Contemporary Risk Scores Research Articles

Comparison of Contemporary Risk Scores in All Groups of Pulmonary Hypertension: A Pulmonary Vascular Research Institute GoDeep Meta-Registry Analysis

BackgroundPulmonary hypertension (PH) is a heterogeneous disease with poor prognosis. Accurate risk stratification is essential for guiding treatment decisions in pulmonary arterial hypertension (PAH). While various risk models were developed for PAH, their comparative prognostic potential requires further exploration. Additionally, the applicability of risk scores in PH groups beyond group 1 remains to be investigated. Research Question.Are risk scores originally developed for PAH predictive in PH group 1–4? Study Design and MethodsWe conducted a comprehensive analysis of outcomes among incident PH patients enrolled in the multicenter worldwide PVRI-GoDeep meta-registry. Analyses were performed across PH groups 1-4 and further subgroups to evaluate the predictive value of PAH-risk scores, including REVEAL Lite 2, REVEAL 2.0, ESC/ERS 2022, COMPERA 3-strata and COMPERA 4-strata. Results8565 patients were included in the study, of whom 3537 patients were assigned to group 1 PH while 1807, 1635, and 1586 patients were diagnosed with group 2, group 3, and group 4 PH. Pulmonary hemodynamics were impaired with median mPAP of 42 [33,52]mmHg and PVR of 7 [4,11]WU. All risk scores were prognostic in the entire PH population and in each of the PH groups 1–4. The REVEAL scores, when used as continuous prediction models, possessed the highest statistical prognostic power and granularity; the COMPERA 4-strata risk score provided sub-differentiation of the intermediate-risk group. Similar results were obtained when separately analyzing various subgroups (PH subgroups 1.1, 1.4.1, 1.4.4; 3.1, 3.2; group 2 with isolated post-capillary-PH versus combined pre-/post-capillary-PH; patients of all groups with concomitant cardiac comorbidities; severe [> 5 WU] versus non-severe PH). Interpretation.This comprehensive study with real-world data from 15 PH-centers showed that PAH-designed risk scores possess predictive power in a large PH cohort, whether considered as common group or calculated separately for each PH group (1-4) and various subgroups.

A discrete choice experiment to understand public preferences and priorities for risk-stratified bowel cancer screening programmes in the UK

ObjectivePublic acceptability of bowel cancer screening programmes must be maintained, including if risk stratification is introduced. We aimed to describe and quantify preferences for different attributes of risk-stratified screening programmes amongst the UK population, focussing on who to invite for bowel screening. MethodsWe conducted a discrete choice experiment (DCE) including the following attributes: risk factors used to estimate bowel cancer risk (age plus/minus sex, lifestyle factors and genetics); personalisation of risk feedback; risk stratification strategy plus resource implications; default screening in the case of no risk information; number of deaths prevented by screening; and number experiencing physical harm from screening. We used the results of conditional logit regression models to estimate the importance of each attribute, willingness to trade-off between the attributes, and preferences for different programmes using contemporary risk scores and models. Results1196 respondents completed the survey, generating 21,528 DCE observations. Deaths prevented was the most influential attribute on respondents' decision-making (contributing to 58.8% of the choice), followed by harms experienced (15.9%). For every three additional deaths prevented, respondents were willing to accept an additional screening harm per 100,000 people. Risk factors and risk stratification strategy contributed to just 11.1% and 3.6% of the choice, respectively. Although the influence on decision-making was small, respondents favoured more personalised feedback. ConclusionsBowel cancer screening programmes that improve cancer outcomes, particularly by preventing more deaths amongst those screened, are most preferred by the public. Optimising risk prediction models, developing public communication, and readying infrastructure should be prioritised for implementation.

Soluble ST2, BCN-Bio-HF calculator and MAGGIC-HF score in long-term risk prediction after an urgent visit for heart failure.

