Traditional Risk Scores Research Articles

Assessing the cardiovascular risk in patients with systemic lupus erythematosus: QRISK and GAPSS scores head-to-head

ObjectivesWe aimed to apply and compare the QRISK3 and the adjusted Global AntiPhospholipid Syndrome (APS) Score (aGAPSS) in a cohort of systemic lupus erythematosus (SLE) patients, with and without a concomitant diagnosis of APS, in order to assess their augmented risk of developing cardiovascular diseases (CVDs). MethodsPatients (25–85 yo) with a diagnosis of SLE and/or of Secondary APS (SAPS) were included. QRISK3 was calculated using the official online calculator; aGAPSS using the validated point-values based on aPL-profile and independent risk factors. ResultsThe cohort included 142 SLE patients: 34 SAPS (23.9%) and 108 SLE patients without APS (76.1%).When considering all the cohort, patients with cerebrovascular/coronary events showed higher values of aGAPSS (10.1 ± 6.2 vs. 5.8 ± 6.1; p = 0.007), but not of the QRISK3. Furthermore, a significant association was observed between the occurrence of these events and high-risk aGAPSS: p = 0.03 for aGAPSS≥8, p = 0.01 for aGAPSS ≥9, p = 0.008 for aGAPSS ≥10. aGAPSS strongly correlated with the occurrence of any thrombotic event, both at the uni- and multivariate analysis (p = 0.012 and p = 0.009). Male gender also resulted to positively correlate with the occurrence of any thrombotic event at both uni- and multivariate analysis (p = 0.017 and p = 0.03). Focusing on aPL-profile, regardless the diagnosis, we found a statistical significance only for aGAPSS (aPL+ =9.6 ± 6.3 vs. aPL- = 4.1 ± 5.1; p < 0.001). ConclusionsDespite QRISK3 being more accurate than traditional risk score in predicting CVD risk in SLE patients, aGAPSS appears to be the most valuable tool for this purpose.

FIPS Score for Prediction of Survival After TIPS Placement: External Validation and Comparison With Traditional Risk Scores in a Cohort of Chinese Patients With Cirrhosis.

BACKGROUND. Various prognostic scores for patients with chronic liver disease have been applied for predicting survival after TIPS placement. In 2021, the Freiburg index of post-TIPS survival (FIPS) score was developed specifically for predicting survival after TIPS placement. The score has exhibited variable performance in initial investigations conducted in German and U.S. cohorts. OBJECTIVE. The purpose of this study was to compare the utility of the FIPS score and traditional scoring systems for predicting post-TIPS survival in a cohort of Chinese patients with cirrhosis. METHODS. This retrospective validation study compared four prognostic scores (model for end-stage liver disease [MELD], sodium MELD [MELD-Na], Chronic Liver Failure Consortium acute decompensation [CLIF-C AD], and FIPS) in 383 patients (mean age, 54.9 ± 11.7 years; 249 men, 134 women) with cirrhosis who underwent TIPS placement (341 for variceal bleeding, 42 for refractory ascites) at Wuhan Union Hospital between January 2016 and August 2021. Model performance was assessed in terms of discrimination (using concordance index) and calibration (using Brier score and observed-to-predicted ratios) for 6-, 12-, and 24-month post-TIPS survival. Discrimination was further stratified by TIPS indication. Risk stratification was performed using previously proposed cutoffs for each score. RESULTS. During postprocedural follow-up, 72 (18.8%) patients died. Discriminative performance for 6-month survival was highest for FIPS score (concordance index, 0.784), followed by CLIF-C AD (0.743), MELD-Na (0.699), and MELD (0.694). FIPS score also showed the highest calibration in terms of lower Brier scores and observed-to-predicted ratios closer to 1 and showed the strongest prognostic performance for 12- and 24-month survival and in subgroups of patients who underwent TIPS placement for either variceal bleeding or refractory ascites (except for similar performance of FIPS and CLIF-C AD in the refractory ascites subgroup). When prior cutoffs were applied, further application of FIPS score was significantly associated with survival among patients classified as low risk by the other scores. CONCLUSION. FIPS score outperformed traditional risk scores in predicting post-TIPS survival in patients with cirrhosis. CLINICAL IMPACT. The findings support utility of FIPS score in differentiating patients who are optimal candidates for TIPS placement versus those at high risk who may instead warrant close monitoring and early liver transplant.

