Evaluating a new algorithm for linking maternal and newborn medical records

Area of Science:

• Public Health
• Biostatistics
• Medical Informatics

Background:

• Linking maternal and newborn medical records is crucial for understanding fetal outcomes.
• Accurate linkage is essential for epidemiological studies and healthcare quality assessment.
• The Center for Disease Control (CDC) developed a new computer-based record linkage system.

Purpose of the Study:

• To evaluate the accuracy of the CDC's maternal and newborn record linkage system.
• To identify reasons for incorrect matches and non-matches.
• To assess the impact of system revisions on matching accuracy and non-match rates.

Main Methods:

• Utilized a dataset of 10,068 newborn records and associated maternal records from seven Ohio hospitals.
• Applied the CDC's weighted variable computer program to match maternal and newborn records.
• Analyzed matched and nonmatched records to determine accuracy and causes of discrepancies.

Main Results:

• The initial system achieved 98% matching accuracy, with 5.7% of newborn records remaining nonmatched.
• Nonmatched records were associated with higher rates of birth defects and prematurity.
• Program revisions reduced nonmatched records to 3% but also decreased matching accuracy.

Conclusions:

• The maternal and newborn record linkage system is a valuable tool but requires refinement.
• Reducing nonmatched records is important due to associated adverse outcomes.
• Balancing matching accuracy and the rate of nonmatched records is a key challenge in system improvement.