Mortality Prediction Following Cardiac Surgery in Children - An STS Congenital Heart Surgery Database Analysis

The Society of Thoracic Surgeons Congenital Heart Surgery Database: 2023 Update on Outcomes and Research

The Society of Thoracic Surgeons (STS) Congenital Heart Surgery Database (CHSD) continues to be the most comprehensive database of congenital and pediatric cardiothoracic surgical procedures in the world and contains information on 664,210 operations as of June 30, 2023. The 35th harvest of the STS CHSD data was undertaken in Spring 2023, spanning the 4-year period January 1, 2019, through December 31, 2022, and included 144,919 operations performed at 114 participating sites in North America. The harvest analysis was successfully executed by the STS Research and Analytic Center. The overall unadjusted mortality rate was 2.68% and has remained stable over the 4 years included in the current harvest window. Mortality is highest in neonates (7.4%) and lowest in children (1.1%). As in prior analyses, observed mortality and postoperative length of stay in the database increase with an increase in STS-European Association for Cardio-Thoracic Surgery (STAT) Congenital Heart Surgery Mortality Categories. This quality report summarizes contemporary outcomes, provides the odds ratios for the CHSD risk model variables based on this analysis, and describes on-going efforts to improve data collection and augment analytical approaches. Lastly, 5 research publications completed in the last year using data from the CHSD are also summarized.

Prognostic utility of a novel risk prediction model of 1-year mortality in patients surviving to discharge after surgery for congenital or acquired heart disease

OBJECTIVE

We sought to develop a novel risk prediction model of 1-year mortality after congenital heart surgery that accounts for clinical, anatomic, echocardiographic, and socioeconomic factors.

METHODS

This was a single-center, retrospective review of consecutive index operations for congenital or acquired heart disease, from January 2011 to January 2021, among patients with known survival status at 1 year after discharge from the index hospitalization. The primary outcome was postdischarge mortality at 1 year. Variables of interest included age, prematurity, noncardiac anomalies or syndromes, the Childhood Opportunity Index, primary procedure, major adverse postoperative complications, and the Residual Lesion Score. Logistic regression was used to develop a weighted risk score for the primary outcome. Internal validation using a bootstrap-resampling approach was performed.

RESULTS

Of 10,412 consecutive operations for congenital or acquired heart disease, 8808 (84.6%) cases met entry criteria, including survival to discharge. There were 190 (2.2%) deaths at 1 year postdischarge. A weighted risk score was formulated on the basis of the variables in the final risk prediction model, which included all aforementioned risk factors of interest. This model had a C-statistic of 0.82 (95% confidence interval, 0.80-0.85). The median risk score was 6 (interquartile range, 4-8) points. Patients were categorized as low (score 0-5), medium (score 6-10), high (score 11-15), or very high (score 16-20) risk. The expected probability of mortality was 0.4% ± 0.2%, 2.0% ± 1.1%, 10.1% ± 5.0%, and 36.6% ± 9.6% for low-risk, medium-risk, high-risk, and very high-risk patients, respectively.

CONCLUSIONS

A risk prediction model of 1-year mortality may guide prognostication and follow-up of patients after discharge after surgery for congenital or acquired heart disease.

Congenital Heart Surgery Machine Learning-Derived In-Depth Benchmarking Tool

BACKGROUND

We previously showed that machine learning-based methodologies of optimal classification trees (OCTs) can accurately predict risk after congenital heart surgery and assess case-mix-adjusted performance after benchmark procedures. We extend this methodology to provide interpretable, easily accessible, and actionable hospital performance analysis across all procedures.

METHODS

The European Congenital Heart Surgeons Association Congenital Cardiac Database data subset of 172,888 congenital cardiac surgical procedures performed in European centers between 1989 and 2022 was analyzed. OCT models (decision trees) were built predicting hospital mortality (area under the curve [AUC], 0.866), prolonged postoperative mechanical ventilatory support time (AUC, 0.851), or hospital length of stay (AUC, 0.818), thereby establishing case-adjusted benchmarking standards reflecting the overall performance of all participating hospitals, designated as the "virtual hospital." OCT analysis of virtual hospital aggregate data yielded predicted expected outcomes (both aggregate and for risk-matched patient cohorts) for the individual hospital's own specific case-mix, readily available on-line.