Evaluation of the Incremental Value of a Coronary Artery Calcium Score Beyond Traditional Cardiovascular Risk Assessment

Coronary artery calcium scores (CACS) are used to help assess patients' cardiovascular status and risk. However, their best use in risk assessment beyond traditional cardiovascular factors in primary prevention is uncertain. To find, assess, and synthesize all cohort studies that assessed the incremental gain from the addition of a CACS to a standard cardiovascular disease (CVD) risk calculator (or CVD risk factors for a standard calculator), that is, comparing CVD risk score plus CACS with CVD risk score alone. Eligible studies needed to be cohort studies in primary prevention populations that used 1 of the CVD risk calculators recommended by national guidelines (Framingham Risk Score, QRISK, pooled cohort equation, NZ PREDICT, NORRISK, or SCORE) and assessed and reported incremental discrimination with CACS for estimating the risk of a future cardiovascular event. From 2772 records screened, 6 eligible cohort studies were identified (with 1043 CVD events in 17 961 unique participants) from the US (n = 3), the Netherlands (n = 1), Germany (n = 1), and South Korea (n = 1). Studies varied in size from 470 to 5185 participants (range of mean [SD] ages, 50 [10] to 75.1 [7.3] years; 38.4%-59.4% were women). The C statistic for the CVD risk models without CACS ranged from 0.693 (95% CI, 0.661-0.726) to 0.80. The pooled gain in C statistic from adding CACS was 0.036 (95% CI, 0.020-0.052). Among participants classified as being at low risk by the risk score and reclassified as at intermediate or high risk by CACS, 85.5% (65 of 76) to 96.4% (349 of 362) did not have a CVD event during follow-up (range, 5.1-10.0 years). Among participants classified as being at high risk by the risk score and reclassified as being at low risk by CACS, 91.4% (202 of 221) to 99.2% (502 of 506) did not have a CVD event during follow-up. This systematic review and meta-analysis found that the CACS appears to add some further discrimination to the traditional CVD risk assessment equations used in these studies, which appears to be relatively consistent across studies. However, the modest gain may often be outweighed by costs, rates of incidental findings, and radiation risks. Although the CACS may have a role for refining risk assessment in selected patients, which patients would benefit remains unclear. At present, no evidence suggests that adding CACS to traditional risk scores provides clinical benefit.

Oral anticoagulation in patients with atrial fibrillation and end-stage kidney disease

Atrial fibrillation is highly prevalent in patients with end -stage kidney disease. For patients with atrial fibrillation but normal kidney function, the net clinical benefit of oral anticoagulation with vitamin K antagonists and non -vitamin K oral anticoagulants in preventing stroke against the hemorrhagic risk is well -known. However, dialysis patients present unique risk factors that increase ischemic stroke and bleeding events that are not included in the traditional risk scores. These points, in addition to the pharmacokinetic issues related to the use of oral anticoagulants in end -stage kidney disease, raise the question of the safety of these drugs, making it difficult to predict whether the benefit outweighs the risk. Data on this topic in chronic kidney disease stages 4, 5, and 5D patients is sparse and the clinical guidelines are also inconsistent in which is the best approach for these patients. This review gives an overview of the state of the art on this topic, specifically the recommendations of the newest guidelines, then discusses the various risk scores and their pitfalls, and finally summarizes the current knowledge on the possible benefits of Non -vitamin K oral anticoagulants versus Vitamin K antagonists in dialysis patients.

IPOscore: An interactive web-based platform for postoperative surgical complications analysis and prediction in the oncology domain

Background: The performance of traditional risk score systems to predict (post)-operative outcomes is limited. This weakness reduces confidence in its use to support clinical risk mitigation decisions. However, the rapid growth of health data in the last years offers principles to deal with some of these limitations. In this regard, the data allows the extraction of relevant information for both patients stratification and the rigorous identification of associated risk factors. The patients can then be targeted to specific preoperative optimization programs, thus contributing to the reduction of associated morbidity and mortality. Objectives: The main goal of this work is, therefore, to provide a clinical decision support system (CDSS) based on data-driven modeling methods for surgical risk prediction specific for cancer patients in Portugal. Results: The result is IPOscore, a single web-based platform aimed at being an innovative approach to assist clinical decision-making in the surgical oncology domain. This system includes a database to store/manage the clinical data collected in a structured format, data visualization and analysis tools, and predictive machine learning models to predict postoperative outcomes in cancer patients. IPOscore also includes a pattern mining module based on biclustering to assess the discriminative power of a pattern towards postsurgical outcomes. Additionally, a mobile application is provided to this end. Conclusions: The IPOscore platform is a valuable tool for surgical oncologists not only for clinical data management but also as a preventative and predictive healthcare system. Currently, this clinical support tool is being tested at the Portuguese Institute of Oncology (IPO-Porto), and can be accessed online at https://iposcore.org.

