Background: National-level data are often used to identify groups of women at greater risk of not obtaining mammography, who might then receive targeted interventions. An important question, however, is how well results of national-level analyses match results from smaller samples of the same dataset. This study investigated the consistency of results about correlates of mammography from a single national-level analysis versus the results from analyses within each of five quintiles of mammography rates and nine Census subdivisions.
Methods: The sample for all analyses were women aged 42-79 from the Year 2002 United States' Behavioral Risk Factor Surveillance System (N = 80,283). Recent mammography was defined as self-report of a mammogram within the 2 years prior to the interview. Independent variables included sociodemographics, health practices, and a combined insurance status/usual source of care variable.
Results: Only smoking status, Pap testing, dental visit, and health insurance/source of care had consistent results with mammography status across all levels of analysis. Results for the other covariates, including standard sociodemographics, showed varying degrees of consistency.
Conclusions: Caution is advised when using national data to inform regional or local intervention planning. Local and regional data are necessary to target programs to groups at greatest need for intervention.