RESULTS

Raw average rates were hospital mortality, 4.9%; mechanical ventilatory support time, 14.5%; and length of stay, 15.0%. Of 146 participating centers, compared with each hospital's overall case-adjusted predicted hospital mortality benchmark, 20.5% statistically (<90% CI) overperformed and 20.5% underperformed. An interactive tool based on the OCT analysis automatically reveals 14 hospital-specific patient cohorts, simultaneously assessing overperformance or underperformance, and enabling further analysis of cohort strata in any chosen time frame.

CONCLUSIONS

Machine learning-based OCT benchmarking analysis provides automatic assessment of hospital-specific case-adjusted performance after congenital heart surgery, not only overall but importantly, also by similar risk patient cohorts. This is a tool for hospital self-assessment, particularly facilitated by the user-accessible online-platform.

Center-level factors associated with shorter length of stay following stage 1 palliation: An analysis of the national pediatric cardiology quality improvement collaborative registry

BACKGROUND

Stage 1 single ventricle palliation (S1P) has the longest length of stay (LOS) of all benchmark congenital heart operations. Center-level factors contributing to prolonged hospitalization are poorly defined.

METHODS

We analyzed data from infants status post S1P included in the National Pediatric Cardiology Quality Improvement Collaborative Phase II registry. Our primary outcome was patient-level LOS with days alive and out of hospital before stage 2 palliation (S2P) used as a balancing measure. We compared patient and center-level characteristics across quartiles for median center LOS, and used multivariable regression to calculate center-level factors associated with LOS after adjusting for case mix.

RESULTS

Of 2,510 infants (65 sites), 2037 (47 sites) met study criteria (61% male, 61% white, 72% hypoplastic left heart syndrome). There was wide intercenter variation in LOS (first quartile centers: median 28 days [IQR 19, 46]; fourth quartile: 62 days [35, 95], P < .001). Mortality prior to S2P did not differ across quartiles. Shorter LOS correlated with more pre-S2P days alive and out of hospital, after accounting for readmissions (correlation coefficient -0.48, P < .001). In multivariable analysis, increased use of Norwood with a right ventricle to pulmonary artery conduit (aOR 2.65 [1.1, 6.37]), shorter bypass time (aOR 0.99 per minute [0.98,1.0]), fewer additional cardiac operations (aOR 0.46 [0.22, 0.93]), and increased use of NG tubes rather than G tubes (aOR 7.03 [1.95, 25.42]) were all associated with shorter LOS centers.

CONCLUSIONS

Modifiable center-level practices may be targets to standardize practice and reduce overall LOS across centers.

Hospital Mortality and Adverse Events Following Repair of Congenital Heart Defects in Developing Countries

BACKGROUND

Mortality associated with the correction of congenital heart disease has decreased to approximately 2% in developed countries and major adverse events are uncommon. Outcomes in developing countries are less well defined. The World Database for Pediatric and Congenital Heart Surgery was utilized to compare mortality and adverse events in developed and developing countries.

METHODS

A total of 16,040 primary procedures were identified over a two-year period. Centers that submitted procedures were dichotomized to low/middle income (LMI) and high income (HI) by the Gross National Income per capita categorization. Mortality was defined as any death following the primary procedure to discharge or 90 days inpatient. Multiple logistic regression models were utilized to identify independent predictors of mortality.

RESULTS

Of the total number of procedures analyzed, 83% (n  =  13,294) were from LMI centers. Among all centers, the mean age at operation was 2.2 years, with 36% (n  =  5,743) less than six months; 85% (n  =  11,307) of procedures were STAT I/II for LMI centers compared with 77% (n = 2127) for HI centers (P < .0001). Overall mortality across the cohort was 2.27%. There was a statistical difference in mortality between HI centers (0.55%) versus LMI centers (2.64%) (P < .0001). After adjustment for other risk factors, the risk of death remained significantly higher in LMI centers (odds ratio: 2.36, 95% confidence interval: 1.707-3.27).

CONCLUSION

Although surgical expertise has increased across the globe, there remains a disparity with some outcomes associated with the correction of congenital heart disease between developing and developed countries. Further studies are needed to identify specific opportunities for improvement.