Abstract ES7-3:/AI

Abstract The tools of AI offer promise to uncover and amplify data, previously hidden in images of the human body, and expand the field of radiomics to more accurately predict breast cancer risk across diverse patient populations. Although deep learning products trained to predict mammographic breast density and risk of breast cancer are now available, their performance in routine clinical settings is largely unknown. Since the creation of the Gail model in 1989, risk models have supported risk-adjusted screening and prevention, and their continued evolution has been a central pillar of breast cancer research. Mammographic breast density is now being incorporated into well-established clinical risk models, which are used to determine eligibility for supplemental imaging and other services for patients at increased risk. However, mammographic breast density assessment is subjective and varies widely between and within radiologists. For example, Sprague et al demonstrated considerable variability of qualitative Breast Imaging Reporting and Data System (BI-RADS) density assessments with a range of 6% to 85% of mammograms being assessed as heterogeneously or extremely dense across 83 radiologists. To limit variability and subjectivity, various automated breast density assessment methods were commercially developed, but clinical evaluation has yielded mixed results. Density is only one limited feature of any woman’s mammogram. Deep learning models can operate over the full resolution of mammogram images to assess a patient’s future breast cancer risk. Rather than manually identifying discriminative image patterns, machine learning models can discover these patterns directly from the data. Specifically, models are trained with full resolution mammograms and the outcome of interest, namely whether the patient developed breast cancer within five years from the date of the examination. Our recent work demonstrates that application of novel artificial intelligence applications to imaging data can significantly improve breast density assessment and breast cancer risk prediction. In addition, unlike traditional models, our DL models perform equally well across varied races, ages, and family histories and we have built a clinical platform which is currently in use to support implementation of our density and risk models into routine clinical care. From this clinical platform, we found that use of a deep learning model more accurately predicts mammographic breast density, reducing human variation and “overcalling” of breast density, and can help healthcare systems more appropriately utilize limited supplemental screening resources as well as provide patients with more accurate information regarding their breast density. We have also found a DL breast cancer risk model, generated in routine clinical mammography screening programs, can predict future breast cancer risk and can more accurately identify women who will most benefit from screening mammography. This DL score can be used in diverse clinical settings that currently depend on traditional risk scores, which are at best modest in performance. DL risk scores based on the mammogram alone can inform shared decision making with patients and their healthcare providers regarding more personalized risk reduction and early detection strategies. Citation Format: C Lehman, B Dontchos, LR Lamb./AI [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr ES7-3.

How Communicating Polygenic and Clinical Risk for Atherosclerotic Cardiovascular Disease Impacts Health Behavior: an Observational Follow-up Study.

Prediction tools that combine polygenic risk scores with clinical factors provide a new opportunity for improved prediction and prevention of atherosclerotic cardiovascular disease, but the clinical utility of polygenic risk score has remained unclear. We collected a prospective cohort of 7342 individuals (64% women, mean age 56 years) and estimated their 10-year risk for atherosclerotic cardiovascular disease both by a traditional risk score and a composite score combining the effect of a polygenic risk score and clinical risk factors. We then tested how returning the personal risk information with an interactive web-tool impacted on the participants' health behavior. When reassessed after 1.5 years by a clinical visit and questionnaires, 20.8% of individuals at high (>10%) 10-year atherosclerotic cardiovascular disease risk had seen a doctor, 12.4% reported weight loss, 14.2% of smokers had quit smoking, and 15.4% had signed up for health coaching online. Altogether, 42.6% of persons at high risk had made one or more health behavioral changes versus 33.5% of persons at low/average risk such that higher baseline risk predicted a favorable change (OR [CI], 1.53 [1.37-1.72] for persons at high risk versus the rest, P<0.001), with both high clinical (P<0.001) and genomic risk (OR [CI], 1.10 [1.03-1.17], P=0.003) contributing independently. Web-based communication of personal atherosclerotic cardiovascular disease risk-data including polygenic risk to middle-aged persons motivates positive changes in health behavior and the propensity to seek care. It supports integration of genomic information into clinical risk calculators as a feasible approach to enhance disease prevention.

Development and validation of a novel risk score to predict 5-year mortality in patients with acute myocardial infarction in China: a retrospective study.

