Application of Classification and Regression Trees (CART) methodology to evaluate a behavioral intervention impacting poor pregnancy outcomes among minority women: A randomized clinical trial

The objective of this study was to apply Classification and Regression Trees (CART) analysis to assess the impact of a psychosocial/behavioral intervention in reducing adverse pregnancy outcomes (low birthweight (LBW) and prematurity) and to identify women who benefited most from the intervention.

The study was conducted in 6 clinics in Washington, DC. Pregnant women were screened for eligibility and risks: smoking, environmental tobacco smoke exposure, depression, and intimate partner violence (IPV). 1,044 African American women were randomized to the integrated intervention (IG) or usual care (UCG) groups. IG received educational/counseling sessions targeting reported risks. 2 follow-up interviews were conducted. CART, was used to identify sociodemographic subgroups at the highest risk for LBW and prematurity and to characterize women who benefited most from this intervention.

Births <34 weeks were more common among women with IPV in the UCG vs. IG (12.0% vs. 3.0%, p=.009). Controlling for sociodemographics and risk factors, IG was less likely to have very premature babies than UCG (2.9% vs. 6.1%, p=.039). Women with 4+ intervention sessions were less likely to have LBW (p<.001) and those with 3+ sessions were least likely to have very LBW (p=.002). Women with 6+ sessions were less likely to have premature (<37 wks) babies (p<.001) and those with 3+ sessions were less likely to have very premature babies (<34 wks) (p<.001).

CART analyses elucidated the specific associations between risk and outcome and identified subcohorts benefiting most from this behavioral intervention. IPV was a marker for reproductive morbidity and responded to the intervention.