Application of Propensity Score Adjustment to Reduce the Survivorship Bias in Assessing Neurocognitive and Functional Outcomes of Children after Early Cardiac Surgery for Congenital Heart Disease

Abstract

Adjusting for survivorship bias for neurocognitive and functional outcome using propensity score and k-means clustering methodDescriptionIn recent years, the number of congenital heart disease survivors has increased with cardiac detection and surgery and perioperative care improvements. However, some survivors may experience poor neurocognitive and functional outcomes after complex cardiac surgery.It could be related to underestimating survivor outcomes due to a lack of practical statistical methods. We proposed an analytical approach to adjust for the severity of illness, using pre-operative and intra-operative differences among children to develop a propensity score and k-means clustering.The propensity score and k-means clustering were used to assess the impact of the severity of illness confounders and compare them with crude results.The analysis included n=235 children with single ventricle congenital heart disease registered at age ≤ 6 weeks in the Western Canadian Complex Pediatric Therapies Follow-up Program between 1997 and 2016.Preoperative, intraoperative, and postoperative variables were collected. The severity of illness propensity score was calculated based on selected variables. Then, a logistic regression model was set up accordingly. Neurocognitive and functional outcomes’ linear time trends showed that 4.5 years after surgery, FSIQ scores stayed the same, VMI scores increased, and ABAS scores decreased over time in high-risk children.Using propensity score adjustment helps clarify the actual trend of neurocognitive and functional outcomes in babies with complex conditions.