BackgroundThe disease burden from ischaemic heart disease remains heavy in the Chinese population. Traditional risk scores for estimating long-term mortality in patients with acute myocardial infarction (AMI) have been developed without sufficiently considering advances in interventional procedures and medication. The goal of this study was to develop a risk score comprising clinical parameters and intervention advances at hospital admission to assess 5-year mortality in AMI patients in a Chinese population.MethodsWe performed a retrospective observational study on 2,722 AMI patients between January 2013 and December 2017. Of these patients, 1,471 patients from Changsha city, Hunan Province, China were assigned to the development cohort, and 1,251 patients from Xiangtan city, Hunan Province, China, were assigned to the validation cohort. Forty-five candidate variables assessed at admission were screened using least absolute shrinkage and selection operator, stepwise backward regression, and Cox regression methods to construct the C2ABS2-GLPK score, which was graded and stratified using a nomogram and X-tile. The score was internally and externally validated. The C-statistic and Hosmer-Lemeshow test were used to assess discrimination and calibration, respectively.ResultsFrom the 45 candidate variables obtained at admission, 10 potential predictors, namely, including Creatinine, experience of Cardiac arrest, Age, N-terminal Pro-Brain Natriuretic Peptide, a history of Stroke, Statins therapy, fasting blood Glucose, Left ventricular end-diastolic diameter, Percutaneous coronary intervention and Killip classification were identified as having a close association with 5-year mortality in patients with AMI and collectively termed the C2ABS2-GLPK score. The score had good discrimination (C-statistic = 0.811, 95% confidence intervals (CI) [0.786–0.836]) and calibration (calibration slope = 0.988) in the development cohort. In the external validation cohort, the score performed well in both discrimination (C-statistic = 0.787, 95% CI [0.756–0.818]) and calibration (calibration slope = 0.976). The patients were stratified into low- (≤148), medium- (149 to 218) and high-risk (≥219) categories according to the C2ABS2-GLPK score. The predictive performance of the score was also validated in all subpopulations of both cohorts.ConclusionThe C2ABS2-GLPK score is a Chinese population-based risk assessment tool to predict 5-year mortality in AMI patients based on 10 variables that are routinely assessed at admission. This score can assist physicians in stratifying high-risk patients and optimizing emergency medical interventions to improve long-term survival in patients with AMI.

Change in left atrioventricular coupling index to predict hard cardiovascular disease: The Multi-Ethnic Study of Atherosclerosis (MESA)

Both left atrial (LA) and left ventricular (LV) functional parameters influence the prognosis of patients with cardiovascular diseases (CVD). This study aimed to investigate the prognostic value of a novel left atrioventricular coupling index (LACI), as well as annual change in LACI (ΔLACI) for predicting hard CVD in a population without history of CVD at baseline. In the Multi-Ethnic Study of Atherosclerosis (MESA), 2221 study participants, free of CVD at baseline, had LACI assessed with CMR imaging at baseline (Exam 1, 2000–2002), and ten years later (Exam 5, 2010–2012). LACI was defined by the ratio of the LA end-diastolic volume divided by the LV end-diastolic volume. Cox proportional hazard models were used to evaluate the associations of LACI and average annualized change in LACI (ΔLACI) with hard CVD defined by myocardial infarction, resuscitated cardiac arrest, fatal and nonfatal stroke, or CHD death. Among the 2221 participants (62.8 ± 10.1 years, 47.9% men), 223 cardiovascular events were observed (mean follow-up 13.0 ± 3.2 years). After adjustment for traditional risk factors, greater LACI and ΔLACI were independently associated with hard CVD (HR: 1.23; 95% CI [1.13–1.34] and HR: 1.71, 95% CI [1.50–1.94]; both P < 0.0001). Adjusted models for LACI and ΔLACI showed significant improvement in model discrimination compared to traditional risk score model (C-statistic: 0.79 vs. 0.77, and C-statistic: 0.80 vs. 0.78, respectively) ( Fig. 1 ). LACI is a strong predictor for the incidence of hard CVD. LACI has incremental prognostic value to predict hard CVD over traditional risk factors, and better discrimination and reclassification power compared to individual LA or LV parameters.

504 Comparison of cardiovascular risk scores to predict anthracycline-induced cardiotoxicity in early breast cancer patients

Abstract Aims Cardiovascular (CV) toxicities related to chemotherapy have been reported with increasing frequency. Thus, latest cardio-oncology guidelines focused on the importance of a baseline CV risk assessment before starting a potentially cardiotoxic regimen, in order to identify individuals at greater risk of developing CV toxicities and to prevent them. Despite these recommendations, there are still no validated systems to predict cardiotoxicity risk in oncological patients. The aim of the present study was to compare the average risk assessed by conventional CV risk scores (SCORE, QRISK2, ASCVD) and two recently developed cardiotoxicity risk scores (Baseline cardiovascular risk stratification proformas, Multivariable prediction model for major adverse cardiovascular events after early stage breast cancer) in a cohort of 41 (100% female; mean age 52.9 ± 11) early breast cancer (EBC) patients treated with epirubicin and to investigate the relationship between CV and cardiotoxicity risk and the rate of CV outcome, defined as biomarkers (high sensibility Troponin I and BNP) increase during the follow-up. Results A baseline CV risk assessment was performed in all patients before starting chemotherapy according to prespecified scores, showing a higher mean CV risk in our cohort compared to general population. Most patients were at low risk of developing cardiotoxicity according to prespecified scores. During the follow-up (mean 221 ± 75 days), the primary outcome occurred in 9.8% (n = 4). The average CV risk scores of patients developing the primary outcome were higher than the controls according to both conventional and cardiotoxicity risk scores (Figure 1). The QRISK2 achieved the best ratio cases average score/controls average score (r = 3) and the ROC curve confirmed QRISK2 as the best CV risk score to predict the outcome in our cohort, reaching the highest sensitivity and specificity (Figure 2). Conclusions In a cohort of young patients affected by EBC the mean CV risk is higher than the general young female population according to traditional CV risk scores. Moreover, patients developing the primary outcome during the Follow-up were those at higher CV risk and high risk of cardio toxicity. Surprisingly the best score to predict the incidence of outcome was the QRISK2, confirming the crucial role of baseline CV risk in cancer patients.

