Ecological and small-area studies
Ecological studies compare groups, such as districts. Small-area studies examine patterns in small places, using area totals, individual records, or both.
Why this level?
Specify the research question: intervention, diagnostic accuracy, prognosis, aetiology, or screening intervention. Other question types are outside this hierarchy. NHMRC 2009 sourceEcological and small-area studies cover descriptive and causal questions, with different data levels. The label alone does not assign one evidence level.
Fit
When to use it, and when not to
Use it when
- Your question concerns area patterns in disease burden, exposure or access to care.
- You have counts and population denominators for comparable places and periods.
Not when
- You want an individual exposure effect from area averages alone. Use a cohort with individual exposure and confounder data. Read: Prospective cohort study
- You only need a prevalence estimate for individual people. Use a prevalence study with suitable sampling. Read: Prevalence survey
Before you start
What you need
- The unit of exposure and outcomeState whether each measure describes an area or a person. Small-area does not automatically mean ecological.
- Comparable counts, boundaries and population denominatorsAlign dates and area borders. Use person-time when people enter and leave, and standardize for age when needed.
- Dependence within areas and between nearby areasPeople in one area share conditions. Nearby areas can share exposures, services and reporting systems.
- A plan for gaps and suppressed cellsRecord which areas lack exposure, outcome or denominator data. Suppressed small counts are not zeros.
- Group associations do not identify individual associations.How to check: The ecological fallacy can reverse the direction of an association. Seek individual data before making an individual claim.
- Individual outcomes sharing area exposure are clustered.How to check: Use a random intercept or cluster-robust standard errors, or give an explicit justification. Address spatial dependence between areas too.
- Rates refer to comparable people, places and time.How to check: Check age mix, migration, boundary changes and case definitions. Try alternative area boundaries to assess sensitivity.
- Missing records are handled under a named assumption.How to check: Missing completely at random (MCAR) means gaps are unrelated to recorded and unrecorded values; this is rarely plausible. Missing at random (MAR) means gaps depend only on recorded factors. Under MAR, consider multiple imputation; it is not automatically better than a justified complete-case analysis. Missing not at random (MNAR) is plausible if unobserved values themselves influence whether records are absent. Under MNAR, test departures with a delta or tipping-point analysis that shifts the unobserved values.
Plain language
How it works
Define areas and align their counts, exposures and denominators in time. Model counts with an offset, which scales expected counts to population size. Allow for age mix and other measured causes. Small areas often have sparse counts. Smoothing shares information across areas, but depends on a model and can hide local variation.
An area-level outcome describes that area. Individual outcomes linked to an area exposure need clustered inference or an explicit justification. Spatial correlation is similarity between nearby areas. It can remain after accounting for within-area clustering. Neither adjustment nor a smooth map resolves the ecological fallacy.
Worked example
A worked example
- Setting
- A registry compares admission counts in two sets of districts over one year.
- Question
- Are crude admission risks higher in districts with greater air pollution?
- Approach
- Pool counts within each district set for description. Model area counts with population denominators and assess spatial dependence for inference.
- Result
- Admission occurs in 80 of 1,000 people in the higher-exposure set (8 in 100), versus 40 of 1,000 (4 in 100). The difference is 4 per 100; the ratio is 2.
- What it means
- These pooled area risks do not identify which people were exposed or why they were admitted. They cannot establish an individual causal effect.
In practice
Recent studies PubMed search · not vetted
Watch out
Common mistakes
The one to remember
An association between areas does not establish an association between individuals. For group-level exposure with individual outcomes, account for clustering or state a justification. Area averages do not remove individual confounding.
- Treating a colored map as evidence that an exposure caused disease.Instead: Check confounding, reporting and spatial dependence. Report estimates with uncertainty, not only ranks or colors.
- Treating thousands of people as independent when exposure varies across a few areas.Instead: Count the independent areas and use clustered inference. Few areas need methods suited to small samples.
- Dropping all areas with missing exposure values.Instead: Describe selection bias and model the gaps under a stated assumption. Compare sensitivity analyses that retain those areas.
Inside Mallard
How Mallard uses this
Plans do not link to this design yet. Mallard has no dedicated sizing form for ecological and small-area studies.
design.ecological, the id a plan cue records.How to report it
No reporting guideline dedicated to ecological or small-area studies was found in a PubMed search on 2026-11-06. The search used design names and reporting-guideline or checklist terms in titles. Use STROBE as a reporting framework for the observational comparison, while stating its limits for aggregate data. Report area boundaries, denominators, age standardization, spatial dependence, missingness and sensitivity to grouping. STROBE does not establish that the ecological association identifies an individual effect. A reporting guideline is not a design manual: use it to make the study transparent, not as a substitute for careful design. Check the target journal's own requirements as well, which this page cannot know.
Related Learn pages
Related designs and methods
Evidence
Sources & further reading
- Small-area assessment of temperature-related mortality risks in England and Wales: a case time series analysis (opens in a new tab) · Lancet Planet Health · doi 10.1016/S2542-5196(22)00138-3
- Monitoring the spatiotemporal epidemiology of Covid-19 incidence and mortality: A small-area analysis in Germany (opens in a new tab) · Spat Spatiotemporal Epidemiol · doi 10.1016/j.sste.2021.100433
- The potential for bias in Cohen's ecological analysis of lung cancer and residential radon (opens in a new tab) · J Radiol Prot · doi 10.1088/0952-4746/22/2/302
- Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies (opens in a new tab) · BMJ · doi 10.1136/bmj.39335.541782.AD
- Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls (opens in a new tab) · BMJ · doi 10.1136/bmj.b2393