Soluble ST2 (sST2) is the expression of a pathogenic process related to adverse remodeling that ultimately leads to increased mortality in heart failure (HF). Risk score models provide a comprehensive approach for mortality prediction, beyond the use of biomarkers alone. The objective was to determine the additional value of sST2 and two well-validated contemporary risk scores, BCN-Bio-HF and MAGGIC-HF, in predicting mortality and readmission in the acute setting. This prospective study included 129 patients (mean age 75 ± 9 years; 52% males) after an urgent HF visit. Baseline sST2 levels were measured and the two risk scores were calculated. The primary endpoint was all-cause mortality, and the secondary endpoint was HF readmissions. The follow-up period was 3.6 ± 1.9 years. Patients who died (46%) had higher sST2 concentrations (80.5 vs. 42.7 ng/ml; p < 0.001). The BCN-Bio-HF calculator with sST2 demonstrated the best discriminative ability for mortality prediction (area under the ROC curve: 0.792; p < 0.001). In multivariate analysis for each risk score, the MAGGIC-HF score retained its predictive value only in the model without sST2 (3-year risk: HR = 1.036; 95% CI 1.019-1.054; p < 0.001). However, the BCN-Bio-HF score maintained its prognostic value with sST2 (HR = 1.032; 95%CI 1.020-1.044; p < 0.001), as well as without sST2 (HR = 1.035; 95% CI 1.021-1.049; p < 0.001). sST2 was not associated with readmission, and only the BCN-Bio-HF risk of HF hospitalization showed independent predictive value (OR = 1.040; 95% CI 1.005-1.076; p = 0.023). For predicting long-term mortality in HF in the emergency department, the BCN-Bio-HF calculator with sST2 demonstrated superior discrimination and allows estimation of the risk of HF hospitalizations.

A novel polygenic risk score improves prognostic prediction of heart failure with preserved ejection fraction in the Chinese Han population.

Mortality risk assessment in patients with heart failure (HF) with preserved ejection fraction (HFpEF) presents a major challenge. We sought to construct a polygenic risk score (PRS) to accurately predict the mortality risk of HFpEF. We first carried out a microarray analysis of 50 HFpEF patients who died and 50 matched controls who survived during 1-year follow-up for candidate gene selection. The HF-PRS was developed using the independent common (MAF > 0.05) genetic variants that showed significant associations with 1-year all-cause death (P < 0.05) in 1442 HFpEF patients. Internal cross-validation and subgroup analyses were performed to evaluate the discrimination ability of the HF-PRS. In 209 genes identified by microarray analysis, 69 independent variants (r < 0.1) were selected to develop the HF-PRS model. This model yielded the best discrimination capability for 1-year all-cause mortality with an area under the curve (AUC) of 0.852 (95% CI 0.827-0.877), which outperformed the clinical risk score consisting of 10 significant traditional risk factors for 1-year all-cause mortality (AUC 0.696, 95% CI 0.658-0.734, P = 4 × 10-11), with net reclassification improvement (NRI) of 0.741 (95% CI 0.605-0.877; P < 0.001) and integrated discrimination improvement (IDI) of 0.181 (95% CI 0.145-0.218; P < 0.001). Individuals in the medium and the highest tertile of the HF-PRS had nearly a five-fold (HR = 5.3, 95% CI 2.4-11.9; P = 5.6 × 10-5) and 30-fold (HR = 29.8, 95% CI 14.0-63.5; P = 1.4 × 10-18) increased risk of mortality compared to those in the lowest tertile, respectively. The discrimination ability of the HF-PRS was excellent in cross validation and throughout the subgroups regardless of comorbidities, gender, and patients with or without a history of heart failure. The HF-PRS comprising 69 genetic variants provided an improvement of prognostic power over the contemporary risk scores and NT-proBNP in HFpEF patients.

Predicting 5-Year Clinical Outcomes After Transcatheter or Surgical Aortic Valve Replacement (a Risk Score from the SURTAVI Trial)

Risk prediction scores for long-term outcomes after transcatheter aortic valve implantation (TAVI) or surgical aortic valve replacement (SAVR) are lacking. This study aimed to develop preprocedural risk scores for 5-year clinical outcomes after TAVI or SAVR. This analysis included 1,660 patients at an intermediate surgical risk with severe aortic stenosis randomly assigned to TAVI (n=864) or SAVR (n=796) from the SURTAVI (Surgical Replacement and Transcatheter Aortic Valve Implantation) trial. The primary end point was a composite of all-cause mortality or disabling stroke at 5 years. The secondary end point was a composite of cardiovascular mortality or hospitalizations for valve disease or worsening heart failure at 5 years. Preprocedural multivariable predictors of clinical outcomes were used to calculate a simple risk score for both procedures. At 5 years, the primary end point occurred in 31.3% of the patients with TAVI and 30.8% of the patients with SAVR. Preprocedural predictors differed between TAVI and SAVR. Baseline anticoagulant use was a common predictor for events in both procedures, whereas male sex and a left ventricular ejection fraction <60% were significant predictors for events in patients with TAVI and SAVR, respectively. A total of 4 simple scoring systems were created based on these multivariable predictors. The C-statistics of all models were modest but performed better than the contemporary risk scores. In conclusion, preprocedural predictors of events differ between TAVI and SAVR, necessitating separate risk models. Despite the modest predictive value of the SURTAVI risk scores, they appeared superior to other contemporary scores. Further research is needed to strengthen and validate our risk scores, possibly by including biomarker and echocardiographic parameters.

