Background The COVID-19 pandemic necessitated major reallocation of healthcare services. Our aim was to assess the impact on paediatric congenital heart disease (CHD) procedures during different pandemic periods compared with the prepandemic period, to inform appropriate responses to future major health services disruptions.Methods and results We analysed 26 270 procedures from 17 860 children between 1 January 2018 and 31 March 2022 in England, linking them to primary/secondary care data. The study period included prepandemic and pandemic phases, with the latter including three restriction periods and corresponding relaxation periods. We compared procedure characteristics and outcomes between each pandemic period and the prepandemic period. There was a reduction in all procedures across all pandemic periods, with the largest reductions during the first, most severe restriction period (23 March 2020 to 23 June 2020), and the relaxation period following second restrictions (3 December 2020 to 4 January 2021) coinciding with winter pressures. During the first restrictions, median procedures per week dropped by 51 compared with the prepandemic period (80 vs 131 per week, p=4.98×10−08). Elective procedures drove these reductions, falling from 96 to 44 per week (p=1.89×10−06), while urgent (28 vs 27 per week, p=0.649) and life-saving/emergency procedures (7 vs 6 per week, p=0.198) remained unchanged. Cardiac surgery rates increased, and catheter-based procedure rates reduced during the pandemic. Procedures for children under 1 year were prioritised, especially during the first four pandemic periods. No evidence was found for differences in postprocedure complications (age-adjusted OR 1.1 (95% CI 0.9, 1.4)) or postprocedure mortality (age and case mix adjusted OR 0.9 (95% CI 0.6, 1.3)).Conclusions Prioritisation of urgent, emergency and life-saving procedures during the pandemic, particularly in infants, did not impact paediatric CHD postprocedure complications or mortality. This information is valuable for future major health services disruptions, though longer-term follow-up of the effects of delaying elective surgery is needed.

Atrial fibrillation (AF) is the most frequent cardiac arrhythmia, with an estimated five million cases globally. This condition increases the likelihood of developing cardiovascular complications such as thromboembolic events, with a fivefold increase in risk of both heart failure and stroke. Contemporary challenges include a better understanding AF pathophysiology and optimizing therapeutical options due to the current lack of efficacy and adverse effects of antiarrhythmic drug therapy. Hence, the identification of novel biomarkers in biological samples would greatly impact the diagnostic and therapeutic opportunities offered to AF patients. Long noncoding RNAs, micro RNAs, circular RNAs, and genes involved in heart cell differentiation are particularly relevant to understanding gene regulatory effects on AF pathophysiology. Proteomic remodeling may also play an important role in the structural, electrical, ion channel, and interactome dysfunctions associated with AF pathogenesis. Different devices for processing RNA and proteomic samples vary from RNA sequencing and microarray to a wide range of mass spectrometry techniques such as Orbitrap, Quadrupole, LC-MS, and hybrid systems. Since AF atrial tissue samples require a more invasive approach to be retrieved and analyzed, blood plasma biomarkers were also considered. A range of different sample preprocessing techniques and bioinformatic methods across studies were examined. The objective of this descriptive review is to examine the most recent developments of transcriptomics, proteomics, and bioinformatics in atrial fibrillation.

OBJECTIVES:The last 2 decades have seen an incremental use of biological over mechanical prostheses. However, while short-term clinical outcomes are largely equivalent, there is still controversy about long-term outcomes. METHODS:All patients between the ages of 50 and 70 years undergoing elective/urgent isolated aortic valve replacement at our institute between 1996 and 2023 were included. Trends, early, and long-term outcomes were investigated. RESULTS:A total of 1708 (61% male) patients with a median age of 63.60 (interquartile range: 58.28-67.0) years were included of which 1191 (69.7%) received a biological prosthesis. After inverse propensity score weighting, there were no short-term differences when comparing patients receiving biological and mechanical valves. However, patients who received mechanical prostheses had better long-term survival (P < 0.001). Sub-group analysis revealed that patients with biological size 19 mm prosthesis had the worst long-term survival. Patients with a size 21-mm mechanical prosthesis had better survival compared to both size 19-mm [hazard ratio (HR) 0.25, 95% confidence interval (CI) 0.17-0.37, P < 0.001], 21-mm (HR 0.33, 95% CI 0.23-0.48, P < 0.001) and 23-mm (HR 0.40, 95% CI 0.27-0.60, P < 0.001) biological prosthesis. Additionally, patients with severe patient-prosthesis mismatch exhibited the lowest survival rate compared to those with moderate or no (HR 1.56, 95% CI 1.21-2.00, P < 0.001). CONCLUSIONS:Patients aged between 50 and 70 years with a mechanical aortic prosthesis had better long-term survival compared to those with a biological prosthesis. Our study underscores the need for a critical re-evaluation of prosthesis selection strategies in this age group.