60. Creation and Comparison of a Machine Learning Decision Tree and Traditional Risk Score to Predict Ceftriaxone Resistance in Cancer Patients with E. coli Bacteremia

BackgroundThere are several clinical tools that have been developed to predict the likelihood of extended-spectrum β-lactamase producing Enterobacterales; however, the creation of these tools included few patients with cancer or otherwise immunosuppressed. The objectives of this retrospective cohort study were to develop a decision tree and traditional risk score to predict ceftriaxone resistance in cancer patients with Escherichia coli (E. coli) bacteremia as well as to compare the predictive accuracy between the tools.MethodsAdults age ≥ 18 years old with E. coli bacteremia at The University of Texas MD Anderson Cancer Center from 1/2018 to 12/2019 were included. Isolates recovered within 1 week from the same patient were excluded. The decision tree was constructed using classification and regression tree analysis, with a minimum node size of 10. The risk score was created using a multivariable logistic regression model derived by using stepwise variable selection with backward elimination at level 0.2. The decision tree and risk score statistical metrics were compared.ResultsA total of 629 E. coli isolates were screened, of which 580 isolates met criteria. Ceftriaxone-resistant (CRO-R) E. coli accounted for 36% of isolates. The machine learning-derived decision tree included 5 predictors whereas the logistic regression-derived risk score included 7 predictors. The risk score cutoff point of ≥ 5 points demonstrated the most optimized overall classification accuracy. The positive predictive value of the decision tree was higher than that of the risk score (88% vs 74%, respectively), but the area under the receiver operating characteristic curve and model accuracy of the risk score was higher than that of the decision tree (0.85 vs 0.73 and 82% vs 74%, respectively). Figure 1. Clinical Decision Tree Table 1. Regression Model and Assigned Points for Clinical Risk Score Table 2. Statistical Metrics of Clinical Decision Tree and Clinical Risk Score ConclusionThe decision tree and risk score can be used to determine the likelihood of whether a cancer patient with E. coli bacteremia has a CRO-R infection. In both clinical tools, the strongest predictor was a history of CRO-R E. coli colonization or infection in the last 6 months. The decision tree was more user-friendly, has fewer variables, and has a better positive predictive value in comparison to the risk score. However, the risk score has a significantly better discrimination and model accuracy than that of the decision tree.Disclosures Samuel L. Aitken, PharmD, MPH, BCIDP, Melinta Therapeutoics (Individual(s) Involved: Self): Consultant, Grant/Research Support

Abstract 10074: Cardiovascular Outcomes in Patients with Concurrent Atrial Fibrillation and Myeloproliferative Neoplasms

Introduction: Myeloproliferative neoplasms (MPNs), including polycythemia vera (PV), essential thrombocytosis (ET) and myelofibrosis (MF), are clonal stem cell neoplasms. Therapy is instituted to prevent thrombosis, a leading cause of morbidity/mortality in this population. Atrial fibrillation (AF) is also associated with increased risk of thromboembolism. Risk scores, including CHA 2 DS 2 VASC, are used to determine benefit of anticoagulation (AC) but have not been validated in patients with MPN. Methods: This is a single-center, retrospective cohort study of patients with concurrent MPN and AF from 2000 to 2020 (n = 142; 62 ET, 54 PV, 26 MF). Outcomes included composite of arterial thrombotic events (ATE; MI, ischemic stroke, or peripheral arterial embolism) and cardiovascular (CV) death, bleeding requiring hospitalization, and all-cause mortality (ACM). Results: AC was used in 69% of all MPN patients (73% of ET/ PV, 50% of MF). CHA 2 DS 2 VASC scores were significantly higher in patients with ET/PV compared with MF (4 ± 2 vs 3 ± 1, p = 0.0019). After median follow-up (FU) of 49 months (IQR 18, 86) the composite outcome occurred in 39% of patients with MPN (ET/PV 36%, MF 50%). Bleeding occurred in 30% of patients (ET/PV 29%, MF 30%) and ACM in 45% of all patients (ET/PV 40%, MF 83%). Patients with MF had worse composite-free survival (log-rank p = 0.0043, Fig-A) and no difference in bleed-free survival (Fig-B) compared with ET/PV patients. After multivariable logistic regression including CHA 2 DS 2 VASC, age, sex, MF, AC, CAD, hypertension, diabetes, prior stroke, aspirin use, MPN treatment and FU time as co-variables, MF was associated with the composite (aOR 4.1, 95% CI 1.2-14.9), CHA 2 DS 2 VASC was not (aOR 1.3, 95% CI 0.8-2.1). Conclusions: Patients with MPN and AF have high rates of adverse outcomes with MF having worse composite-free survival compared to ET/PV. Traditional risk scores may underperform in this patient population underscoring the need for further investigation.

