Ecological studies: advantages and disadvantagesBMJ 2014; 348 doi: https://doi.org/10.1136/bmj.g2979 (Published 02 May 2014) Cite this as: BMJ 2014;348:g2979
- Philip Sedgwick, reader in medical statistics and medical education
Researchers examined the association between child wellbeing and economic status in rich developed societies. An ecological cross sectional study design was used, with 23 of the richest 50 countries in the world included in the analysis. Child wellbeing was measured by the Unicef index of child wellbeing. Three macro-economic measures were used—material living standards (average income), the scale of differentiation in social status (income inequality), and social exclusion (children in relative poverty).1
The overall Unicef index has 40 items that measure six dimensions—material wellbeing, health and safety, education, peer and family relationships, behaviours and risks, and young people’s own subjective sense of wellbeing. An item measuring relative poverty was removed before calculating the index of child wellbeing. Low scores indicated worse outcomes. Income inequality was measured as the ratio of the total annual household income received by the richest 20% of the population to that received by the poorest 20%. Therefore, larger values indicated greater inequality between the richest and poorest within a country. Child relative poverty was measured as the proportion of children aged 0-17 years in households with an income equivalent to less than the national median.
It was reported that the overall index of child wellbeing was negatively correlated with income inequality (r=−0.64, P=0.001) and proportion of children in relative poverty (r=−0.67, P=0.001), but not with average income (r=0.15, P=0.50). Therefore, countries with lower measures of child wellbeing had greater inequality of income and a higher proportion of children in relative poverty. The researchers concluded that improvements in child wellbeing in rich societies may depend more on reductions in inequality than on further economic growth.
Which of the following statements, if any, are true?
a) The unit of analysis was the country
b) It can be inferred that children with lower levels of wellbeing were more likely to be living in relative poverty
c) It can be inferred that greater income inequality in a country caused worse outcomes for child wellbeing
Statement a is true, whereas b and c are false.
The aim of the study was to examine the association between child wellbeing and economic status in rich developed societies. An ecological cross sectional study design was used. An ecological study is observational in design. In an observational study the investigators do not intervene in any way but record factors such as the health, behaviour, attitudes, lifestyle choices, or economic status of the study participants. The ecological study was a cross sectional one, with 23 of the richest 50 countries in the world included in the analysis. As the name of the study design implies, the researchers took a cross section of the richest countries, with the aim of obtaining a representative sample.
In an ecological study the unit of analysis is a group of people rather than the individual. In the study above the unit of analysis was the country (a is true), with data aggregated across each country for each of the outcome measures. The association between the Unicef index of child wellbeing and each of the three macro-economic measures was investigated using Pearson’s correlation coefficient. Pearson’s correlation coefficient, described in a previous question,2 was used to describe the strength of the association between the outcome measures across the 23 countries.
The example above is typical of an ecological study—routinely collected data were used to explore the association between child wellbeing and economic status. Ecological studies are therefore usually cheap and easy to perform. However, data are unlikely to be collected for all members of the group of people—the unit of analysis. Thus, when the data are aggregated across the unit of analysis in an ecological study, the outcome measures are likely to be inadequate or biased.
Results from ecological studies are prone to the ecological fallacy. The ecological fallacy is a term used when collected data are analysed at a group level and the results are assumed to apply to associations at the individual level. For example, the above study reported that a negative correlation existed at the country level between child wellbeing and the proportion of children in relative poverty. To assume that these results could be applied to the individual child—that is, a child with lower levels of wellbeing was more likely to be living in relative poverty—is an erroneous belief termed the ecological fallacy (b is false). Although countries with the highest proportion of children living in relative poverty have the lowest levels of child wellbeing, we cannot assume that children from families living in relative poverty were those with lowest levels of wellbeing.
Ecological studies typically investigate the association between outcome measures using correlation analysis. However, care is needed when interpreting the results of such analyses. For example, the index of child wellbeing was significantly negatively correlated with income inequality (r=−0.64, P=0.001; figure⇓). Therefore, those countries with lower measures of child wellbeing had greater income inequality. A significant correlation does not mean that a cause and effect association can be implied (c is false). That is, it cannot be inferred that the decrease in child wellbeing for a country would have been caused by a rise in income inequality. Correlation, described in a previous question,3 measures the strength of association between two variables.
The purpose of an ecological study is to make large scale comparisons between groups of people—an example of its use would be in assessing the health status of countries. Ecological studies do not examine individuals but large groups of people and the results are therefore open to bias, as described above. However, they do allow an initial examination of the health status and needs of communities. Ecological studies should be seen as a means of generating hypotheses rather than deriving definitive information about associations between risk factors and health outcomes.
Cite this as: BMJ 2014;348:g2979
Competing interests: None declared.