and straightforward interpretation That is for dissemination, advocacy, but for diagnostics, there is no replacement for a deep understanding of the way poverty and inequality, epidemiology and health systems interact in a society There are a number of measures to consider based on summary health indicators: Comparison of extremes (differences or ratios) Absolute and relative Gaps Concentration index Regression-based measures
Comparison of extremes Source: Peru, DHS 2000 Health indicator: immunization of children under 5 Comparing richest to poorest deciles Difference: 51-22=29 % points Ratio: 51/22=2.3 times Limitation: does not consider differences in what happens between extremes
Concentration curves, index and inequalities: is lower inequality in rural areas good? 1.00 Rural 0.75 Urbano Undernourishment concentration index 0.50 Global -36.09 0.25 Urban -38.05 Rural -11.88 0.00 0.00 0.25 0.50 0.75 1.00 Poblacion acumulada por gasper
Chronic malnutrition rates: importance of looking at the whole distribution 70 60 decil 1 Tasa de desnutrición 50 40 30 20 decil 1 decil 10 10 decil 10 0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 Promedio del ln del gasto per cápita Urbana Rural Malnutrition rate in rural areas is 40% in rural areas, but it drops to 12% in urban areas The issue is: rural kids face higher nutritional risk than their SES-equivalent urban counterparts.
Final comments on measures of health inequalities Summary measures are often simple (maybe not the concentration index) but there is a risk to hide very important information on the nature of health inequalities You can easily think of the last example in terms of inter-temporal comparisons (inequalities can de reduced by making everybody worse off) Comparisons can also be made for other stratifiers: gender, race, ethnicity, etc Recall that comparisons across groups are relevant as we see that inequalities are unnecessary and evitable in some broad sense
Grupo de Análisis para el Desarrollo www.grade.org.pe