Frequency and predictors of concurrent complications in multi-suture release for syndromic craniosynostosis

This retrospective cohort study used the American College of Surgeons (ACS) National Surgical Quality Improvement Program-Pediatric (NSQIP-P) database. NSQIP-P includes 30-day perioperative and surgical outcomes for individuals ≤ 18 years. Variables are collected from > 60 pediatric institutions on an 8-day systematic sampling cycle that allows for proportional diversity in selection [11]. The NSQIP-P data are audited and validated continually and organized into Participant User Files (PUFs). For the current investigation, the 2012–2020 data were used [12]. The use of ICD-9 and ICD-10 codes was based on the method of coding in the NSQIP-P records during the inclusion years of the study (2012–20). Because NSQIP-P PUF datasets do not have patient identifying information, the study was exempt from institutional ethics review.

Patients ≤ 3 years of age with a syndromic diagnosis who underwent multi-suture craniosynostosis release and other cranial reconstruction procedures were included. The classification system for diagnoses of “congenital malformations” in the NSQIP-P was the International Classification of Diseases-9 and for procedures was the Current Procedural Terminology (CPT). The included diagnoses were Apert, Pfeiffer, or Saethre-Chotzen syndrome (ICD-9 code 755.55), craniofrontal dysplasia (759.81), Carpenter syndrome (759.89), and Crouzon’s/other congenital and musculoskeletal anomalies (756). The included procedures were craniotomy for craniosynostosis—frontal or parietal bone flap and bifrontal bone flap (CPT codes 61556 and 61557); extensive craniectomy for multiple suture craniosynostosis, not requiring and requiring bone grafts (61558 and 61559); reconstruction, bifrontal, superior-lateral orbital rims and lower forehead, advancement or alteration (e.g., plagiocephaly, trigonocephaly, brachycephaly), with or without grafts (includes obtaining autografts) (21175); and repair, revision, and/or reconstruction procedures on the head (21180). Only primary CPT codes were used for identification. Patients who underwent emergent surgery or had prior surgery within 30 days were excluded from the analysis.

The primary outcome was concurrent complications (≥ 2 complications occurring intra- or perioperatively up to 30 days affecting multiple organ systems). Secondary outcomes included non-home discharge, days of mechanical ventilation, readmission, length of stay, and death.

Frequencies of each complication were calculated, cross-tabulated, and ranked to identify the most frequent concurrent complications.

A random forest analysis was performed using Python (v1.1.2, https://​www.​python.​org/​) to identify patient and treatment factors associated with the risk of concurrent complications. Top factors were identified, and binary logistic regression was performed to identify odds ratios (ORs) for the most influential factors for the primary outcome. The comparison group was children with syndromic disease who experienced zero or one complication.

Correlation analyses were used to identify relationships between the number of perioperative complications and the secondary outcomes. Because of the high rate of blood transfusion, univariate logistic regression was performed to identify the odds of having any complication when the patient received blood transfusion.

Descriptive and logistic regression analyses were performed using the SPSS statistical software (v26). A p-value of < 0.05 was considered statistically significant. The STROBE checklist was used for study reporting.

A total of 5,848 children who underwent multi-suture synostosis surgery were included (Fig. 1). The frequency of procedures steadily increased over time (Fig. 2, top). The median age of the entire cohort was 0.69 years [interquartile range (IQR) 0.43–1.00], and 3771 (64.48%) of children were male. Most children were white (4215, 72.08%), and 4158 (71.10%) had a body mass index < 18.5 kg/m² (Table 1).

Nearly 700 (11.94%) children were documented as having impaired cognitive status at time of surgery, and 226 (3.86%) children were on nutritional support. Oxygen support was present in 68 (1.16%) children and 48 (0.82%) children were ventilator dependent at the time of surgery. Nearly 2% of children (116) had a history of asthma and 161 (2.75%) had a bronchopulmonary disorder; 497 (8.50%) had a history of gastrointestinal disease. Cardiac risk factors were present in ~ 10% of children. Additionally, 77 (1.32%) children had a seizure disorder, 1236 (21.14%) were diagnosed with a structural central nervous system abnormality, 151 (2.58%) had a hematologic disorder, and 32 (0.55%) were on inotropic support at the time of surgery.