Abstract 12965: Race-Specific Models to Predict In-Hospital Mortality in Patients With Heart Failure Using Machine Learning: The American Heart Association Get With the Guidelines Registry

Background: Prior prognostic models for acute decompensated HF (ADHF) have incorporated race as a covariate may not have completely account for the social and biological factors that underlie the racial disparities in outcomes. We developed race-specific models to improve risk prediction and better identify race-specific predictors of mortality in patients with ADHF using machine learning (ML) methods. Methods: Race-specific ML-models for in-hospital mortality were developed and validated in ADHF patients enrolled in the GWTG-HF registry between 2007 to 2020 (14,586 Black and 50,351 Non-Black adults) using over 40 candidate variables. External validation was performed among participants from the ARIC study with ADHF (n=1,115 Black and 2,028 Non-Black adults). The discrimination and calibration was compared with a previously validated risk score and cohort-specific logistic regression (LR) models. The changes in model reclassification performance with addition of zip-code level data of socioeconomic (SES) parameters was also assessed. Results: In the GWTG-HF cohort, the ML models had superior performance in both Black and Non-Black participants compared with the traditional GWTG-HF risk score ( Table ). The superior performance of the ML-model was also observed in the external validation cohort as compared with the other LR models ( Table ). Addition of measures of SES did not meaningfully change the performance but improved the reclassification metrics of the ML-models with net up-classification of risk in Black [NRI = 0.21 (0.07, 0.35)] and down-classification in Non-Black patients [NRI = -0.08 (-0.12, -0.04)]. SES was responsible for 21.7% of the total attributable mortality risk in Black adults compared to 0.4% in Non-Black adults. Conclusion: Race-specific, ML-based mortality models demonstrated superior performance when compared to traditional HF mortality risk models. Socioeconomic distress is an important contributor to risk in Black adults.

Using machine learning to predict atrial fibrillation diagnosed after ischemic stroke

BackgroundSelecting best candidates for prolonged poststroke cardiac monitoring in acute ischemic stroke (AIS) patients is still challenging. We aimed to develop a machine learning (ML) model to select AIS patients at high risk of poststroke atrial fibrillation (AF) for prolonged cardiac monitoring and then to compare ML model with traditional risk scores and classic statistical logistic regression (classic-LR) model. MethodsAIS patients from July 2012 to September 2020 across Nanjing First Hospital were collected. We performed the LASSO regression for selecting the critical features and built five ML models to assess the risk of poststroke AF. The SHAP and partial dependence plot (PDP) method were introduced to interpret the optimal model. We also compared ML model with CHADS2 score, CHA2DS2-VASc score, AS5F score, HAVOC score, and classic-LR model. ResultsA total of 3929 AIS patients were included. Among the five ML models, deep neural network (DNN) was the model with best performance. It also exhibited superior performance compared with CHADS2 score, CHA2DS2-VASc score, AS5F score, HAVOC score and classic-LR model. The results of SHAP and PDP method revealed age, cardioembolic stroke, large-artery atherosclerosis stroke, and NIHSS score at admission were the top four important features and revealed the DNN model had good interpretability and reliability. ConclusionThe DNN model achieved best performance and improved prediction performance compared with traditional risk scores and classic-LR model. The DNN model can be applied to identify AIS patients at high risk of poststroke AF as best candidates for prolonged poststroke cardiac monitoring.

Predicting cardiovascular risk using a novel risk score in young and middle-age adults with HIV: associations with biomarkers and carotid atherosclerotic plaque.