BackgroundRedo sternotomy aortic root surgery is technically demanding, and the evidence on outcomes is mostly from retrospective, small sample, single-centre studies. We report the trend, early clinical results and outcome predictors of redo aortic root replacement over 20 years in the United Kingdom.MethodsWe retrospectively analysed collected data from the UK National Adult Cardiac Surgery Audit (NACSA) on all redo sternotomy aortic root replacements performed between 30th January 1998 and 19th March 2019. We analysed trends in the volume of operations, characteristics of hospital survivors vs. non-survivors, and predictors of in-hospital outcomes.ResultsDuring the study period, 1,107 redo sternotomy aortic root replacements were performed (median age 59, 26% of patients were females). Eighty-four per cent of cases (N = 931) underwent a composite root replacement, 11% (N = 119) had homograft root replacement and valve-sparing root replacement was performed in 5.1% (N = 57) of cases. There was a steady increase in the volume of redo sternotomy root replacements beyond 2006, from an annual volume of 22 procedures in 2006 to 106 procedures in 2017. Hospital mortality was 17% (n = 192), postoperative stroke or TIA occurred in 5.2% (n = 58), and postoperative dialysis was required in 11% (n = 109) of patients. Return to the theatre for bleeding/tamponade was required in 9% (n = 102) and median in-hospital stay was 9 days. Age >59 (OR: 2.99, CI: 1.92–4.65, P < 0.001), recent myocardial infarction (OR: 6.42, CI: 2.24–18.41, P = 0.001) were associated with increased in-hospital mortality. Emergency surgery (OR: 3.95, 2.27–6.86, P < 0.001), surgery for endocarditis (OR: 2.05, CI: 1.26–3.33, P = 0.001), salvage coronary artery bypass grafting (OR: 2.20, CI: 1.37–3.54, P < 0.001), arch surgery (OR: 2.47, CI: 1.30–3.61, P = 0.018) and aortic cross-clamp longer than 169 min (OR: 2.17, CI: 1.00–1.01, P = 0.003) were associated with increased risk of mortality. We found no effect of the centre or surgeon volume on mortality (P > 0.05).ConclusionsRedo sternotomy aortic root replacement still carries significant morbidity and mortality and is sporadically performed across surgeons and centres in the UK.

OBJECTIVES:A study of the performance of in-hospital/30-day mortality risk prediction models using an alternative machine learning algorithm (XGBoost) in adults undergoing cardiac surgery. METHODS:Retrospective analyses of prospectively routinely collected data on adult patients undergoing cardiac surgery in the UK from January 2012 to March 2019. Data were temporally split 70:30 into training and validation subsets. Independent mortality prediction models were created using sequential backward floating selection starting with 61 variables. Assessments of discrimination, calibration, and clinical utility of the resultant XGBoost model with 23 variables were then conducted. RESULTS:A total of 224,318 adults underwent cardiac surgery during the study period with a 2.76% (N = 6,100) mortality. In the testing cohort, there was good discrimination (area under the receiver operator curve 0.846, F1 0.277) and calibration (especially in high-risk patients). Decision curve analysis showed XGBoost-23 had a net benefit till a threshold probability of 60%. The most important variables were the type of operation, age, creatinine clearance, urgency of the procedure and the New York Heart Association score. CONCLUSIONS:Feature-selected XGBoost showed good discrimination, calibration and clinical benefit when predicting mortality post-cardiac surgery. Prospective external validation of a XGBoost-derived model performance is warranted.

