Mortality Risk Stratification Research Articles

ABC2-SPH risk score for in-hospital mortality in COVID-19 patients: development, external validation and comparison with other available scores

ObjectivesThe majority of available scores to assess mortality risk of coronavirus disease 2019 (COVID-19) patients in the emergency department have high risk of bias. Therefore, this cohort aimed to develop and validate a score at hospital admission for predicting in-hospital mortality in COVID-19 patients and to compare this score with other existing ones. MethodsConsecutive patients (≥ 18 years) with confirmed COVID-19 admitted to the participating hospitals were included. Logistic regression analysis was performed to develop a prediction model for in-hospital mortality, based on the 3978 patients admitted between March–July, 2020. The model was validated in the 1054 patients admitted during August–September, as well as in an external cohort of 474 Spanish patients. ResultsMedian (25–75th percentile) age of the model-derivation cohort was 60 (48–72) years, and in-hospital mortality was 20.3%. The validation cohorts had similar age distribution and in-hospital mortality. Seven significant variables were included in the risk score: age, blood urea nitrogen, number of comorbidities, C-reactive protein, SpO2/FiO2 ratio, platelet count, and heart rate. The model had high discriminatory value (AUROC 0.844, 95% CI 0.829–0.859), which was confirmed in the Brazilian (0.859 [95% CI 0.833–0.885]) and Spanish (0.894 [95% CI 0.870–0.919]) validation cohorts, and displayed better discrimination ability than other existing scores. It is implemented in a freely available online risk calculator (https://abc2sph.com/). ConclusionsAn easy-to-use rapid scoring system based on characteristics of COVID-19 patients commonly available at hospital presentation was designed and validated for early stratification of in-hospital mortality risk of patients with COVID-19.

Inflammation-based assessment for the risk stratification of mortality in patients with heart failure

The Glasgow Prognostic Score (GPS) has been established as a useful resource to evaluate inflammation and malnutrition and predict prognosis in several cancers. However, its prognostic significance in patients with heart failure (HF) is not well established. To investigate the association between the GPS and mortality in patients with HF, we assessed 870 patients who were 20 years old and more and had been admitted for acute decompensated HF. The GPS ranged from 0 to 2 points as previously reported. Over the 18-month follow-up (follow-up rate, 83.9%), 143 patients died. Increasing GPS was associated with higher HF severity assessed by New York Heart Association functional class and B-type natriuretic peptide (BNP) levels. Kaplan–Meier analysis showed significant associations for mortality and increased GPS. In multivariate analysis, compared to the GPS 0 group, the GPS 2 group was associated with high mortality (hazard ratio 2.92, 95% confidence interval 1.77–4.81, p < 0.001) after adjustment for age, sex, blood pressure, HF history, HF severity, hemoglobin, renal function, sodium, BNP, left ventricular ejection fraction, and anti-HF medications. In conclusion, high GPS was significantly associated with worse prognosis in patients with HF. Inflammation-based assessment by the GPS may enable simple evaluation of HF severity and prognosis.

How to Synchronize Longitudinal Patient Data With the Underlying Disease Progression: A Pilot Study Using the Biomarker CRP for Timing COVID-19.

The continued digitalization of medicine has led to an increased availability of longitudinal patient data that allows the investigation of novel and known diseases in unprecedented detail. However, to accurately describe any underlying pathophysiology and allow inter-patient comparisons, individual patient trajectories have to be synchronized based on temporal markers. In this pilot study, we use longitudinal data from critically ill ICU COVID-19 patients to compare the commonly used alignment markers “onset of symptoms,” “hospital admission,” and “ICU admission” with a novel objective method based on the peak value of the inflammatory marker C-reactive protein (CRP). By applying our CRP-based method to align the progression of neutrophils and lymphocytes, we were able to define a pathophysiological window that improved mortality risk stratification in our COVID-19 patient cohort. Our data highlights that proper synchronization of longitudinal patient data is crucial for accurate interpatient comparisons and the definition of relevant subgroups. The use of objective temporal disease markers will facilitate both translational research efforts and multicenter trials.

