Understanding the ecological fallacy

Answers

Statements a, b, c, and d are all true.

The aim of the study was to establish whether an association existed between antibiotic prescribing by general practices and antibiotic resistance. The unit of observation was the patient (a is true), whereas the unit of analysis was the general practice (b is true). The unit of observation and unit of analysis are often confused.2 The unit of observation, sometimes referred to as the unit of measurement, is defined statistically as the “who” or “what” for which data are measured or collected. The unit of analysis is defined statistically as the “who” or “what” for which information is analysed and conclusions are made.

In the above study, the data were measured for patients; antibiotic resistance was recorded for those patients who provided urine samples, while the number of dispensed trimethoprim prescriptions per patient was identified from healthcare records for all patients in each of the practices. However, the data were aggregated at the practice level, permitting a comparison of the association between the rate of trimethoprim prescription in practices and the proportion of patients with urine samples identified with trimethoprim resistant Gram negative bacteria across the 28 general practices.

The data were analysed using Spearman’s rank correlation coefficient,3 which measured the strength of the linear association between rates of trimethoprim prescription and the proportion of patients with urine samples identified with trimethoprim resistant Gram negative bacteria across the 28 general practices (c is true). A weak negative association between trimethoprim prescription and resistance across general practices was reported. Hence, there was no evidence that practices with the highest prescribing rates had increased proportions of patients with trimethoprim resistance. However, to assume that these results could be applied to the individual (that is, that patients with prescriptions for trimethoprim do not have trimethoprim resistance) would be an erroneous belief termed the ecological fallacy (d is true). The results suggest that general practices with the highest rates of trimethoprim prescribing have similar proportions of patients with trimethoprim resistance as those practices with the lowest rates of trimethoprim prescribing. However, we cannot assume that patients receiving trimethoprim prescriptions are just as likely to have trimethoprim resistance as those patients not receiving trimethoprim prescriptions.

Although the authors of the above study aggregated and analysed data at the general practice level, they also reported the results of analyses at the individual patient level. It was possible to link patient records indicating prescription of trimethoprim and the test results for the urine samples. In particular, resistance to trimethoprim was significantly associated with age, sex, and the individual’s exposure to trimethoprim or other antibiotics. The association with trimethoprim resistance was strongest for people who had recently been exposed to trimethoprim. The researchers commented that the analysis of data at the practice level obscured important associations between antibiotic prescribing and resistance at the individual level.

More generally, the ecological fallacy is a term used when data collected at a group level are analysed and the results are assumed to apply to associations at the individual patient level. The ecological fallacy is of particular concern when interpreting results from ecological studies.4 An ecological study is observational by design, whereby the investigators do not intervene but record factors such as the health, behaviour, attitudes, lifestyle choices, or economic status of people or patients. The people or patients occur in clusters or natural groupings, such as general practices or countries, and a cross section of the clusters is obtained to represent the population. Although the unit of observation may be the individual person or patient, the unit of analysis is the natural grouping (cluster). Therefore, ecological studies are prone to the ecological fallacy. Ecological studies make large scale comparisons between groups of people and allow an initial examination of the status and needs of communities, for example, the health status of countries.