OBJECTIVES:This nationwide retrospective cohort study assessed the impact of the explanted valve type on reoperative outcomes in aortic valve surgery within the UK over a 23-year period. METHODS:Data were sourced from the National Institute for Cardiovascular Outcomes Research (NICOR) database. All patients undergoing first-time isolated reoperative aortic valve replacement between 1996 and 2019 in the UK were included. Concomitant procedures, homograft implantation or aortic root enlargement were excluded. Propensity score matching was utilized to compare outcomes and risk factors for in-hospital mortality was evaluated through multivariable logistic regression. Final model selection was conducted using Akaike Information Criterion through bootstrapping. The primary end point was in-hospital mortality, and secondary end points included postoperative morbidities. RESULTS:Out of 2371 patients, 24.9% had mechanical and 75% had bioprosthetic valves implanted during the primary procedure. Propensity matched groups of 324 patients each, were compared. In-hospital mortality for mechanical and bioprosthetic valve explants was 7.1% and 5.9%, respectively (P = 0.632). On multivariable logistic regression analysis, valve type was not a risk factor for mortality [odds ratio (OR) 0.62, 95% confidence interval (CI) 0.37-1.05; P = 0.1]. Age (OR 1.03, 95% CI 1.01-1.05; P < 0.05), left ventricular ejection fraction (OR 1.62, 95% CI 1.08-2.42; P < 0.05), creatinine ≥ 200 mg/dl (OR 2.21, 95% CI 1.17-4.04; P < 0.05) and endocarditis (OR 2.66, 95% CI 1.71-4.14; P < 0.05) emerged as risk factors for mortality. CONCLUSIONS:The type of valve initially implanted (mechanical or bioprosthetic) did not determine mortality. Instead, age, left ventricular ejection fraction, renal impairment and endocarditis were significant risk factors for in-hospital mortality.

BACKGROUND:Ensemble tree-based models such as Xgboost are highly prognostic in cardiovascular medicine, as measured by the Clinical Effectiveness Metric (CEM). However, their ability to handle correlated data, such as hospital-level effects, is limited. OBJECTIVES:The aim of this work is to develop a binary-outcome mixed-effects Xgboost (BME) model that integrates random effects at the hospital level. To ascertain how well the model handles correlated data in cardiovascular outcomes, we aim to assess its performance and compare it to fixed-effects Xgboost and traditional logistic regression models. METHODS:A total of 227,087 patients over 17 years of age, undergoing cardiac surgery from 42 UK hospitals between 1 January 2012 and 31 March 2019, were included. The dataset was split into two cohorts: training/validation (n = 157,196; 2012-2016) and holdout (n = 69,891; 2017-2019). The outcome variable was 30-day mortality with hospitals considered as the clustering variable. The logistic regression, mixed-effects logistic regression, Xgboost and binary-outcome mixed-effects Xgboost (BME) were fitted to both standardized and unstandardized datasets across a range of sample sizes and the estimated prediction power metrics were compared to identify the best approach. RESULTS:The exploratory study found high variability in hospital-related mortality across datasets, which supported the adoption of the mixed-effects models. Unstandardized Xgboost BME demonstrated marked improvements in prediction power over the Xgboost model at small sample size ranges, but performance differences decreased as dataset sizes increased. Generalized linear models (glms) and generalized linear mixed-effects models (glmers) followed similar results, with the Xgboost models also excelling at greater sample sizes. CONCLUSIONS:These findings suggest that integrating mixed effects into machine learning models can enhance their performance on datasets where the sample size is small.

Journal Article Accepted manuscript Reply to Rajakumar Get access Tim Dong, Tim Dong Division of Cardiac Surgery, Bristol Heart Institute, Translational Health Sciences, University of Bristol Search for other works by this author on: Oxford Academic PubMed Google Scholar Shubhra Sinha, Shubhra Sinha Division of Cardiac Surgery, Bristol Heart Institute, Translational Health Sciences, University of Bristol Search for other works by this author on: Oxford Academic PubMed Google Scholar Gianni D Angelini Gianni D Angelini Division of Cardiac Surgery, Bristol Heart Institute, Translational Health Sciences, University of Bristol Corresponding Author: Prof Gianni D Angelini, BHF Professor of Cardiac Surgery, Bristol Royal Infirmary, Bristol, BS2 8HW. E-mail: [email protected] Phone : 01173423165 Search for other works by this author on: Oxford Academic PubMed Google Scholar European Journal of Cardio-Thoracic Surgery, ezaf002, https://doi.org/10.1093/ejcts/ezaf002 Published: 17 January 2025 Article history Received: 17 December 2024 Accepted: 14 January 2025 Published: 17 January 2025

