Evaluating a new algorithm for linking maternal and newborn medical records

Linking material and newborn medical records is a valuable tool for assessing the relationship between maternal variables and fetal outcome. This study evaluated the Center for Disease Control's newly developed maternal and newborn medical record linkage system, a computer program that uses weighted variables to determine the most likely maternal and newborn pairs. Any newborn record not achieving a set minimum score with a maternal record remains nonmatched. The objectives of the study were to estimate the program's matching accuracy, determine causes of incorrect matches and nonmatches, develop suggestions for program revisions, and evaluate the effects of the revisions. The study sample included 521 matched and 247 nonmatched maternal and newborn medical records from seven Ohio hospitals. Of all available newborn records (10,068), 574 (5.7%) did not match with maternal records; for those in which a match occurred, the authors ascertained a 98% matching accuracy and determined explanations for nonmatched and incorrectly matched records. The authors noted a greater prevalence of birth defects and prematurity among newborns with nonmatched records than among those with matched records. Program revisions, therefore, focused on reducing the prevalence of nonmatched records. The revised program reduced the prevalence of nonmatched records from 5.7% to 3% but reduced matching accuracy.