Contemporary risk models for infective endocarditis surgery: a narrative review.

Infective endocarditis is a complex heterogeneous condition involving the infection of the endocardium and heart valves, leading to severe complications, including death. Surgery is often indicated in patients with infective endocarditis but is associated with elevated risk compared with other forms of cardiac surgery. Risk models play an important role in many cardiac surgeries as they can help inform clinicians and patients regarding procedural risk, decision-making to proceed or not, and influence perioperative management; however, they remain under-utilized in the infective endocarditis settings. Another crucial role of such risk models is to assess predicted versus found mortality, thereby allowing an assessment of institutional performance in infective endocarditis surgery. Traditionally, general cardiac surgery risk models such as European System for Cardiac Operative Risk Evaluation (EuroSCORE), EuroSCORE II, and Society of Thoracic Surgeon's score have been applied to endocarditis surgery. However, there has been the development of many endocarditis surgery-specific scores over the last decade. This review aims to discuss clinical characteristics and applications of all contemporary risk scores in the setting of surgical treatment of infective endocarditis.

Risk scores’ performance and their impact on operative decision-making in left-sided endocarditis: a cohort study

The accuracy of contemporary risk scores in predicting perioperative mortality in infective endocarditis (IE) remains controversial. The aim is to evaluate the performance of existent mortality risk scores for cardiovascular surgery in IE and the impact on operability at high-risk thresholds. A single-center retrospective review of adult patients diagnosed with acute left-sided IE undergoing surgery from May 2014 to August 2019 (n = 142) was done. Individualized risk calculation was obtained according to the available mortality risk scores: EuroScore I and II, PALSUSE, Risk-E, Costa, De Feo-Cotrufo, AEPEI, STS-risk, STS-IE, APORTEI, and ICE-PCS scores. A cross-validation analysis was performed on the score with the best area under the curve (AUC). The 30-day survival was 96.5% (95%CI 91–98%). The score with worse area under the curve (AUC = 0.6) was the STS-IE score, while the higher was for the RISK-E score (AUC = 0.89). The AUC of the majority of risk scores suggested acceptable performance; however, statistically significant differences in expected versus observed mortalities were common. The cross-validation analysis showed that a large number of survivors (> 75%) would not have been operated if arbitrary high-risk threshold estimates had been used to deny surgery. The observed mortality in our cohort is significantly lower than is predicted by contemporary risk scores. Despite the reasonable numeric performance of the analyzed scores, their utility in judging the operability of a given patient remains questionable, as demonstrated in the cross-validation analysis. Future guidelines may advise that denial of surgery should only follow a highly experienced Endocarditis Team evaluation.

Incremental prognostic value of lung ultrasound on contemporary heart failure risk scores.

Introduction: Over the last decades, several scores have been developed to aid clinicians in assessing prognosis in patients with heart failure (HF) based on clinical data, medications and, ultimately, biomarkers. Lung ultrasound (LUS) has emerged as a promising prognostic tool for patients when assessed at discharge after a HF hospitalization. We hypothesized that contemporary HF risk scores can be improved upon by the inclusion of the number of B-lines detected by LUS at discharge to predict death, urgent visit, or HF readmission at 6- month follow-up. Methods: We evaluated the discrimination improvement of adding the number of B-lines to 4 contemporary HF risk scores (Get with the Guidelines -GWTG-, MAGGIC, Redin-SCORE, and BCN Bio-HF) by comparing the change in the area under the receiver operating curve (AUC), the net reclassification index (NRI), and the integrated discrimination improvement (IDI). The population of the study was constituted by the 123 patients enrolled in the LUS-HF trial, adjusting the analyses by the intervention. Results: The AUC of the GWTG score increased from 0.682 to 0.789 (p = 0.02), resulting in a NRI of 0.608 and an IDI of 0.136 (p < 0.05). Similar results were observed when adding the number of B-lines to the MAGGIC score, with an AUC that increased from 0.705 to 0.787 (p < 0.05). This increase translated into a NRI of 0.608 and an IDI of 0.038 (p < 0.05). Regarding Redin-SCORE at 1-month and 1-year, the AUC increased from 0.714 to 0.773 and from 0.681 to 0.757, although it did not reach statistical significance (p = 0.08 and p = 0.06 respectively). Both IDI and NRI were significantly improved (0.093 and 0.509 in the 1-month score, p < 0.05; 0.056 and 0.111 in the 1-year score, p < 0.05). Lastly, the AUC for the BCN Bio-HF score increased from 0.733 to 0.772, which was statistically non-significant, with a NRI value of 0.363 (p = 0.06) and an IDI of 0.092 (p < 0.05). Conclusion: Adding the results of LUS evaluated at discharge improved the predictive value of most of the contemporary HF risk scores. As it is a simple, fast, and non-invasive test it may be recommended to assess prognosis at discharge in HF patients.

