Perioperative risk stratification is an integral part of informed consent and medical decision-making processes. It also guides resource allocation and appropriate escalation of care, such as transfer to intensive care settings or tertiary-care hospitals. Fortunately, the incidence of perioperative mortality in children is extremely low.

While the ASA Physical Status Classification System (ASA PS) is commonly used to characterize the patient’s preoperative status, it is a subjective score that does not take into account the complexity of the planned surgical procedure. A recent study examining pediatric-adapted examples of ASA PS demonstrated only moderate interrater reliability among pediatric anesthesiologists (Anesth Analg 2021;132:807-17).

Pediatric perioperative mortality can be roughly predicted based on the child’s comorbidities, physical status at the time of surgery, and the intrinsic risk of the surgical procedure (Anesthesiology 2019;130:971-80). Over the last decade, several risk stratification models have been developed to predict major perioperative adverse events, including death, in children. A recent systematic review identified nine studies that reported the development and internal validation of scores, including a study that validated an existing score (Anesthesiology 2022;137:555-67).

The unadjusted rate of perioperative 30-day mortality ranged from 0.3 to 1.5% when considering the entire pediatric population (0 to 18 years) (Anesthesiology 2022;137:555-67). As expected, 30-day mortality in neonates was much higher, 3.6%, since neonates only have surgery when it is urgently needed and cannot be postponed. All risk models showed good discrimination upon internal validation, but internal validation does not preclude the potential for bias. No single score emerged as qualitatively better than others. The only score included in the systematic review that has undergone an external temporal validation is the Pediatric Risk Assessment (PRAm) score (Anesth Analg 2017;124:1514-19).

The PRAm score predicts perioperative mortality in children undergoing noncardiac surgery and is based on a multivariable regression model using data from the American College of Surgeons National Surgical Quality Improvement Program Pediatric (ACS NSQIP-P) databases (Anesth Analg 2017;124:1514-19). The PRAm score ranges from 0 to 13. The PRAm score considers five domains: whether the patient is undergoing an urgent surgical procedure (+1 point); the presence of at least one of the following comorbidities: respiratory disease, congenital heart disease, preoperative acute or chronic kidney disease, neurologic disease, or hematologic disease (+2 points); the presence of at least one indicator of critical illness (preoperative mechanical ventilation within 48 hours of the operation, inotropic support, or preoperative cardiopulmonary resuscitation; +3 points); age less than 12 months at the time of surgery (+3 points); and whether the patient has a neoplasm (+4 points) (Table).

Table: PRAm score to predict postoperative mortality in infants and children undergoing noncardiac surgery.

Table: PRAm score to predict postoperative mortality in infants and children undergoing noncardiac surgery.

The PRAm score’s internal bootstrap validation was highly accurate in predicting perioperative mortality in pediatric patients undergoing noncardiac surgery. The area under the curve (AUC) was 0.950 in both the derivation cohort (2012 and 2013 ACS NSQIP-P database) and the validation cohort (2014 ACS NSQIP-P database) (Anesth Analg 2017;124:1514-19). The incidence of perioperative mortality was 0.5% (563/115,229) in the derivation cohort and 0.4% (290/67,904) in the validation cohort (Figure 1). A secondary analysis of patients assigned ASA PS classification ≥4 showed wide variability in PRAm scores, further supporting the conclusion that “ASA PS is not suitable as a predictive score for perioperative mortality in individual patients” (Anesth Analg 2017;124:1514-19). It is important to note that this was compared to the ASA PS prior to the pediatric ASA PS publication (Anesth Analg 2021;132:807-17; asamonitor.pub/3vhW4TQ).

Figure 1: Distribution of PRAm score values in the derivation cohort (blue bars) in relation to the observed in-hospital mortality rate (smoothed line) for each score. Reproduced with permission from Nasr VG et al. Development of a pediatric risk assessment score to predict perioperative mortality in children undergoing noncardiac surgery (Anesth Analg 2017;124:1514-9).

Figure 1: Distribution of PRAm score values in the derivation cohort (blue bars) in relation to the observed in-hospital mortality rate (smoothed line) for each score. Reproduced with permission from Nasr VG et al. Development of a pediatric risk assessment score to predict perioperative mortality in children undergoing noncardiac surgery (Anesth Analg 2017;124:1514-9).

External validation of the PRAm score was conducted by prospectively calculating the PRAm score for 13,530 surgical cases at a single institution and performing a receiver operating characteristic curve analysis (Anesth Analg 2019;129:1014-20). The PRAm score was again found to accurately predict 30-day mortality with an AUC of 0.956 (95% confidence interval [CI], 0.938-0.974). PRAm scores of ≥6 and ≤3 were found to be optimal cutoffs for determining an increased or decreased risk for perioperative mortality, respectively. Gray zone analysis identified an inconclusive risk of mortality in 6.93% (938/13,530) of patients who had PRAm scores of 4 or 5 (sensitivity or specificity <90%, respectively), therefore refining the optimal cutoff point to be a PRAm score of ≥6. Mortality in children with an ASA PS ≤3 increased eight-fold when their PRAm score was ≥6 (Anesth Analg 2019;129:1014-20).