INTRODUCTION:Congenital heart disease (CHD) is the most common congenital anomaly, representing a significant global disease burden. Limitations exist in our understanding of aetiology, diagnostic methodology and screening, with metabolomics offering promise in addressing these. OBJECTIVE:To evaluate maternal metabolomics and lipidomics in prediction and risk factor identification for childhood CHD. METHODS:We performed an observational study in mothers of children with CHD following pregnancy, using untargeted plasma metabolomics and lipidomics by ultrahigh performance liquid chromatography-high resolution mass spectrometry (UHPLC-HRMS). 190 cases (157 mothers of children with structural CHD (sCHD); 33 mothers of children with genetic CHD (gCHD)) from the children OMACp cohort and 162 controls from the ALSPAC cohort were analysed. CHD diagnoses were stratified by severity and clinical classifications. Univariate, exploratory and supervised chemometric methods were used to identify metabolites and lipids distinguishing cases and controls, alongside predictive modelling. RESULTS:499 metabolites and lipids were annotated and used to build PLS-DA and SO-CovSel-LDA predictive models to accurately distinguish sCHD and control groups. The best performing model had an sCHD test set mean accuracy of 94.74% (sCHD test group sensitivity 93.33%; specificity 96.00%) utilising only 11 analytes. Similar test performances were seen for gCHD. Across best performing models, 37 analytes contributed to performance including amino acids, lipids, and nucleotides. CONCLUSIONS:Here, maternal metabolomic and lipidomic analysis has facilitated the development of sensitive risk prediction models classifying mothers of children with CHD. Metabolites and lipids identified offer promise for maternal risk factor profiling, and understanding of CHD pathogenesis in the future.

Objective. The learning curve of coronary artery bypass grafting with multiple arterial grafting and without the use of cardiopulmonary bypass (off‐pump) is perceived as an advanced subspeciality associated with increased surgical risk. We compared the trends and early clinical outcomes between trainees and consultants as the first operator in the United Kingdom. Methods. All patients who underwent elective or urgent isolated coronary artery bypass grafting from 1996 to 2019 were extracted from the National Adult Cardiac Surgery Audit database. Trends and early clinical outcomes between trainees and consultants as the first operator were compared in the whole cohort and after propensity score matching. Results. Of the total coronary artery bypass graft procedures, trainees performed 24.39% (n = 79759/327025). Trainees performed 27.10% (63934/235920) on‐pump without multiple arterial graft procedures compared to consultants. The consultants had a shorter cardiopulmonary bypass time (82.81 (SD: 35.36) vs 86.21 (SD: 30.07) minutes, p < 0.001) and aortic cross‐clamp time (48.05 (SD: 22.46) vs 50.66 (SD: 19.49) minutes, p < 0.001). However, consultants had a higher mortality (1.6% vs 1.0%, p < 0.001) and incidence of postoperative dialysis (2.1% vs 1.5%, p < 0.001). Trainees performed 16.78% (8089/48220) multiple arterial graft procedures, with no differences compared with consultants for in‐hospital mortality (1.0% vs 0.9%, p = 0.42), cerebral vascular accident (transient ischaemic attack (0.5% vs 0.5%) and permanent stroke (0.6% vs 0.4%), p = 0.33), return to theatre (4.2% vs 4.47%, p = 0.089), postoperative renal dialysis (1.4% vs 1.1%, p = 0.076), and deep sternal wound infection (0.6% vs 0.6%, p = 0.87). Trainees performed 17.17% (8661/41778) off‐pump cases. Consultants had a higher in‐hospital mortality (1.2% vs 0.9%, p = 0.045) with no differences in cerebral vascular accident (transient ischaemic attack (0.2% vs 0.3%) and permanent stroke (0.4% vs 0.4%), p = 0.27), return to theatre (3.8% vs 3.9%, p = 0.69), postoperative renal dialysis (2.0% vs 1.6%, p = 0.059), and deep sternal wound infection (1.0% vs 0.8%, p = 0.66). Conclusion. Trainees in the United Kingdom have adequate exposure to advanced coronary surgery without compromising patients’ safety.

