Developing a multiphase claims-based algorithm for non-live pregnancy outcomes research
Recommended Citation
Secrest M, Phillips S, Shen SW, Woodcroft KJ, Oliveria SA, and Simon TA. Developing a multiphase claims-based algorithm for non-live pregnancy outcomes research. Pharmacoepidemiol Drug Saf 2018; 27:87.
Document Type
Conference Proceeding
Publication Date
2018
Publication Title
Pharmacoepidemiol Drug Saf
Abstract
Background: Non-live outcomes occur in about 30% of pregnancies and may be mediated by exposure to medication during pregnancy, depending on various factors including drug type, length, of exposure and/or time of exposure (eg, foetus age/proximity to conception). Understanding the safety of medication exposure during pregnancy is critical but presents methodological challenges. Registries are common, but often yield few cases. Administrative claims data can be used to study large numbers of women if the claims data accurately identify/estimate non-live outcomes, gestational age (GA) at non-live outcome, and medication exposure during pregnancy ending in non-live outcome. Objectives: To develop a multiphase claims-based algorithm that identifies non-live outcomes (Phase [Ph]1), estimates GA at non-live outcome (Ph2) and estimates medication exposure during pregnancy ending in non-live outcome (Ph3). Methods: A multiphase algorithm is being developed in stages among women aged ≥15 and ≤50 years with ≥1 end of pregnancy (EOP) ICD-9 code and enrolment and prescription coverage 340 days prior to EOP in the Henry Ford Health System in the United States between January 1, 2013, and September 30, 2015. Algorithms have been (Ph1) or will be (Ph2-3) developed, applied to claims data, and validated against electronic medical records using estimated positive predictive value (PPV), sensitivity, and corresponding 95% confidence interval (CI). The best-performing algorithm in each phase is used in the next phase. Based on previous work for live outcomes presented at ICPE 2017 (abstract 154), Ph1 Algorithm 1 (≥1 definitive ICD-9 EOP code) validation was modified to ascertain events ±30 days from the EOP date (vs 0-7 days). Work is ongoing to assess algorithms that estimate GA at non-live outcome by assigning an estimated GA to 3 or 5 categories of non-live outcomes (Ph2) and to assess drug exposure during pregnancy to long-term (eg, antidepressants) and short-term (antibiotics) medications based on estimated GA (Ph3). Results: A total of 698 women met the inclusion criteria, and 145 had non-live EOP codes. When validating EOP codes within 0-7 days of the EOP date, the Ph1 Algorithm 1 PPV was 79% (95% CI: 71, 85), and the sensitivity was 97% (95% CI: 96, 99). When validating EOP codes ±30 days from the EOP date, the PPV was 85% (95% CI: 78, 90), and the sensitivity was 100% (95% CI: 97, 100). Conclusions: Non-live EOP outcomes can be identified in claims data with high PPV and sensitivity. Further analyses are underway to validate algorithms for Ph2 and Ph3.
Volume
27
Issue
87