6B-02 - Designing an algorithm to identify pregnancy episodes among cancer survivors using claims data

Background:
Fertility is an essential issue for cancer patients at childbearing age while receiving cancer treatments. Investigating cancer treatment and fertility is often unsuitable for clinical trials, making observational studies using real-world data a more feasible solution. Currently, the necessary computable phenotype algorithms for identifying pregnancy episodes using real-world data are limited.

Objectives:
The study designed an algorithm to identify pregnancy episodes and to investigate fertility trends among female cancer patients in Taiwan.

Methods:
We identified pregnancy episodes (live births and non-live birth outcomes, including stillbirth, ectopic pregnancy, and spontaneous/ induced abortion) using a hierarchy approach and clinically plausible periods for subsequent episodes. This method was then applied to the National Health Insurance Research Database (NHIRD) and further validated with the Birth Certificate Application database (live births and stillbirths registry) using the positive predictive value (PPV). A retrospective cohort of women aged 15-49 with cancer (2007–2021) was identified from the Taiwan Cancer Registry; controls were cancer-free women of the same age. The standardized incidence rates (SIR, cancer vs. general population) were calculated with weights adjusted for each 5-year age interval and analyzed by Poisson regression.

Results:
A total of 2.3 million women with 4.7 million birth-related events were identified to develop the algorithm. The PPVs were 99.8% for live birth and 97.8% for stillbirth. The SIR of pregnancy increased from 0.50 (95% CI: 0.46–0.54) in 2007 to 0.59 (95% CI: 0.56–0.61) in 2021, showing a gradual upward trend (p-value for trend < .001). In 2021, the SIR of non-live birth outcomes was 0.58 (95% CI: 0.53–0.64) in childbearing-age cancer survivors and 1.01 (95% CI: 0.93–1.11) in pregnant cancer survivors. Childhood cancer survivors have a slightly higher pregnancy SIR (0.73, 95%CI: 0.59–0.87 in 2021, ranging 0.59 to 0.73, p-value for trend= 0.74) than adolescent and young adult (AYA) cancer survivors (0.64, 95%CI: 0.61–0.68, ranging 0.52 to 0.64, p-value for trend < .001). AYA survivors had lower pregnancy rates than childhood cancer survivors (incidence rate ratio =0.90; 95% CI: 0.81–0.99). The pregnancy SIR for lymphoma survivors was slightly higher than for leukemia survivors.

Conclusions:
This study lays the groundwork for identifying pregnancy status using health insurance claims data in Taiwan. Pregnancy rates in female cancer survivors, especially those diagnosed between the ages of 15 and 39, are increasing. The higher pregnancy rates observed in childhood cancer survivors may be attributed to a more extended recovery period.