Background:The Society of Thoracic Surgeons and European System for Cardiac Operative Risk Evaluation (EuroSCORE) II risk scores are the most commonly used risk prediction models for in-hospital mortality after adult cardiac surgery. However, they are prone to miscalibration over time and poor generalization across data sets; thus, their use remains controversial. Despite increased interest, a gap in understanding the effect of data set drift on the performance of machine learning (ML) over time remains a barrier to its wider use in clinical practice. Data set drift occurs when an ML system underperforms because of a mismatch between the data it was developed from and the data on which it is deployed. Objective:In this study, we analyzed the extent of performance drift using models built on a large UK cardiac surgery database. The objectives were to (1) rank and assess the extent of performance drift in cardiac surgery risk ML models over time and (2) investigate any potential influence of data set drift and variable importance drift on performance drift. Methods:We conducted a retrospective analysis of prospectively, routinely gathered data on adult patients undergoing cardiac surgery in the United Kingdom between 2012 and 2019. We temporally split the data 70:30 into a training and validation set and a holdout set. Five novel ML mortality prediction models were developed and assessed, along with EuroSCORE II, for relationships between and within variable importance drift, performance drift, and actual data set drift. Performance was assessed using a consensus metric. Results:A total of 227,087 adults underwent cardiac surgery during the study period, with a mortality rate of 2.76% (n=6258). There was strong evidence of a decrease in overall performance across all models (P<.0001). Extreme gradient boosting (clinical effectiveness metric [CEM] 0.728, 95% CI 0.728-0.729) and random forest (CEM 0.727, 95% CI 0.727-0.728) were the overall best-performing models, both temporally and nontemporally. EuroSCORE II performed the worst across all comparisons. Sharp changes in variable importance and data set drift from October to December 2017, from June to July 2018, and from December 2018 to February 2019 mirrored the effects of performance decrease across models. Conclusions:All models show a decrease in at least 3 of the 5 individual metrics. CEM and variable importance drift detection demonstrate the limitation of logistic regression methods used for cardiac surgery risk prediction and the effects of data set drift. Future work will be required to determine the interplay between ML models and whether ensemble models could improve on their respective performance advantages.

Introduction:Postoperative Atrial Fibrillation (POAF) is a common complication of cardiac surgery, associated with increased mortality, stroke risk, cardiac failure and prolonged hospital stay. Our study aimed to assess the patterns of release of systemic cytokines in patients with and without POAF.Methods:A post-hoc analysis of the Remote Ischemic Preconditioning (RIPC) trial, including 121 patients (93 males and 28 females, mean age of 68 years old) who underwent isolated coronary artery bypass grafting (CABG) and aortic valve replacement (AVR). Mixed-effect models were used to analyze patterns of release of cytokines in POAF and non-AF patients. A logistic regression model was used to assess the effect of peak cytokine concentration (6 h after the aortic cross-clamp release) alongside other clinical predictors on the development of POAF.Results:We found no significant difference in the patterns of release of IL-6 (p = 0.52), IL-10 (p = 0.39), IL-8 (p = 0.20) and TNF-α (p = 0.55) between POAF and non-AF patients. Also, we found no significant predictive value in peak concentrations of IL-6 (p = 0.2), IL-8 (p = >0.9), IL-10 (p = >0.9) and Tumour Necrosis Factor Alpha (TNF-α)(p = 0.6), however age and aortic cross-clamp time were significant predictors of POAF development across all models.Conclusions:Our study suggests no significant association exists between cytokine release patterns and the development of POAF. Age and Aortic Cross-clamp time were found to be significant predictors of POAF.