Bleeding Risk Assessment in End-Stage Kidney Disease: Validation of Existing Risk Scores and Evaluation of a Machine Learning-Based Approach.

Patients with end-stage kidney disease (ESKD) on hemodialysis (HD) are at increased risk for bleeding. However, despite relevant clinical implications regarding dialysis modalities or anticoagulation, no bleeding risk assessment strategy has been established in this challenging population. Analyses on bleeding risk assessment models were performed in the population-based Vienna InVestigation of Atrial fibrillation and thromboemboLism in patients on hemoDialysIs (VIVALDI) study including 625 patients. In this cohort study, patients were prospectively followed for a median observation period of 3.5 years for the occurrence of major bleeding. First, performances of existing bleeding risk scores (i.e., HAS-BLED, HEMORR2HAGES, ATRIA, and four others) were evaluated in terms of discrimination and calibration. Second, four machine learning-based prediction models that included clinical, dialysis-specific, and laboratory parameters were developed and tested using Monte Carlo cross-validation. Of 625 patients (median age: 66 years, 37% women), 89 (14.2%) developed major bleeding, with a 1-year, 2-year, and 3-year cumulative incidence of 6.1% (95% confidence interval [CI]: 4.2-8.0), 10.3% (95% CI: 8.0-12.8), and 13.5% (95% CI: 10.8-16.2), respectively. C-statistics of the seven contemporary bleeding risk scores ranged between 0.54 and 0.59 indicating poor discriminatory performance. The HAS-BLED score showed the highest C-statistic of 0.59 (95% CI: 0.53-0.66). Similarly, all four machine learning-based predictions models performed poorly in internal validation (C-statistics ranging from 0.49 to 0.55). Existing bleeding risk scores and a machine learning approach including common clinical parameters fail to assist in bleeding risk prediction of patients on HD. Therefore, new approaches, including novel biomarkers, to improve bleeding risk prediction in patients on HD are needed.

Validation of Japanese Bleeding Risk Criteria in Patients After Percutaneous Coronary Intervention and Comparison With Contemporary Bleeding Risk Criteria.

Background: The utility of the Japanese version of high bleeding risk (J-HBR) criteria compared with contemporary bleeding risk criteria, including Academic Research Consortium for High Bleeding Risk criteria, has not been fully investigated.Methods and Results: This study included patients who underwent percutaneous coronary intervention between 2010 and 2019. The J-HBR score was calculated by assigning 1 point for each major criterion and 0.5 points for each minor criterion in the J-HBR criteria. Among 1,643 patients, 1,143 (69.6%) met the J-HBR criteria. Accumulated major bleeding event rates at 1 year were higher among those who met the J-HBR criteria (4.8% vs. 0.6%; P<0.001). J-HBR criteria had higher sensitivity (94.8%) and lower specificity (31.4%) than contemporary bleeding risk criteria in predicting major bleeding. Bleeding events increased with increasing J-HBR score. The C statistic for the J-HBR score for predicting major bleeding at 1 year was 0.75 (95% confidence interval 0.69–0.81), and is comparable to that of other risk scores. In multivariate analysis, of the factors included in J-HBR criteria, chronic kidney disease, heart failure, and active malignancy were associated with major bleeding.Conclusions: J-HBR criteria identified patients at high bleeding risk with high sensitivity and low specificity. Bleeding risk was closely related to J-HBR score and its individual components. The discriminative ability of the J-HBR score was comparable to that of contemporary bleeding risk scores.

Mortality Risk Prediction Dynamics After Heart Failure Treatment Optimization: Repeat Risk Assessment Using Online Risk Calculators.