Operative mortality after Society of Thoracic Surgeons-European Association for Cardio-Thoracic Surgery Mortality Category 1 to 3 procedures: Deficiencies and opportunities for quality improvement

OBJECTIVES

We examined cases of operative mortality at a single quaternary academic center for patients undergoing relatively lower-risk (Society of Thoracic Surgeons-European Association for Cardio-Thoracic Surgery Mortality Category 1-3) procedures, as a means of identifying systemic weaknesses and opportunities for quality improvement.

METHODS

A retrospective review of all operative mortality events for patients who underwent a Society of Thoracic Surgeons-European Association for Cardio-Thoracic Surgery Mortality Category 1, 2, or 3 index procedure (2009-2020) at our institution was performed. After a detailed chart review was performed by 2 independent faculty for each case, factors and system deficiencies that contributed to mortality were identified.

RESULTS

A total of 42 mortalities were identified. A total of 37 patients (88%) had at least 1 Society of Thoracic Surgeons-designated risk factor, including prior cardiac operations (48%), extracardiac malformations (43%), and preoperative ventilation (33%). Eight patients (19%) had non-Society of Thoracic Surgeons-designated preoperative patient-level variables considered as at potential risk, including severe ventricular dysfunction, pulmonary hypertension, lung hypoplasia, and undiagnosed severe coronary abnormalities. Four patients (10%) had no identified preoperative risk factors. After detailed chart review, 5 broad categories were identified: patient-related factors (n = 33; 78%), postoperative infection (n = 13; 31%), postoperative residual lesions (n = 7; 17%), Fontan physiology failure (n = 4; 10%), and unexplained left ventricular failure after tetralogy of Fallot repair (n = 3; 7%). A total of 74% of patients had at least 1 preoperative, intraoperative, or postoperative system deficiency. A total of 50% of surgeries were urgent or emergency.

CONCLUSIONS

Operative mortality after Society of Thoracic Surgeons-European Association for Cardio-Thoracic Surgery Mortality Category 1 to 3 procedures is related to the presence of multifactorial risk patterns (Society of Thoracic Surgeons and non-Society of Thoracic Surgeons-designated patient-level risk factors and variables, broad risk categories, system deficiencies, emergency surgery). A multidisciplinary approach to care, with early recognition and treatment of modifiable additional burdens, could reduce this risk.

A machine learning approach to predicting 30-day mortality following paediatric cardiac surgery: findings from the Australia New Zealand Congenital Outcomes Registry for Surgery (ANZCORS)

OBJECTIVES

We aim to develop the first risk prediction model for 30-day mortality for the Australian and New Zealand patient populations and examine whether machine learning (ML) algorithms outperform traditional statistical approaches.

METHODS

Data from the Australia New Zealand Congenital Outcomes Registry for Surgery, which contains information on every paediatric cardiac surgical encounter in Australian and New Zealand for patients aged <18 years between January 2013 and December 2021, were analysed (n = 14 343). The outcome was mortality within the 30-day period following a surgical encounter, with ∼30% of the observations randomly selected to be used for validation of the final model. Three different ML methods were used, all of which employed five-fold cross-validation to prevent overfitting, with model performance judged primarily by the area under the receiver operating curve (AUC).

RESULTS

Among the 14 343 30-day periods, there were 188 deaths (1.3%). In the validation data, the gradient-boosted tree obtained the best performance [AUC = 0.87, 95% confidence interval = (0.82, 0.92); calibration = 0.97, 95% confidence interval = (0.72, 1.27)], outperforming penalized logistic regression and artificial neural networks (AUC of 0.82 and 0.81, respectively). The strongest predictors of mortality in the gradient boosting trees were patient weight, STAT score, age and gender.

CONCLUSIONS

Our risk prediction model outperformed logistic regression and achieved a level of discrimination comparable to the PRAiS2 and Society of Thoracic Surgery Congenital Heart Surgery Database mortality risk models (both which obtained AUC = 0.86). Non-linear ML methods can be used to construct accurate clinical risk prediction tools.