ObjectiveSurgical aortic valve replacement (SAVR) is traditionally the gold-standard treatment in patients with aortic valve disease. The advancement of transcatheter aortic valve replacement (TAVR) provides an alternative treatment to patients with high surgical risks and those who had previous cardiac surgery. We aim to evaluate the trend, early clinical outcomes, and the choice of prosthesis use in isolated SAVR in the United Kingdom.MethodsAll patients (n = 79,173) who underwent elective or urgent isolated surgical aortic valve replacement (SAVR) from 1996 to 2018 were extracted from the National Adult Cardiac Surgery Audit database. Patients who underwent additional procedures and emergency or salvage SAVR were excluded from the study. Trend and clinical outcomes were investigated in the whole cohort. Patients who had previous cardiac surgery, high-risk groups (EuroSCORE II >4%), and predicted/observed mortality were evaluated. Furthermore, the use of biological prostheses in five different age groups, that are <50, 50–59, 60–69, 70–79, and >80, was investigated. Clinical outcomes between the use of mechanical and biological aortic valve prostheses in patients <65 years old were analyzed.ResultsThe number of isolated SAVR increased across the study period with an average of 4,661 cases performed annually after 2010. The in-hospital/30-day mortality rate decreased from 5.28% (1996) to 1.06% (2018), despite an increasing trend in EuroSCORE II. The number of isolated SAVR performed in octogenarians increased from 596 to 2007 (the first year when TAVR was introduced in the UK) to 872 in 2015 and then progressively decreased to 681 in 2018. Biological prosthesis usage increased across all age groups, particularly in the 60–69 group, from 24.59% (1996) to 81.87% (2018). There were no differences in short-term outcomes in patients <65 years old who received biological or mechanical prostheses.ConclusionSurgical aortic valve replacement remains an effective treatment for patients with isolated aortic valve disease with a low in-hospital/30-day mortality rate. The number of patients with high-risk and octogenarians who underwent isolated SAVR and those requiring redo surgery has reduced since 2016, likely due to the advancement in TAVR. The use of biological aortic prostheses has increased significantly in recent years in all age groups.

Perioperative atrial fibrillation (AF) is associated with increased mortality, morbidity, and excess healthcare costs. The objective of our study was to assess if preoperative AF in patients undergoing coronary artery bypass grafting is a predictor of operative mortality, postoperative stroke, and need for postoperative dialysis by interrogating a large registry database. We included all isolated procedures performed between February 1996 and March 2019. We used a generalized linear mixed model to assess the effect of preoperative AF on mortality stroke and the need for postoperative dialysis after adjusting for the relevant confounders derived from EuroSCORE 2. Confounders considered included age, gender, neurological dysfunction, renal dysfunction, recent myocardial infarction, pulmonary disease, unstable angina, NYHA class, pulmonary hypertension, diabetes on insulin and peripheral vascular disease, and urgency of the operation. We treated the hospital and operating consultant as random effect variables. We also performed LV function subgroup analyses to assess the effect of preoperative AF on the outcomes of interest. The incidence of pre-existent AF in the cohort of patients we analyzed (N = 356,040 patients) was 3.5% (N = 12,664). In the unadjusted baseline characteristics, preoperative AF patients had more associated comorbidities. After adjustment, preoperative AF remained a significant predictor of increased mortality (odds ratio [OR]: 1.63, confidence interval [CI] 1.48-1.79, p < 0.001), stroke (OR: 1.33, CI 1.16-1.54, p = 0.001), and need for renal dialysis (OR:1.61, CI 1.46-1.78, p < 0.001). Preoperative AF was a significant predictor of adverse outcomes in patients with moderate and good LV function but not in patients with poor LV function (EF <30%). Our study suggests that preoperative AF is associated with an increased risk for perioperative mortality and stroke in patients undergoing coronary artery bypass grafting.

Abstract OBJECTIVES To perform a systematic comparison of in-hospital mortality risk prediction post-cardiac surgery, between the predominant scoring system—European System for Cardiac Operative Risk Evaluation (EuroSCORE) II, logistic regression (LR) retrained on the same variables and alternative machine learning techniques (ML)—random forest (RF), neural networks (NN), XGBoost and weighted support vector machine. METHODS Retrospective analyses of prospectively routinely collected data on adult patients undergoing cardiac surgery in the UK from January 2012 to March 2019. Data were temporally split 70:30 into training and validation subsets. Mortality prediction models were created using the 18 variables of EuroSCORE II. Comparisons of discrimination, calibration and clinical utility were then conducted. Changes in model performance, variable-importance over time and hospital/operation-based model performance were also reviewed. RESULTS Of the 227 087 adults who underwent cardiac surgery during the study period, there were 6258 deaths (2.76%). In the testing cohort, there was an improvement in discrimination [XGBoost (95% confidence interval (CI) area under the receiver operator curve (AUC), 0.834–0.834, F1 score, 0.276–0.280) and RF (95% CI AUC, 0.833–0.834, F1, 0.277–0.281)] compared with EuroSCORE II (95% CI AUC, 0.817–0.818, F1, 0.243–0.245). There was no significant improvement in calibration with ML and retrained-LR compared to EuroSCORE II. However, EuroSCORE II overestimated risk across all deciles of risk and over time. The calibration drift was lowest in NN, XGBoost and RF compared with EuroSCORE II. Decision curve analysis showed XGBoost and RF to have greater net benefit than EuroSCORE II. CONCLUSIONS ML techniques showed some statistical improvements over retrained-LR and EuroSCORE II. The clinical impact of this improvement is modest at present. However the incorporation of additional risk factors in future studies may improve upon these findings and warrants further study.