ObjectivesHeart failure (HF) management has significantly improved over the past two decades, leading to better survival. This study aimed to assess changes in predicted mortality risk after 12 months of management in a multidisciplinary HF clinic.Materials and MethodsOut of 1,032 consecutive HF outpatients admitted from March-2012 to November-2018, 357 completed the 12-months follow-up and had N-terminal pro-B-type natriuretic peptide (NTproBNP), high sensitivity troponin T (hs-TnT), and interleukin-1 receptor-like-1 (known as ST2) measurements available both at baseline and follow-up. Three contemporary risk scores were used: MAGGIC-HF, Seattle HF Model (SHFM), and the Barcelona Bio-HF (BCN Bio-HF) calculator, which incorporates the three above mentioned biomarkers. The predicted risk of all-cause death at 1 and 3 years was calculated at baseline and re-evaluated after 12 months.ResultsA significant decline in predicted 1-and 3-year mortality risk was observed at 12 months: MAGGIC ~16%, SHFM ~22% and BCN Bio-HF ~15%. In the HF with reduced ejection fraction (HFrEF) subgroup guideline-directed medical therapy led to a complete normalization of left ventricular ejection fraction (≥50%) in almost a third of the patients and to a partial normalization (41–49%) in 30% of them. Repeated risk assessment after 12 months with SHFM and BCN Bio-HF provided adequate discrimination for all-cause 3-year mortality (C-Index: MAGGIC-HF 0.762, SHFM 0.781 and BCN Bio-HF 0.791).ConclusionMortality risk declines in patients with HF managed for 12 months in a multidisciplinary HF clinic. Repeating the mortality risk assessment after optimizing the HF treatment is recommended, particularly in the HFrEF subgroup.

Abstract 9707: Validation of Japanese High Bleeding Risk Criteria in Patients Undergoing Percutaneous Coronary Intervention and Comparisons With Contemporary Bleeding Risk Scores

Introduction: The Japanese version of high bleeding risk (J-HBR) criteria have been established by modifying the Academic Research Consortium criteria for high bleeding risk (ARC-HBR), since East-Asian people have several distinct risk factors that were not included in ARC-HBR. We aimed to validate the J-HBR criteria for bleeding outcomes and compare with contemporary risk scores. Methods: Consecutive 1643 patients with coronary artery disease who underwent percutaneous coronary intervention were enrolled. We followed-up bleeding events of Bleeding Academic Research Consortium (BARC) 3 or 5. Results: Among 1643 patients, 1143 (69.6%) who met J-HBR criteria had a higher accumulated bleeding rate than those who did not (P &lt; 0.001, left figure). BARC 3 or 5 bleeding event rates at 1 year were 4.8% in patients who met J-HBR criteria and 0.6% in those who did not (P &lt; 0.01). The J-HBR criteria had higher sensitivity than ARC-HBR, PRECISE-DAPT and PARIS bleeding risk score (94.8%, 81.0%, 75.9% and 87.9%, respectively) and lower specificity than those (31.4%, 52.6%, 54.1% and 51.0%, respectively). J-HBR score was calculated by adding 1 point for major criterion and 0.5 point for minor criterion in J-HBR criteria. The bleeding event rate increased with advancing J-HBR score (right figure). The c-statistics of J-HBR score for predicting BARC 3 or 5 bleeding at 1 year was 0.74 (95% CI, 0.67 - 0.80) that was comparable to ARC-HBR score, PRECISE-DAPT score and PARIS bleeding risk score (0.73, 0.73 and 0.72, respectively). In the multivariate analysis, chronic kidney disease, anemia and heart failure were associated with BARC 3 or 5 bleeding (HR 6.39, P &lt; 0.01; HR 3.38, P &lt; 0.01; and HR 2.95, P &lt; 0.01; respectively) among factors included in J-HBR criteria. Conclusions: The J-HBR criteria successfully identified patients at high bleeding risk, with high sensitivity and low specificity. The bleeding risk was closely related to J-HBR score and its individual component.

Comparison of risk scores for predicting outcomes after isolated tricuspid valve surgery.

