Strengths and Limitations
Using the validated and weighted Maternal Comorbidity
Index30,45, capture of the range of maternal
comorbidities was a distinct advantage of using health claims data, as
these data capture utilization before, during and post-delivery as well
as across multiple health care delivery sites accessed by each
patient.46 In comparison, prior studies used
self-report, birth certificate and/or hospital discharge data for birth
outcomes and related morbidity and were limited in identifying the full
range of comorbidities that precede and occur during
pregnancy.3,4 Results of our mediation analysis helps
explain why AYA cancer survivors have more adverse perinatal outcomes,
and illustrates potential use of this analytic approach for future
research aimed at improving outcomes in cancer survivors.
We note several limitations. First, prior cancer remained associated
with adverse outcomes in singleton births, after adjusting for broad
chemotherapy or radiation exposures and comorbidities, leaving
unanswered why cancer itself would be related to these outcomes. One
potential reason is iatrogenic delivery, which cannot be accurately
captured in administrative data. Restricting to commercial insurance
enrollees due to the nature of the dataset excludes women with Medicaid
(which provides coverage for large proportions of cancer survivors and
pregnant individuals in the U.S.47), limiting
generalizability. As we relied on billing codes (ICD, CPT, DRG, etc.) to
identify exposures, covariates, and outcomes, misclassification could
have occurred but anticipated to be non-differential by cancer status.
While both severity of preterm births (<28 weeks,
<32 weeks) and spontaneous versus iatrogenic preterm births
are outcomes of interest, claims cannot accurately capture them. Due to
using both ICD-9 and ICD-10-CM/PCS codes, potential disruptions in
observed rates relating to the coding transition and coding errors could
contribute to misclassification bias. Confounders including smoking,
obesity, and prior preterm birth are not reliable in health claims data
and thus could not be included. Finally, there may be detection bias of
comorbidities and outcomes in AYA cancer survivors that contributed to
our findings.