“Predicting mortality is not enough. The mortality in children undergoing noncardiac procedures has not changed over the last 10 years. It has remained approximately 0.34%, based on preliminary longitudinal data analysis of the NSQIP pediatric database.”

A limitation of the PRAm score is that intraoperative and postoperative complications that affect perioperative mortality are not included in it (Anesth Analg 2017;124:1514-19). Of note, the PRAm score was not designed to predict perioperative mortality in cardiac surgery, trauma, or in children undergoing solid organ transplantation.

Comprehensive risk assessment should include the intrinsic risk of mortality associated with a specific procedure in addition to the patient’s comorbidities. Using 367,065 surgical cases in the NSQIP Pediatric Database, the intrinsic risk of specific surgical procedures represented by Current Procedural Terminology rather than by broad categorization was determined (Anesth Analg 2021;132:807-17). Four risk quartiles associated with increased 30-day mortality were identified. The predicted risk of 30-day mortality ranged from 0% with no comorbidities to 4.7% when all comorbidities were present among low-risk surgical procedures and from 0.07% to 46.7% among high-risk surgical procedures (Figure 2). Using an external validation cohort of 110,474 observations, the multivariable predictive risk model displayed good calibration and excellent discrimination with an AUC of 0.95 (95% CI, 0.94 to 0.96; P < 0.001).

Figure 2: Probability of 30-day mortality depends on the number of significant comorbidities (body weight <5 kg, ASA PS ≥3, preoperative sepsis, inotropic support, and ventilator dependence). The impact of intrinsic surgical complexity on 30-day mortality is magnified among patients with a great number of comorbidities. Reproduced with permission from Nasr VG, et al. Pediatric risk stratification is improved by integrating both patient comorbidities and intrinsic surgical risk (Anesthesiology 2019;130:971-80).

Figure 2: Probability of 30-day mortality depends on the number of significant comorbidities (body weight <5 kg, ASA PS ≥3, preoperative sepsis, inotropic support, and ventilator dependence). The impact of intrinsic surgical complexity on 30-day mortality is magnified among patients with a great number of comorbidities. Reproduced with permission from Nasr VG, et al. Pediatric risk stratification is improved by integrating both patient comorbidities and intrinsic surgical risk (Anesthesiology 2019;130:971-80).

What are the next steps? The availability of a user-friendly web-based calculator, such as the ACS NSQIP pediatric calculator, will expand the clinical use of the PRAm score (riskcalculator.facs.org/peds; Anesthesiology 2022;137:555-67). The PRAm has been automated from data in the electronic record and validated in a retrospective observational study of children <18 years who underwent noncardiac surgery from 2017 through 2021 at Boston Children’s Hospital (Anesth Analg 2023;136:738-44). The automated PRAm score was developed via electronic derivation of International Classification of Diseases (ICD)-9 and -10 codes. Agreement and correlation among PRAm scores obtained via automation, NSQIP data, and manual physician entry from the same Boston Children’s cohort, as well as the discriminatory ability of the three PRAm versions, was determined. Of the 6,014 patients with NSQIP and automated PRAm scores (manual scores: n = 5,267), the rate of 30-day mortality was 0.18% (n = 11). Agreement and correlation were greater between the NSQIP and automated scores (rho = 0.78; 95% CI, 0.76-0.79; P <.001; ICC = 0.80; 95% CI, 0.79-0.81; Fleiss kappa = 0.66; 95% CI, 0.65-0.67) versus the NSQIP and manual scores (rho = 0.73; 95% CI, 0.71-0.74; P < .001; ICC = 0.78; 95% CI, 0.77-0.79; Fleiss kappa = 0.56; 95% CI, 0.54-0.57). Receiver operating characteristic analysis with AUC showed manual scores to have an AUC of 0.976 (95% CI, 0.959-0.993). The NSQIP score’s AUC was 0.904 (95% CI, 0.792-0.999), and the automated score’s AUC was 0.880 (95% CI, 0.769-0.999). This study demonstrated that it is feasible to develop an electronically derived, automated PRAm score that maintains good discrimination for 30-day mortality in neonates, infants, and children after noncardiac surgery.

The automation of the PRAm score or other risk scores may reduce the preoperative clerical workload and provide an efficient and accurate means to risk stratify neonatal and pediatric surgical patients with the goal of improving clinical outcomes and resource utilization. The next steps are external validation for risk assessment outside the United States and in low-resource settings. In addition, it needs to be determined if the availability of prediction scores influences physician behavior and improves patient outcomes (Anesthesiology 2022;137:526-8).

Scores that predict mortality help us understand the actual risk of death within 30 days of surgery in children. However, predicting mortality is not enough. The mortality in children undergoing noncardiac procedures has not changed over the last 10 years. It has remained approximately 0.34%, based on preliminary longitudinal data analysis of the NSQIP pediatric database. As anesthesiologists with expertise in both patient safety and pediatric physiology, we are well positioned to both identify and mitigate the causes of 30-day mortality after surgery. This formidable challenge represents the next step in our unceasing efforts to improve the safety of patients undergoing surgery.