The Society of Thoracic Surgeons Congenital Heart Surgery Database: 2022 Update on Outcomes and Research

The Society of Thoracic Surgeons (STS) Congenital Heart Surgery Database continues to be one of the most comprehensive clinical outcomes registries capturing almost all pediatric cardiothoracic surgical operations undertaken in the United States. The latest analysis of aggregate outcomes was performed after the 33rd data harvest and included congenital and pediatric cardiac operations performed between July 1, 2017 and June 30, 2021. This article summarizes these contemporary outcomes and provides a context for the interpretation of these outcomes. In addition this article describes ongoing efforts to improve data collection and augment analytical approaches. Finally, research activities undertaken in the last year using data from the database are also summarized.

Variation in Case-Mix Across Hospitals: Analysis of The Society of Thoracic Surgeons Congenital Heart Surgery Database

BACKGROUND

The Society of Thoracic Surgeons (STS) Congenital Heart Surgery Database was queried to document variation of patient characteristics, procedure types, and programmatic case-mix.

METHODS

All index cardiac operations in patients less than 18 years of age in the STS Congenital Heart Surgery Database (July 2016 to June 2020) were eligible for inclusion except patients weighing ≤2.5 kg undergoing isolated patent ductus arteriosus closure. At the hospital level, we describe variations in patient and procedural characteristics known from previous analyses to be associated with outcomes. We also report variations across hospitals of programmatic case-mix.

RESULTS

Data were analyzed from 117 sites (90 322 total operations, 87 296 total index cardiac operations eligible for STAT [STS-European Association for Cardio-Thoracic Surgery] 2020 Mortality Score). The median annual total index cardiac operations eligible for STAT 2020 Mortality Score per hospital was 157 (interquartile range [IQR], 94-276). Wide variability was documented in total annual index cardiac operations eligible for STAT 2020 Mortality Score per hospital (ratio 90th/10th percentile = 9.01), operations in neonates weighing <2.5 kg (ratio 90th/10th percentile = 4.09), operations in patients with noncardiac anatomic abnormalities (ratio 90th/10th percentile = 3.46), and operations in patients with preoperative mechanical ventilation (ratio 90th/10th percentile = 3.97). At the hospital level, the median percentage of all index cardiac operations in STAT 2020 Mortality Category 5 was 3.7% (IQR, 1.7%-4.9%), the median percentage of all index cardiac operations in STAT 2020 Mortality Category 4 or 5 was 24.4% (IQR, 19.0%-28.4%), the median hospital-specific mean STAT Mortality Category was 2.39 (IQR, 2.20-2.47), and the median hospital-specific mean STAT Mortality Score was 0.86 (IQR, 0.73-0.91).

CONCLUSIONS

Substantial variation of patient characteristics, procedure types, and case-mix exists across pediatric and congenital cardiac surgical programs. Knowledge about programmatic case-mix augments data about indirectly standardized programmatic observed-to-expected (O/E) mortality. Indirectly standardized O/E ratios do not provide a complete description of a given pediatric and congenital cardiac surgical program. The indirectly standardized programmatic O/E ratios associated with a given program apply only to its specific case-mix of patients and may represent a quite different case-mix than that of another program.

Development of a Novel Society of Thoracic Surgeons Adult Congenital Mortality Risk Model

BACKGROUND

Operative mortality risk models for adults with congenital heart disease (ACHD) undergoing cardiac operations are essential, given the growing population of these patients, yet they are currently unavailable. Existing adult Society of Thoracic Surgeons (STS) models exclude congenital procedures, whereas existing congenital models exclude operations for acquired disease. We aimed to develop an STS mortality risk model for ACHD patients undergoing cardiac operations.

METHODS

Leveraging a comprehensive list of diagnostic and procedure codes, ACHD patients who underwent cardiac operations were identified from the STS Adult Cardiac Surgery Database (versions: v2.73, v2.81, and v2.9) between 2011 and 2019. The model was developed and validated in the ACHD population using a 60/40 development/validation split. Univariate analyses and clinical expertise informed the addition of ACHD-relevant procedure and diagnosis variables to existing STS adult risk model variables. Model performance was assessed overall and in 38 subgroups based on patient demographics, procedures, and diagnoses.

RESULTS

Forty-seven procedure and diagnosis variables relevant to ACHD were added to existing STS adult risk model variables. The derived ACHD model for operative mortality was well calibrated within demographic, procedural, and diagnosis subgroups and the overall ACHD population, and discrimination in the validation cohort was excellent (C statistic, 0.815) compared with the model using only existing STS adult risk model variables (C statistic, 0.79; P < .0001).

