A: The study’s purpose, design, data, confidentiality considerations, and funder requirements should inform which geographic unit of analysis is most appropriate. For a broader discussion of geographic units, see Chapter 5. Geography and Disparities in Health Care (Ricketts, TC) in Guidance for the National Healthcare Disparities Report.
Studies that work with small cell sizes, especially in small geographic units, should consider the risk of deductive disclosure, where an individual’s identity may be ascertained using known characteristics, such as race and age, even when direct identifiers, such as name and address, are removed.
Studies that use multiple datasets at varying geographic units can aggregate or approximate the data to larger geographic areas. For example, the FutureDocs Forecasting Tool created tertiary service areas by aggregating groups of counties to approximate the Dartmouth Atlas Hospital Referral Regions, which are built on ZIP Codes.
Analyses using sample data should evaluate the data’s sampling unit and frame when determining an appropriate geographic unit of analysis. Data derived from a sample will have some degree of uncertainty associated with the estimates. Generally, the smaller the sample the larger the sampling error. Relying on a small geographic unit may further exacerbate the uncertainty around the estimates and prevent researchers from producing reliable statistics. Therefore one should carefully consider the data-generating process before considering the geographic unit of analysis.