Risk models play important roles in stratification and decision-making towards cardiac surgery. Isolated tricuspid valve surgery is a high risk but increasingly performed the operation, however, the performance of risk models has not been externally evaluated in these patients. We compared the prognostic utility of contemporary risk scores for isolated tricuspid valve surgery. Consecutive patients undergoing isolated tricuspid valve surgery at Cleveland Clinic during 2004-2018 were evaluated in this cohort study. EuroSCORE II, Society of Thoracic Surgeon's tricuspid (STS-TVS) score, and the Model for End-stage Liver Disease (MELD) score were retrospectively calculated, and their performance for predicting operative mortality, postoperative complications, and mortality during follow-up was assessed. Amongst 207 patients studied, the mean age was 54.1 ± 17.9 years, 116 (56.0%) were female, 92 (44.4%) had secondary tricuspid regurgitation, and 151 (72.9%) had a surgical repair. Mean EuroSCORE II, STS-TVS, and MELD scores were 6.3 ± 6.6%, 5.5 ± 6.2%, and 9.8 ± 4.7, respectively. C-statistics (95% confidence intervals) for operative mortality were 0.83 (0.74-0.93) for EuroSCORE II, 0.60 (0.45-0.75) for STS-TVS score, and 0.74 (0.58-0.89) for MELD score, while observed/expected ratios were 0.78 and 0.89 for the first two scores. All three scores were associated with mortality during follow-up and discriminated most postoperative complications. EuroSCORE II was superior to STS-tricuspid score for isolated TVS risk assessment. Although surgical risk scores traditionally underestimated operative mortality after isolated tricuspid valve surgery, they did not in our cohort, reflecting the excellent surgical results. The simple MELD score performed similarly to the EuroSCORE II, especially for discriminating morbidities.

Head-to-head comparison of contemporary heart failure risk scores.

Several heart failure (HF) web-based risk scores are currently used in clinical practice. Currently, we lack head-to-head comparison of the accuracy of risk scores. This study aimed to assess correlation and mortality prediction performance of Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC-HF) risk score, which includes clinical variables + medications; Seattle Heart Failure Model (SHFM), which includes clinical variables + treatments + analytes; PARADIGM Risk of Events and Death in the Contemporary Treatment of Heart Failure (PREDICT-HF) and Barcelona Bio-Heart Failure (BCN-Bio-HF) risk calculator, which also include biomarkers, like N-terminal pro B-type natriuretic peptide (NT-proBNP). A total of 1166 consecutive patients with HF from different aetiologies that had NT-proBNP measurement at first visit were included. Discrimination for all-cause mortality was compared by Harrell's C-statistic from 1 to 5 years, when possible. Calibration was assessed by calibration plots and Hosmer-Lemeshow test and global performance by Nagelkerke's R2 . Correlation between scores was assessed by Spearman rank test. Correlation between the scores was relatively poor (rho value from 0.66 to 0.79). Discrimination analyses showed better results for 1-year mortality than for longer follow-up (SHFM 0.817, MAGGIC-HF 0.801, PREDICT-HF 0.799, BCN-Bio-HF 0.830). MAGGIC-HF showed the best calibration, BCN-Bio-HF overestimated risk while SHFM and PREDICT-HF underestimated it. BCN-Bio-HF provided the best discrimination and overall performance at every time-point. None of the contemporary risk scores examined showed a clear superiority over the rest. BCN-Bio-HF calculator provided the best discrimination and overall performance with overestimation of risk. MAGGIC-HF showed the best calibration, and SHFM and PREDICT-HF tended to underestimate risk. Regular updating and recalibration of online web calculators seems necessary to improve their accuracy as HF management evolves at unprecedented pace.

Abstract P122: Leveraging Machine Learning Models For Dynamic Prediction Of Bleeding And Ischemic Risk After Drug-eluting Stent Implantation

Background: Contemporary risk scores for ischemic or bleeding endpoint prediction after drug-eluting stent (DES) implantation have limited predictive accuracy and fixed prediction windows. Objective: This study aimed to dynamically predict the ischemic and bleeding risks in different follow-up windows for patients with DES leveraging cutting-edge machine learning models. Methods: Using the Cerner Health Facts, we identified 98,236 adult patients who received DES with more than three months of follow-up after discharge. Three traditional machine learning models (LR, RF, and LGBM), three RNN-based deep learning models (GRU, LSTM, RETAIN), and one transformer-based model (Med-BERT) were applied for the risk stratification. The total follow-up window covered two years after discharge. ML models were tested in three types of prediction windows with varying lengths (i.e., one year, six months, and three months). Model performance was assessed according to a range of learning metrics including area under the receiver operating characteristic curve (AUC) and precision-recall curve. Results: For bleeding event prediction, the best AUC for the three-month window was 77.02% (GRU), for the six-month window was 81.11% (BiGRU + MedBert), and for the one-year window is 71.86% (LGBM). For ischemic event prediction, the best AUC for the three-month window was 71.09% (GRU), for the six-month window was 68.97% (LGBM), for the one-year window was 69.41% (LSTM). The transformer-based model and RNN-based models generally performed better than the traditional ML in both bleeding and ischemic risk prediction. Conclusions: Our machine learning model-based approaches for dynamically predicting the ischemic and bleeding risks after stenting is feasible and effective. The best-performing models show good discriminative capabilities in different prediction scenarios and have the potential to guide the clinical decision-making of the optimum DAPT duration in a personalized way.