CONCLUSIONS

A novel, high-performing STS ACHD mortality risk model has been developed on the basis of contemporary patient data. The ACHD risk model represents an important expansion of the STS portfolio. Implementation with an online risk calculator is planned.

Performance of a novel risk model for deep sternal wound infection after coronary artery bypass grafting

Clinical prediction models for deep sternal wound infections (DSWI) after coronary artery bypass graft (CABG) surgery exist, although they have a poor impact in external validation studies. We developed and validated a new predictive model for 30-day DSWI after CABG (REPINF) and compared it with the Society of Thoracic Surgeons model (STS). The REPINF model was created through a multicenter cohort of adults undergoing CABG surgery (REPLICCAR II Study) database, using least absolute shrinkage and selection operator (LASSO) logistic regression, internally and externally validated comparing discrimination, calibration in-the-large (CL), net reclassification improvement (NRI) and integrated discrimination improvement (IDI), trained between the new model and the STS PredDeep, a validated model for DSWI after cardiac surgery. In the validation data, c-index = 0.83 (95% CI 0.72–0.95). Compared to the STS PredDeep, predictions improved by 6.5% (IDI). However, both STS and REPINF had limited calibration. Different populations require independent scoring systems to achieve the best predictive effect. The external validation of REPINF across multiple centers is an important quality improvement tool to generalize the model and to guide healthcare professionals in the prevention of DSWI after CABG surgery.

Re-evaluating Congenital Heart Surgery Center Performance Using Operative Mortality

BACKGROUND

The Society of Thoracic Surgeons' Congenital Heart Surgery Database provides observed to expected (O/E) operative mortality ratios to more than 100 congenital heart centers in North America. We compared the current approach (STS-CHSD) for estimating O/E ratios to approaches incorporating information on diagnosis as moderators of procedures, other unutilized risk factors, and additional variation in confidence interval (CI) construction to characterize center performance.

METHODS

Bayesian additive regression trees (BART) and lasso models linked operative mortality to diagnosis-procedure categories, procedure-specific risk factors, and syndromes/abnormalities. Bootstrapping accounted for variation in STS-CHSD (STS Bootstrap) and lasso CIs. We compared O/E estimates, interquartile range (IQR) of CI widths, and concordance of center performance categorizations (worse-than-, as-, or better-than-expected mortality) of the new approaches to the STS-CHSD.

RESULTS

In 110 surgical centers including 98,822 surgical operative encounters, there were 2818 (2.85%) operative mortalities (center range: 0.37% to 10%). Compared to the STS-CHSD, BART and lasso estimated O/E ratios varied more and had narrower CIs (IQR of CI: STS-CHSD = 1.11, STS Bootstrap = 0.98; lasso = 0.80; BART = 0.96). Concordance of performance categorization with the STS-CHSD ranged from 84% (lasso) to 91% (STS Bootstrap); more than 70% of discordant centers improved categories. Discordant centers had smaller volumes, fewer operative mortalities, and treated more patients with congenital lung abnormalities.

CONCLUSIONS

Relative to the STS-CHSD, up to 16% of hospitals changed performance categories, most improving performance. Given the significance of quality reports for congenital heart centers, inclusion of additional risk factors and unaddressed variation should be considered.

Organization of Pediatric Echocardiography Laboratories: Impact of Sonographers on Clinical, Academic, and Financial Performance

Echocardiography has evolved the first-line imaging for diagnosis and management of pediatric and congenital heart disease all over the world. While it recognized as essential component of pediatric cardiac care delivery, organization of pediatric echocardiography services is very heterogeneous across the world, mainly related to significant differences in material and human resources in heterogeneous health care systems. In this paper, we focus on the role of pediatric sonographers, defined as expert technicians in pediatric echocardiography. While in some services sonographers are an essential part of the organizational structure, other laboratories operate only with physicians trained in echocardiography. The impact of sonographers on clinical, academic and financial performance will be discussed. Two organizational models (with and without sonographers) will be compared, and the advantages and disadvantages of each model will be evaluated. Different models of care provision are possible and decisions on organizational models need to be adjusted to the demands and available resources.