Applying recursive partitioning to a prospective study of factors associated with adherence to mammography screening guidelines.

Although a number of predictors of adherence to mammography screening guidelines have been identified using traditional statistical methods, many women are not screening according to these guidelines. Recursive partitioning may aid in developing novel intervention strategies to promote this screening behavior by identifying subgroups of women that differ on adherence across predictor variables. In a prospective study of 1,229 African-American and White women in Connecticut whose adherence to mammography screening guidelines was ascertained over a 26-month follow-up period from initial screening in 1996-1998, recursive partitioning selected six of 22 candidate predictors and identified subgroups that differed on adherence across predictors by age (40-49 and 50-79 years). Among the five subgroups identified for women aged 50-79 years, the subgroup most adherent to screening guidelines during follow-up included four predictors: a history of adherence, annual family income of 15,000 dollars or more, a belief that mammograms were very useful, and low or moderate perceived breast cancer susceptibility. Among the three subgroups identified for women aged 40-49 years, the most adherent subgroup included only one predictor: receipt of a health-care provider's recommendation to obtain a mammogram. These findings suggest that recursive partitioning may be a useful statistical tool and may aid in developing interventions to promote adherence to mammography screening guidelines.