Mortality risk prediction in elderly patients with cardiogenic shock: results from the CardShock study.

AimsThis study aimed to assess the utility of contemporary clinical risk scores and explore the ability of two biomarkers [growth differentiation factor‐15 (GDF‐15) and soluble ST2 (sST2)] to improve risk prediction in elderly patients with cardiogenic shock.Methods and resultsPatients (n = 219) from the multicentre CardShock study were grouped according to age (elderly ≥75 years and younger). Characteristics, management, and outcome between the groups were compared. The ability of the CardShock risk score and the IABP‐SHOCK II score to predict in‐hospital mortality and the additional value of GDF‐15 and sST2 to improve risk prediction in the elderly was evaluated. The elderly constituted 26% of the patients (n = 56), with a higher proportion of women (41% vs. 21%, P < 0.05) and more co‐morbidities compared with the younger. The primary aetiology of shock in the elderly was acute coronary syndrome (84%), with high rates of percutaneous coronary intervention (87%). Compared with the younger, the elderly had higher in‐hospital mortality (46% vs. 33%; P = 0.08), but 1 year post‐discharge survival was excellent in both age groups (90% in the elderly vs. 88% in the younger). In the elderly, the risk prediction models demonstrated an area under the curve of 0.75 for the CardShock risk score and 0.71 for the IABP‐SHOCK II score. Incorporating GDF‐15 and sST2 improved discrimination for both risk scores with areas under the curve ranging from 0.78 to 0.84.ConclusionsElderly patients with cardiogenic shock have higher in‐hospital mortality compared with the younger, but post‐discharge outcomes are similar. Contemporary risk scores proved useful for early mortality risk prediction also in the elderly, and risk stratification could be further improved with biomarkers such as GDF‐15 or sST2.

Scoring systems for the triage and assessment of short-term cardiovascular risk in patients with acute chest pain.

Acute chest pain is a common emergency department (ED) chief complaint. Evaluating patients for acute coronary syndrome is challenging because missing the diagnosis carries substantial morbidity, mortality, and medicolegal consequences. However, over-testing is associated with increased cost, overcrowding, and possible iatrogenic harm. Over the past two decades, multiple risk scoring systems have been developed to help emergency providers evaluate patients with acute chest pain. The ideal risk score balances safety by achieving high sensitivity and negative predictive value for major adverse cardiovascular events while also being effective in identifying a large proportion of patients for early discharge from the ED. This review examines contemporary risk scores used to risk stratify patients with acute chest pain.

Abstract 15341: Multi-center Validation of a Novel Echocardiography Artificial Intelligence System to Predict Post-operative Right Ventricular Failure in LVAD Patients

Introduction: Post-operative right ventricular failure (RV failure) is the single largest contributor to short-term mortality in patients with left ventricular assist devices (LVAD); yet predicting which patient is at risk of developing this complication in the pre-operative setting has remained beyond the abilities of experts in the field. We hypothesized that deep artificial intelligence (AI) driven characterization of subtle pre-operative myocardial motion features, could predict post-operative RV failure in LVAD patients. Methods: We developed a novel echocardiography AI system using an improved dense trajectory algorithm that tracks motion vectors in an unsupervised fashion, and a 3-dimensional convolutional neural network that tracks spatiotemporal features from videos. We used pre-operative ECHO videos from a 536 patient multicenter echocardiographic and clinical dataset, and via a standard 10-fold cross validation, trained and validated the AI system to predict severe or higher grades of post-operative RV failure at the time of index hospitalization. Finally, we independently benchmarked our AI system against clinicians equipped with contemporary clinical risk scores to predict post-operative RV failure (Penn and CRITT score) and manually derived echocardiographic metrics of RV function. Results: 173 (32%) patients were adjudicated to have severe post-operative RV failure. The area under the receiver operator characteristic curve (AUC ROC) for the AI system was 0.86 (95% CI 0.824-0.896). The performance of our AI system exceeded that of both the best performing clinical risk score and manually derived echocardiographic metric by a significant margin (ΔAUC +0.30, 95% CI 0.27 - 0.33 and ΔAUC +0.24, 95% CI 0.21 - 0.27; both p &amp;lt 0.0001). Conclusions: A novel ECHO AI system trained and validated on a multicenter dataset outperformed clinical experts equipped with both contemporary risk scores and manually calculated echocardiographic metrics.

