Assessing the Impact of Blood Loss in Cranial Vault Remodeling: A Risk Assessment Model Using the 2012-2013 Pediatric National Surgical Quality Improvement Program Datasets

Purpose:  Most Cranial Vault Remodeling (CVR) for craniosynostosis is associated with substantial intraoperative blood loss necessitating transfusion.  The transfusion of over 25 mL/kg of red blood cells has long been considered an important safety threshold, with the National Surgical Quality Improvement Program-Pediatric (NSQIP-P) recently identifying this threshold as a significant marker of health care quality. Thus we sought to ascertain risk factors for transfusion in CVR and to quantify the effect of transfusion volume on post-operative complications.

Methods:  Patients who underwent CVR for craniosynostosis were identified from the NSQIP-P database.  Multivariate regression models were used to identify independent risk factors for transfusion and to assess the impact of transfusion on subsequent outcomes.  Multiple methods were used to determine if a transfusion threshold independently predictive of complications existed. 

Results: 1,059 patients met inclusion criteria with 777 patients (73.4%) requiring a transfusion and 520 patients (49.1%) requiring a transfusion in excess of 25 mL/kg. Mean transfusion volume was 52.1±373.8 mL/kg.  Independent risk factors for transfusion included age between 7-9 months (p=0.025), asthma (p=0.035), impaired cognitive status (p=0.012), increasing total procedure RVU (p<0.001), and increasing operative time (p=0.008) (Table 1 and 2).  Neither transfusion nor transfusion volume in excess of 25 mL/kg were associated with an independently increased risk of any type of complication, reoperation, readmission, or increased length of stay following multivariate analysis.

Given that current transfusion thresholds were not predictive of post-operative complications, we attempted to determine a more meaningful threshold.  The top 20% of transfusion volumes were ≥45.28 mL/kg, while recursive partitioning generated a threshold of 62.52 mL/kg.  The threshold obtained via recursive partitioning was able to independently predict a greater number of complications and was associated with higher odds ratios than the quintile method.  A threshold of 60 mL/kg was chosen for simplicity and was independently predictive of overall complications (p=0.009), medical complications (p=0.004), and increased length of stay (p<0.001). 
Conclusions: Transfusion is common in complex CVR.  The current described transfusion occurrence threshold of 25 mL/kg does not accurately convey postoperative risk.  We demonstrate that transfusion in excess of 60 mL/kg significantly increases risk for complications and length of stay in CVR. 

Legends

Table 1 Multivariate Logistic Regression Evaluating Risk Factors for Transfusion

Table 2 Multivariate Logistic Regression Evaluating Risk Factors for Transfusion ≥25mL/kg