OBJECTIVE:Despite the superiority of mitral valve repair, surgical mitral valve replacement (SMVR) remains an important intervention for patients with valve stenosis, infective endocarditis and complex mitral valve degeneration. There has been an increasing popularity in the worldwide use of biological valves due to the avoidance of long-term anti-coagulation and recent advancements in transcatheter techniques. We aim to evaluate the trend, early clinical outcomes and the choice of prostheses use in isolated SMVR over a 23 years period in the United Kingdom. METHODS:All patients (n = 13,147) who underwent elective or urgent isolated SMVR from March 1996 to April 2019 were identified from the National Adult Cardiac Surgery Audit database. Trends in clinical outcomes, predicted/observed mortality of patients and the utilization of biological prostheses across 5 different age groups: <50, 50-59, 60-69, 70-79 and ≥80 years old were investigated. Early clinical outcomes associated with the use of mechanical and biological mitral valve prostheses in patients between the age of 60-70 years old were analysed. RESULTS:The number of isolated SMVR performed has remained stable with approximately 600 cases annually since 2010. The in-hospital/30-day mortality rate has decreased from 7.41% (1996) to 3.92% (2018), despite the EuroScore II increasing from 1.42% in 1996 to 2.43% in 2018. Biological prostheses usage increased across all age group, and particularly in the 60-69 and 70-79 group, from 17.86% and 53.85% in 1996 to 48.85% and 82.38% in 2018, respectively. The use of mechanical prostheses was reduced in patients between the age of 50-59 from 100% in 1996 to 80.65% in 2018. There were no differences in short term outcomes among patients aged 60-70 years who received either a biological or mechanical prostheses. CONCLUSION:There has been a significant reduction in surgical mitral valve replacement early in-hospital mortality, despite an observed increase in the risk profile of patients over 23 years. A shifting trend in valve replacement choices was observed with a rise in the use of biological prostheses, particularly within the 60-69 and 70-79 age group. Early in hospital outcomes for patients aged 60-70 were not determined by the implanted valve type.

Background: Although electronic health records (EHR) provide useful insights into disease patterns and patient treatment optimisation, their reliance on unstructured data presents a difficulty. Echocardiography reports, which provide extensive pathology information for cardiovascular patients, are particularly challenging to extract and analyse, because of their narrative structure. Although natural language processing (NLP) has been utilised successfully in a variety of medical fields, it is not commonly used in echocardiography analysis. Objectives: To develop an NLP-based approach for extracting and categorising data from echocardiography reports by accurately converting continuous (e.g., LVOT VTI, AV VTI and TR Vmax) and discrete (e.g., regurgitation severity) outcomes in a semi-structured narrative format into a structured and categorised format, allowing for future research or clinical use. Methods: 135,062 Trans-Thoracic Echocardiogram (TTE) reports were derived from 146967 baseline echocardiogram reports and split into three cohorts: Training and Validation (n = 1075), Test Dataset (n = 98) and Application Dataset (n = 133,889). The NLP system was developed and was iteratively refined using medical expert knowledge. The system was used to curate a moderate-fidelity database from extractions of 133,889 reports. A hold-out validation set of 98 reports was blindly annotated and extracted by two clinicians for comparison with the NLP extraction. Agreement, discrimination, accuracy and calibration of outcome measure extractions were evaluated. Results: Continuous outcomes including LVOT VTI, AV VTI and TR Vmax exhibited perfect inter-rater reliability using intra-class correlation scores (ICC = 1.00, p < 0.05) alongside high R2 values, demonstrating an ideal alignment between the NLP system and clinicians. A good level (ICC = 0.75–0.9, p < 0.05) of inter-rater reliability was observed for outcomes such as LVOT Diam, Lateral MAPSE, Peak E Velocity, Lateral E’ Velocity, PV Vmax, Sinuses of Valsalva and Ascending Aorta diameters. Furthermore, the accuracy rate for discrete outcome measures was 91.38% in the confusion matrix analysis, indicating effective performance. Conclusions: The NLP-based technique yielded good results when it came to extracting and categorising data from echocardiography reports. The system demonstrated a high degree of agreement and concordance with clinician extractions. This study contributes to the effective use of semi-structured data by providing a useful tool for converting semi-structured text to a structured echo report that can be used for data management. Additional validation and implementation in healthcare settings can improve data availability and support research and clinical decision-making.