Abstract 12871: Comparison of Risk Scores at Predicting Mortality and Morbidities After Isolated Tricuspid Valve Surgery

Background: Risk models play important roles in stratification and decision-making towards cardiac surgery. Isolated tricuspid valve surgery (TVR) is a high risk but increasingly performed operation, however the performance of surgical risk models, including the recently developed Society of Thoracic Surgeon’s (STS) TVR score, have not been externally evaluated in these patients. We compared the prognostic utility of contemporary risk scores for isolated TVR. Methods: Consecutive patients undergoing isolated TVR at Cleveland Clinic during 2004-2018 were evaluated. EuroSCORE II and STS-TVR score were retrospectively calculated, and their performance for predicting 30-day mortality, post-operative complications and mortality during follow-up were assessed. Results: Amongst 207 patients studied, mean age was 54.1±17.9 years, 116 (56.0%) were female, 92 (44.4%) had secondary tricuspid regurgitation, and 151 (72.9%) had surgical repair. Mean EuroSCORE II and STS-TVR scores were 6.3±6.6% and 5.5±6.2% respectively. Operative mortality occurred in 10 (4.9%). C-statistics (95% confidence intervals) for 30-day mortality were 0.83 (0.74-0.93) for EuroSCORE II and 0.60 (0.45-0.75) for STS-TVR score (Table), and observed/expected ratios were 0.78 and 0.89 respectively. EuroSCORE II also had higher c-statistics for composite and most post-operative complications. In univariable Cox regression for mortality during follow-up, hazards ratios (95% confidence intervals) were 1.08 (1.05-1.11) for EuroSCORE II and 1.06 (1.02-1.11) for STS-TVR score. Conclusion: EuroSCORE II was superior to STS-TVR score at predicting adverse outcomes after isolated TVR. Although risk scores traditionally under-estimated 30-day mortality after isolated TVR, our excellent surgical results meant that calibration of both scores were satisfactory for our cohort, and provides a benchmark for future studies of isolated surgical and transcatheter tricuspid valve procedures.

Direct comparison of predictive performance of PRECISE-DAPT versus PARIS versus CREDO-Kyoto: asubanalysis of the ReCre8 trial.

BackgroundMultiple scores have been proposed to guide risk stratification after percutaneous coronary intervention. This study assessed the performance of the PRECISE-DAPT, PARIS and CREDO-Kyoto risk scores to predict post-discharge ischaemic or bleeding events.MethodsA total of 1491 patients treated with latest-generation drug-eluting stent implantation were evaluated. Risk scores for post-discharge ischaemic or bleeding events were calculated and directly compared. Prognostic performance of both risk scores was assessed with calibration, Harrell’s c‑statistics net reclassification index and decision curve analyses.ResultsPost-discharge ischaemic events occurred in 56 patients (3.8%) and post-discharge bleeding events in 34 patients (2.3%) within the first year after the invasive procedure. C‑statistics for the PARIS ischaemic risk score was marginal (0.59, 95% confidence interval (CI) 0.51–0.68), whereas the CREDO-Kyoto ischaemic risk score was moderate (0.68, 95% CI 0.60–0.75). With regard to post-discharge bleeding events, CREDO-Kyoto displayed moderate discrimination (c-statistic 0.67, 95% CI 0.56–0.77), whereas PRECISE-DAPT (0.59, 95% CI 0.48–0.69) and PARIS (0.55, 95% CI 0.44–0.65) had a marginal discriminative capacity. Net reclassification index and decision curve analysis favoured CREDO-Kyoto-derived bleeding risk assessment.ConclusionIn this contemporary all-comer population, PARIS and PRECISE-DAPT risk scores were not resilient to independent testing for post-discharge bleeding events. CREDO-Kyoto-derived risk stratification was associated with a moderate predictive capability for post-discharge ischaemic or bleeding events. Future studies are warranted to improve risk stratification with more focus on robustness and rigorous testing.Electronic supplementary materialThe online version of this article (10.1007/s12471-020-01486-y) contains supplementary material, which is available to authorized users.