In the Canadian population, does caesarean delivery compared to vaginal birth increase risk of early neonatal mortality? An instrumental variable method approach

摘要

BACKGROUND: The lack of adjustment for unmeasured factors which may be associated with both delivery decisions and pregnancy outcomes has likely resulted in an overestimation of the risk associated with caesarean delivery on neonatal mortality. An instrumental variable method (IVM) originating from the field of econometrics has been utilized in modern epidemiological research to reduce the influence of unmeasured selection bias. By accounting for measured, unmeasured, and unknown confounding variables, utilizing the IVM can serve as a more valid approach in determining intervention effects amongst patients in observational studies.udOBJECTIVE: The purpose of this study is to compare the results from traditional multivariate methods and instrumental variable-adjusted analyses to determine if caesarean delivery increases the risk of early neonatal death in comparison to vaginal birth. MATERIALS AND METHODS: This is a retrospective cohort study which compares the outcome of early neonatal mortality between 20 completed weeks of gestation and 7 days post-partum among women who delivered through a caesarean section and women who delivered vaginally. The cohort includes all in-hospital births during the fiscal years of April 1, 2006 - March 31, 2009 across Canada identified in the Discharge Abstract Database (DAD) from the Canadian Institute of Health Information (CIHI), excluding deliveries in Quebec. The effect of mode of delivery, being either caesarean or vaginal delivery, on early neonatal mortality was measured using a bivariate logistic regression, followed by a multivariate logistic regression and instrumental variable-adjusted analysis which controlled for 24 covariates. RESULTS: Multivariate logistic regression indicated that caesarean delivery significantly reduced the risk of early neonatal death in comparison to vaginal birth by 21% (Adjusted OR = 0.79, 95%CI = 0.66-0.93, p = 0.006). Instrumental variable-adjusted regression indicated a lack of association between mode of delivery and early neonatal mortality (ARD = -0.0053, 95%CI = -4.3x10⁻³-3.0x10⁻³, p = 0.781). CONCLUSION: In conclusion, the findings from the IVM analysis suggest that the risk of early neonatal mortality is not influenced by the mode of delivery. However, given the large discrepancy in risk estimates between analytic methods, health-system level recommendations towards altering local caesarean rates should be avoided until its impact on maternal and neonatal morbidities, hospital costs, and resulting factors are better understood. Future researchers should aim to answer these questions using similar analytic methods to help inform health-care policy makers and providers of the safety of caesarean deliveries.