Traditional risk factors associated with cardiovascular disease (CVD) include older age, smoking, poor diet, lack of exercise, obesity, high blood pressure, high cholesterol, and family history. Young-to-middle age adults (YMAA) are less often identified as being at risk of CVD, but traditional risk scores primarily target older adults and do not accurately estimate risk among YMAA. This study examined biomarkers associated with CVD risk in YMAA in the context of HIV and cocaine use; risk was assessed by two methods: (1) a relative cardiovascular (CV) risk score that includes several factors and (2) carotid atherosclerotic plaque. Associations between CVD risk (CV risk score and carotid atherosclerotic plaque) and proinflammatory cytokines, markers of immune activation, HIV status, and cocaine use were examined. Participants (N = 506) included people with and without HIV and people who use or do not use cocaine. Participants' mean age was 36 (SD = 9.53); half (51%) were men. Cocaine use and C-reactive protein were associated with greater relative CV risk scores, but no associations between biomarkers and CV risk emerged. Age and CV risk scores were associated with carotid atherosclerotic plaque, but biomarkers were not. HIV was not associated with CV risk scores or carotid atherosclerotic plaque. Among YMAA, CV risk scores may help providers identify lifestyle changes needed among those at risk for CVD before more advanced risk (e.g., atherosclerotic plaque) is identified. Implications are discussed.

Machine learning with D-dimer in the risk stratification for pulmonary embolism: a derivation and internal validation study.

To develop a machine learning model to predict the diagnosis of pulmonary embolism (PE). We undertook a derivation and internal validation study to develop a risk prediction model for use in patients being investigated for possible PE. The machine learning technique, generalized logistic regression using elastic net, was chosen following an assessment of seven machine learning techniques and on the basis that it optimized the area under the receiver operator characteristic curve (AUC) and Brier score. Models were developed both with and without the addition of D-dimer. A total of 3347 patients were included in the study of whom, 219 (6.5%) had PE. Four clinical variables (O2 saturation, previous deep venous thrombosis or PE, immobilization or surgery, and alternative diagnosis equal or more likely than PE) plus D-dimer contributed to the machine learning models. The addition of D-dimer improved the AUC by 0.16 (95% confidence interval 0.13-0.19), from 0.73 to 0.89 (0.87-0.91) and decreased the Brier score by 14% (10-18%). More could be ruled out with a higher positive likelihood ratio than by the Wells score combined with D-dimer, revised Geneva score combined with D-dimer, or the Pulmonary Embolism Rule-out Criteria score. Machine learning with D-dimer maintained a low-false-negative rate at a true-negative rate of nearly 53%, which was better performance than any of the other alternatives. A machine learning model outperformed traditional risk scores for the risk stratification of PE in the emergency department. However, external validation is needed.

Reduced progression of carotid intima media thickness by pictorial presentation of subclinical atherosclerosis 3-year follow-up from VIPVIZA a randomized controlled trial

Abstract Background Reduced progression of carotid intima media thickness (cIMT) and atherosclerotic plaque assessed by carotid ultrasound has been reported in pharmacological intervention studies. We have previously provide evidence for improved primary prevention of cardiovascular diseases by pictorial presentation of subclinical atherosclerosis severity based on reduction in traditional risk scores in 1-year follow-up. Purpose We aimed to investigate if pictorial presentation of subclinical atherosclerosis severity affected cIMT and plaque progression over a 3-year follow-up period in comparison with a routine primary prevention program. Methods Visualization of asymptomatic atherosclerotic disease for optimum cardiovascular prevention (VIPVIZA) study is a population-based, randomized controlled trial with blinded evaluators (PROBE design). Participants aged 40, 50 and 60 years were enrolled from a prevention program within the routine primary care. Carotid ultrasound examinations were made at baseline and at 3-year follow-up (n: 3154). The ultrasound examinations included standardized measurements of cIMT in the far wall of common carotid artery at predefined angles due to Meijer's arc. Used angles was 240 degrees for left-mean-cIMT and 120 degrees for right-mean-cIMT. Detection of plaque was based on Mannheim consensus and plaque prevalence was evaluated as absent, unilateral or bilateral plaques. Total plaque area (TPA) was a sum of plaque areas and measured off-line form the longitudinal 2D B-mode images. At baseline, the intervention group (n: 1575) and their primary care physicians received a pictorial presentation describing the severity of subclinical atherosclerosis with graphs and colored figures based on measured cIMT and plaque prevalence (Figure 1). The control group with respective physician (n: 1579) did not receive any information about ultrasound results. The bilateral, left and right mean cIMT, plaque prevalence and TPA at 3-year follow-up were compared between groups. Analysis was performed by analysis of covariance and ordinal proportional odds models. Bonferroni correction was applied to account for multiple comparisons, each individual test was performed using α = 0.01. Results Reduced cIMT progression in the intervention group was found in left-mean-cIMT with an estimated group difference of −0.011 mm (p=0.001). Estimated group difference of right- and bilateral-mean-cIMT were both −0.005 mm (p=0.223 and 0.036, respectively) (Figure 2). No significant difference between groups was found for plaque prevalence or TPA. Conclusion Intervention by pictorial presentation of subclinical atherosclerotic severity reduced the progression of cIMT in comparison with a traditional cardiovascular preventive program only. This was largely driven by changes in the left carotid artery. Long-term follow-up will be required to elucidate if the intervention will have a protective effect on future risk of cardiovascular events. Funding Acknowledgement Type of funding sources: Foundation. Main funding source(s): The Heart Foundation of Northern Sweden, Västerbotten County Council. Figure 1Figure 2

