Related Experiment Videos
Constructing reproductive histories by linking vital records
M M Adams1, H G Wilson, D L Casto
1World Health Organization Collaborating Center in Perinatal Care and Health Services Research in Maternal Child Health, Division of Reproductive Health, Atlanta, GA, USA.
American Journal of Epidemiology
|February 15, 1997
Summary
Linking birth and fetal death records created reproductive histories for women. A two-step deterministic and probabilistic approach linked millions of records, but some complex cases were challenging, potentially impacting high-risk groups.
Area of Science:
- Reproductive Health Epidemiology
- Biostatistics
- Public Health Surveillance
Background:
- Linking vital records is crucial for understanding reproductive health trends.
- Previous linkage methods often relied on deterministic approaches, which can be error-prone.
- A combined deterministic and probabilistic strategy offers a more robust linkage.
Purpose of the Study:
- To link live birth and fetal death certificates in Georgia (1980-1992) to reconstruct individual women's reproductive experiences.
- To evaluate the effectiveness and challenges of a hybrid deterministic-probabilistic record linkage methodology.
- To identify potential biases in record linkage, particularly for high-risk populations.
Main Methods:
- Linked 1,449,287 birth and fetal death certificates using a deterministic approach.
- Applied a probabilistic method to validate and correct linkages from the deterministic step.
- Constructed reproductive event chains for individual women, excluding induced abortions and ectopic pregnancies.
Main Results:
- 38% of records could not be linked.
- 369,686 reproductive event chains (two or more events) were created.
- Probabilistic evaluation corrected linkages for 3.0% of records, with longer chains being more affected.
- Difficult-to-link records disproportionately represented high-risk groups.
Conclusions:
- A hybrid deterministic-probabilistic linkage method can successfully create reproductive histories from vital records.
- While effective for most records, challenges remain in linking complex cases, potentially introducing bias.
- Researchers should be aware of potential overrepresentation of high-risk groups among unlinked or difficult-to-link records.