Designs for the combination of group- and individual-level data

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Ecological studies using aggregate data are prone to bias when examining individual-level associations.
• Combining group-level and individual-level data can mitigate these biases.
• While multilevel models are established, guidance on choosing optimal combined designs is lacking.

Purpose of the Study:

• To review and present recently proposed combined group- and individual-level designs for data analysis.
• To provide a framework for researchers to select appropriate designs for their studies.
• To highlight the benefits of integrating multilevel data sources.

Main Methods:

• Review of recently proposed combined group- and individual-level designs.
• Analysis of data at two levels of aggregation: group and individual.
• Illustration using a simulation study based on birth-weight data.

Main Results:

• Methods vary in data elements and statistical techniques for combining group and individual data.
• Careful implementation is needed; ignoring group data may be simpler post-individual data collection.
• Simulation demonstrated the advantages of incorporating group-level information.

Conclusions:

• No single combined design is universally ideal.
• The choice of design hinges on the research model and the characteristics of available group-level data.
• Integrating individual-level data with accessible group-level data offers significant potential for epidemiological research.