Comparison of traditional versus artificial intelligence based coronary artery disease risk prediction scores in young patients with acute coronary syndrome

Abstract Background Ever since the concept of preventive cardiology has come into vogue, several risk identification models have come up which combine several risk factors to create a risk prediction score for occurrence of cardiovascular (CV) event. While carrying a proven validation in Western population, none of the risk prediction model has been satisfactorily evaluated in Indians especially young &amp;lt;40 years old. Objectives To compare Artificial Intelligence based novel risk score with traditional risk scores in young (less than 40 years age) patients presenting with acute coronary syndrome (ACS) and to estimate the relative efficacy of different coronary artery disease (CAD) risk scores in young Indian Patients. Design Single center, Observational, Non-interventional study. Participants Cohort of Patients more than 20 but less than 40 years old with ACS in the department of Cardiology from 1st January 2019 to 31st October 2019. Methods 314 young patients [mean age 36.14±4.17 years] presenting with acute coronary syndrome (ACS) were enrolled. The three clinically most pertinent risk assessment models [Framingham Risk score (FRS), World Health Organization risk prediction charts (WHO/ISH), and QRISK3 scores] and Artificial Intelligence based novel risk score (AICVD) were applied on day 1 of presentation, and tried to see whether one risk score versus other risk score could have predicted the event earlier had we applied it before the occurrence of ACS. Risk factors considered included those already in traditional scoring systems and new risk factors (diet, alcohol, tobacco, dyslipidemia, physical activity, family history of heart disease, history of heart disease, heart rate, respiratory rate, chronic heart symptoms and psychological stress). Results WHO/ISH provided the lowest high risk estimate with only 1 (0.9%) patient estimated to be having &amp;gt;20% 10-year risk. The FRS estimated high risk (&amp;gt;20% 10-year risk) in 3 (1%) patients. The QRISK3 estimated high risk (&amp;gt;10% 10-year risk) in 20 (6.5%) patient. In comparison, AICVD risk prediction model stood tall by identifying 73 (23.2%) patients as high risk and 62.74% patients as more than moderate risk for having CV events at 7 years (p&amp;lt;0.001). Conclusion Perhaps, this is the first study which has compared artificial intelligence based novel risk prediction model with the three most commonly applied models, in the young Indian patients. We found that a cohort of young Indian patients presenting with ACS, when studied retrospectively, was identified as “high risk” most likely by AICVD risk prediction model rather than the traditional counterparts. The WHO/ISH risk prediction charts and FRS were the poorest predictors. Performance of QRISK3 score also remained less than satisfactory. These findings suggested that AICVD risk prediction model is a promising tool to assess for CV risk in Indian population. Funding Acknowledgement Type of funding sources: None. Predictability of risk prediction models

REadmission PREvention in SepSis: Development and Validation of a Prediction Model.

Hospital 30-day readmissions remain a major quality and cost indicator. Traditional readmission risk scores, such as LACE (length of stay, acuity of admission, Charlson comorbidity index, and emergency department visits), may be suboptimal in special patient populations, such as those with sepsis. As sepsis survivorship improves, there is a need to determine which variables might be associated with a decrease in 30-day readmission. We completed a retrospective analysis reviewing patients with sepsis who had unplanned 30-day readmissions. Multivariate regression analysis was performed for the REadmission PREvention in SepSis (REPRESS) model, which evaluated age, length of stay, Charlson disease count, Richmond Agitation-Sedation Scale score, discharge to a skilled nursing facility, and mobility for predictive significance in hospital readmission. Our REPRESS model performed better when compared with LACE for predicting readmission risk in a sepsis population.

External validation of the 4C Mortality Score for patients with COVID-19 and pre-existing cardiovascular diseases/risk factors

ObjectivesPredictive algorithms to inform risk management decisions are needed for patients with COVID-19, although the traditional risk scores have not been adequately assessed in Asian patients. We aimed to evaluate...