The most common procedures were extensive craniectomy for multiple cranial suture craniosynostosis requiring bone grafts (CPT 61559) and bifrontal reconstruction of superior-lateral orbital rims and lower forehead, advancement or alteration (CPT 21175) (approximately 80% of cases). Most (99.84%) patients were coded as Crouzon’s/other syndromic disease (ICD-9 756).

The median length of hospital stay was 5 days [IQR 4–7 days] (Table 2), with 22 (0.38%) requiring > 30 days in the hospital. After surgery, the mean number of days on mechanical ventilation was 0.18 ± 1.81, but 5 patients (0.09%) required > 30 days of mechanical ventilation. At discharge, 5,802 (99.21%) patients went home. One child (0.02%) died before discharge, another died within 30 days of surgery, and 17 (0.29%) were unaccounted for.

Among the 5,848 patients, 1741 (29.77%) did not experience a complication. The overall perioperative complication rate was 70.23% (4107/5848), and the overall concurrent complication rate (≥ 2) was 2.75% (161/5848). Stratified by number of complications, 3946 (67.48%) children experienced one complication and 129 (2.21%), 26 (0.44%), and 6 (0.11%) experienced two, three, and four or more complications, respectively. Table 3 demonstrates the frequency of specific complications. The most common complication was hematologic, with 69.53% of all patients requiring transfusion, despite the overall incidence of blood transfusion steadily decreasing from 2012 to 2020 (Fig. 2, bottom). There were 244 nonhematologic complications, of which superficial infection/surgical site infection was the most prevalent (1.08%).

One hundred sixty-one children (2.75%) experienced > 1 complication. When assessing complications by body system, 82% of concurrent complications were hematologic/infectious and 59% hematologic/wound-related. Table 4 demonstrates the most frequently concurrent complications, which were led by transfusion/superficial infection (27.95%), transfusion/deep incisional surgical site infection (13.04%), and transfusion/sepsis (13.04%). The top 10 concurrent complications all included bleeding with subsequent transfusion. The most frequent concurrent complications excluding bleeding/transfusion were site-specific/systemic infections (5 children, 3.11%), cerebrovascular accident/seizure (4 children, 2.48%), and infection/wound disturbance (4 children, 2.48%).

Six statistically significant factors were associated with the outcome of concurrent complications (Table 5), including five that increased the risk of concurrent complications. Two cardiac-related factors were elucidated: major cardiac risk factors (OR 3.50 [95% CI 1.92–6.38]) and previous cardiac surgery (OR 4.87 [2.36–10.04]). Two pulmonary factors were also identified: preoperative ventilator dependence (OR 3.27 [1.16–9.21]) and structural pulmonary/airway abnormalities (OR 2.67 [1.27–5.63]). Finally, an additional risk factor was preoperative nutritional support (OR 4.05 [2.34–7.01]). Non-premature birth was found to be protective against co-occurring complications (OR 0.48 [0.35–0.67]). No surgical treatment factors were identified as significant risk factors associated with concurrent complications.

Having a greater number of concurrent complications weakly correlated with hospital readmission (r = 0.25; p < 0.001). No correlation existed between number of concurrent complications and total hospital stay, number of ventilation days, non-home discharge, re-admission, and death (Table 6). Given the prevalence of transfusion, further analysis revealed blood transfusion was associated with higher risk of deep incisional surgical site infection (OR 4.62 [1.08–19.73] (p = 0.04).

We found an overall complication rate of 70.23%, which reduced to 30% when bleeding requiring transfusion was excluded, which is significantly higher than reported previously [13]. Perioperative bleeding is significant during craniosynostosis procedures and comprised 94% of all complications in this study because blood transfusion is considered a complication in the NSQIP-P data. Allogenic blood is commonly transfused during CVR, with some institutions reporting transfusion rates as high as 100% [14]. Although some argue that blood transfusion is “expected,” we argue that transfusion should be considered a complication because of the potential consequences of the intervention. Interestingly, the rate of blood transfusion is decreasing over time, reflecting an improvement in surgical technique, attention to blood loss, and methods to reduce intraoperative bleeding.

