Response to Reader Comment on “Risk Factors and Causes of Readmission to the Pediatric Intensive Care Unit After Cardiac Surgery”

We thank A. Sharma et al for their insightful comments and interest in our study on risk factors for unplanned PICU readmission after pediatric cardiac surgery.^1,2 We appreciate the opportunity to clarify the methodological choices and acknowledge the limitations inherent to studying rare outcomes.

Matching and Regression Approach

We agree that conditional logistic regression is generally recommended for individually matched case–control designs to account for the matched sets and avoid biased estimates. ³ Our decision to use unconditional logistic regression was based on the small sample size and sparse data within strata, which led to unstable estimates when conditional models were attempted. This approach is supported in the medical literature. ⁴ We acknowledge this as a limitation and plan to reanalyze using stratified or penalized conditional approaches in our future work with larger cohorts.

Sample Size and Model Complexity

The limited number of events (n = 23) is a recognized constraint. As reported, the recent literature emphasizes that low events-per-variable ratios increases the risk of overfitting and inflates effect sizes. We attempted to mitigate this by restricting candidate predictors and prioritizing clinically plausible variables. Nonetheless, we concur that penalized regression methods such as Least Absolute Shrinkage and Selection Operator (LASSO) could improve stability in small samples, ⁵ even if it is not ideal. These approaches will be considered in future analyses.

Variable Selection

We acknowledge the limitations of univariable screening for variable selection. Methodological reviews caution that this practice can lead to biased estimates and poor reproducibility. Our approach combined statistical screening with clinical judgment to balance parsimony and relevance. We agree that the large odds ratio for single ventricle physiology should be interpreted as hypothesis-generating rather than definitive. However, this result was also validated by clinicians in charge of the patients and resulted in modifications of the management protocol of these patients.

Severity Score Clarification

The severity score used for matching and adjustment was the modified PELOD-2 score. ⁶ We used the modified PELOD 2 score because, as stated in the Table 2 legend of our article, this score excludes neurological assessment that is impaired in all patients due to the sedation in the pediatric intensive care unit after cardiac surgery. We will ensure this is more clearly stated in future reports to facilitate comparability with other cohorts.

In conclusion, we appreciate the constructive feedback. Despite these limitations, we believe our study provides valuable descriptive insights and highlights an important clinical issue warranting further investigation.