Objectives The Society of Thoracic Surgeons (STS), and EuroSCORE II (ES II) risk scores, are the most commonly used risk prediction models for adult cardiac surgery post-operative in-hospital mortality. However, they are prone to miscalibration over time, and poor generalisation across datasets and their use remain controversial. It has been suggested that using Machine Learning (ML) techniques, a branch of Artificial intelligence (AI), may improve the accuracy of risk prediction. Despite increased interest, a gap in understanding the effect of dataset drift on the performance of ML over time remains a barrier to its wider use in clinical practice. Dataset drift occurs when a machine learning system underperforms because of a mismatch between the dataset it was developed and the data on which it is deployed. Here we analyse this potential concern in a large United Kingdom (UK) database. Methods A retrospective analyses of prospectively routinely gathered data on adult patients undergoing cardiac surgery in the UK between 2012-2019. We temporally split the data 70:30 into a training and validation subset. ES II and five ML mortality prediction models were assessed for relationships between and within variable importance drift, performance drift and actual dataset drift using temporal and non-temporal invariant consensus scoring, combining geometric average results of all metrics as the Clinical Effective Metric (CEM). Results A total of 227,087 adults underwent cardiac surgery during the study period with a mortality rate of 2.76%. There was a strong evidence of decrease in overall performance across all models (p < 0.0001). Xgboost (CEM 0.728 95CI: 0.728-0.729) and Random Forest (CEM 0.727 95CI 0.727-0.728) were the best overall performing models both temporally and non-temporally. ES II perfomed worst across all comparisons. Sharp changes in variable importance and dataset drift between 2017-10 to 2017-12, 2018-06 to 2018-07 and 2018-12 to 2019-02 mirrored effects of performance decrease across models. Conclusions Combining the metrics covering all four aspects of discrimination, calibration, clinical usefulness and overall accuracy into a single consensus metric improved the efficiency of cognitive decision-making. All models show a decrease in at least 3 of the 5 individual metrics. CEM and variable importance drift detection demonstrate the limitation of logistic regression methods used for cardiac surgery risk prediction and the effects of dataset drift. Future work will be required to determine the interplay between ML and whether ensemble models could take advantage of their respective performance advantages. Central message ML performance decreases over time due to dataset drift, but remains superior to ES II. Therefore regular assessment and modification of ML models may be preferable. Prospective message A gap in understanding the effect of dataset drift on the performance of ML models over time presents a major barrier to their clinical application. Xgboost and Random Forest have shown superior performance both temporally and non-temporally against ES II. However, a decrease in model performance of all models due to dataset drift suggests the need for regular drift monitoring. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement This work was supported by a grant from the BHF-Turing Institute and the NIHR Biomedical Research Centre at University Hospitals Bristol and Weston NHS Foundation Trust and the University of Bristol. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: The study was approved by the Health Research Authority (HRA) and Health and Care Research Wales (HCRW) in 23 of July 2019, IRAS project ID: 257758 and a waiver for patients' consent was obtained. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable. Yes All data used in this study are from the National Adult Cardiac Surgery Audit (NACSA) dataset. These data may be requested from Healthcare Quality Improvement Partnership (HQIP). <https://www.hqip.org.uk/national-programmes/accessing-ncapop-data/#.Ys6gN-zMLdp> * AUC : area under receiver operating characteristic curve CEM : Clinical Effective Metric ECE : Expected Calibration Error ES II : Euroscore II AI : Artificial intelligence ML : machine learning RF : random forest NN : Neural Network (Neuronetwork) SVM : support vector machine XGBoost : extreme gradient boosted trees : Ensemble using several models to derive a consensus prediction SHAP : (SHapley Additive exPlanations)