Six of the most common concurrent complications identified in this study were hematologic/infectious, suggesting a possible relationship between bleeding requiring transfusion and perioperative infection. We found that blood transfusion was associated with higher risk of deep incisional surgical site infection (OR 4.62 [1.08–19.73]. This is similar to a meta-analysis of 23 prospective controlled clinical trials, which observed that postoperative allogenic blood transfusions across various procedures were associated with a significantly increased odds of postoperative bacterial infections (common OR 5.26 [5.03–5.43]) [15]. The specific causative factor of this association remains unclear; however, it is evident that allogeneic blood leads to a greater degree of postoperative immunosuppression than autologous transfusions or no transfusion. Therefore, it is compelling to augment intraoperative maneuvers with the aim of reducing bleeding/transfusions and subsequent infectious complications as well. A 2022 study suggested that the use of tranexamic acid in a dose-independent manner for craniosynostosis procedures reduced the rates of perioperative bleeding and transfusion [16]. Another study reported that maintaining normothermia during craniosynostosis repair, requiring pretransfusion surgical consultation, and using a preincisional bolus of tranexamic acid and incisional lidocaine with epinephrine reduced the institutional transfusion rate from 100% to 22.7% over 5 years [17]. Institutional adoption of clinical guidelines for anesthesia and surgery may explain the proportional decrease in bleeding/transfusions observed in our study.

Nonhematologic complication rates vary in the literature because of the differing data sources and classification schemes chosen by authors. One small retrospective review observed a postoperative complication rate of 2.86% (one patient experiencing wire protrusion) when bleeding/transfusion was excluded [18]. Despite a reoperation rate of 8.6%, the authors concluded that open craniosynostosis surgery can be performed with minimal complications, low recurrence rates, and satisfactory cosmetic outcomes. A higher complication rate (35.9%) was reported by Shastin et al. [19] using a larger sample size and including only “excessive” bleeding/transfusion as a complication. The inclusion of complications > 30 days in both inpatient and outpatient settings may have contributed to their higher reported rate. A 2012 study comparing complication rates from the NSQIP-P and the Kids’ Inpatient Database (KID) described rates excluding blood transfusion at 2.9% and 3.0%, respectively [20]. For comparison, our nonhematologic complication rate was 4.17%, which is likely higher because we included only children with syndromic diagnoses. Future studies should focus on postoperative complication rates by using common data elements and standardized definitions.

Five clinical factors were identified as having a statistically significant association with increased risk of concurrent complications. These were major cardiac risk factors, previous cardiac surgery, preoperative ventilator dependence, structural pulmonary/airway abnormalities, and preoperative nutritional support.

There are limitations to this investigation. The NSQIP-P database is retrospective, and although craniosynostosis procedures are captured, the database was not built specifically for these patients. However, NSQIP-P accurately captures patient characteristics, treatment factors, and early postoperative complications because of the ACS’s strict data integrity requirements for participating hospitals and trained reviewers. In NSQIP-P, there is no detailed information regarding the severity of deformity and correction, procedural approaches, or surgical techniques. Although outdated, the use of ICD-9 and ICD-10 codes was based on coding methods used in the database during the inclusion years of the study (2012–20). The use of CPT codes is the most accurate way to search and define surgical treatment using the NSQIP-P data. This may limit the global applicability of this study, but there is precedence for a study like this [30] and currently over 60 countries use CPT terminology to date [31].

The procedural codes used here are a modest proxy for risk and severity, although indirect. There is substantial difference in procedural morbidity in syndromic synostosis procedures depending on the indication and specific procedure (often encompassed under the same CPT code) and expected associated complications. It is difficult to parse out complications by procedure because of this heterogeneity, which limits the extrapolation of the results. In addition, 99% of the diagnoses were Crouzon/other syndromic disease (ICD-9 756), which encompasses a wide array of conditions, although using a national administrative database is useful for studying conditions with low incidence and low complication rates.

For this study, we chose to include children ≤ 3 years to consider only primary procedures and prohibit inclusion of older children at higher risk for complications. Although institutional thresholds for blood transfusion are often present, they are not standardized among NSQIP-P centers. Furthermore, it is unclear how NSQIP-P defines developmental delay for infants, making it difficult to ascertain its significance. Furthermore, the NSQIP-P database is based on data collected up to 30 days postoperatively so longer-term complications and concurrent complications are not captured. Finally, we would caution international generalizability with respect to the codes used for study inclusion as the cohort was derived from a US sample.