Evaluation of a Decision Tree for Efficient Antenatal Red Blood Cell Antibody Screening.

BACKGROUND: Hemolytic disease of the fetus and newborn due to maternal red blood cell alloimmunization can have serious consequences. Because early detection enables careful monitoring of affected pregnancies, programs to routinely screen all pregnant women have been widely adopted. Due to the low prevalence of alloimmunization, these require large investments of resources to detect a small number of cases.

METHODS: We conducted a validation study of a decision tree developed in the Netherlands for determining whether to screen for alloimmunization. In a Swedish cohort, we compared the performance of that decision tree to two alternative models that used maternal characteristics, obstetric history, and transfusion history to identify high-risk women for screening or low-risk women who might be exempt from screening. The models were compared for predictive ability and potential reduction in the volume of screening.

RESULTS: The decision tree applied to our study population identified 89% of alloimmunized women with a negative predictive value (NPV) of 99.7% by screening 62% of the population. To achieve the same NPV, our model exempting low-risk women captured 90% of alloimmunizations by screening 63% of the population. In contrast, the model identifying high-risk women for screening while maintaining a similar NPV captured 63% of alloimmunized women by screening 20% of the population.

CONCLUSIONS: We validated that an existing decision tree for selecting women for maternal screening performed well in our population, identifying a large proportion of women who became alloimmunized, with a predictive performance almost identical to that of a more elaborate model.