Prognostic value of plasma exosomal levels of histone H3 protein in patients with heat stroke.

Heat stroke (HS) is a condition that can lead to multiple organ dysfunction syndrome and death; however, there is no reliable method for stratifying mortality risk in HS. The abundance of exosomes in the circulation and their contents may be used as potential biomarkers of HS. The present study aimed to examine whether histone H3 levels in plasma exosomes could be used to determine HS prognosis. Blood samples were collected from patients with HS (36 survivors and 8 non-survivors) at admission to the intensive care unit and 4 days after admission. Blood samples were additionally collected from 15 healthy volunteers. Plasma exosomes were isolated using high-speed differential centrifugation. Correlation between histone H3 level and organ function and disease severity was examined. The results suggested differential expression and enrichment of histone H3 in the plasma exosomes of patients with HS (survivors, 249.3±04.6; non-survivors, 500.4±216.8; healthy controls, 161.1±52.49 pg/100 µg; P<0.05). The increased expression of histone H3 was associated with increased disease severity and duration. Plasma exosomal levels of histone H3 were significantly correlated with both organ dysfunction and disease severity (P<0.0001) and were significantly different between non-survivors and survivors (area under the receiver operating characteristic curve, 0.9668). A cutoff value of 307 pg/100 µg demonstrated optimized sensitivity (95%) and specificity (91.67%) for predicting mortality risk, suggesting that histone H3 levels in plasma exosomes may be a reliable biomarker for HS prognosis.

Mitral Regurgitation International Database (MIDA) Score Predicts Outcome in Patients With Heart Failure Undergoing Transcatheter Edge‐to‐Edge Mitral Valve Repair

BackgroundOptimizing risk stratification in patients undergoing transcatheter mitral valve repair is an ongoing challenge. The Mitral Regurgitation International Database (MIDA) score represents a user‐friendly mortality risk stratification tool that is validated on a large‐scale registry of patients with degenerative mitral regurgitation (MR). We here assessed the potential benefit of the MIDA risk score for patients with functional or degenerative MR undergoing transcatheter mitral valve repair.Methods and ResultsIn total, 680 patients undergoing MitraClip implantation were stratified according to MIDA score tertiles into a low (0–7), intermediate (8–9), and a high (10–12) MIDA score group. MR was assessed in follow‐up echocardiograms in 416 patients at 323±169 days after transcatheter mitral valve repair. During 2‐year follow‐up, 8.2% (15/182) of patients with low, 21.3% (64/300) with intermediate, and 26.3% (52/198) with high MIDA score died (log‐rank test P<0.001). Hazard of all‐cause mortality increased by 13% (95% CI, 3%–25%) with every additional point of the MIDA score. Subanalysis of 431 patients with functional MR showed similar results. Furthermore, rates of a combined end point of mortality and hospitalization for heart failure were higher with increasing MIDA score (30% [54/182], 38% [113/300] and 48% [94/198], respectively, log‐rank test P=0.001). Frequency of residual MR ≥II at follow‐up increased with increasing MIDA score group (33%, 44%, and 59%, respectively, P<0.001).ConclusionsThe MIDA mortality risk score maintains its predictive utility in patients undergoing transcatheter mitral valve repair, regardless of MR cause. Moreover, it was predictive of worse event‐free survival regarding a combined end point of mortality and hospitalization for heart failure, and was associated with postprocedural residual MR ≥II and MR recurrence.

Is estimated cardiorespiratory fitness an effective predictor for cardiovascular and all-cause mortality? A meta-analysis

Background and aimsEstimated cardiorespiratory fitness (eCRF) derived from algorithm correlates well with exercise testing-measured CRF, yet its clinical use for mortality risk stratification has not been systematically evaluated. This meta-analysis with dose-response analysis was conducted to quantify its association with risk of cardiovascular and all-cause mortality. MethodsElectronic databases were searched for prospective cohort studies that investigated the association of eCRF with risk of cardiovascular and all-cause mortality. Study-specific multivariate-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) per 1-metabolic equivalent (MET) higher of eCRF were pooled using a random-effects model. ResultsTwenty-five datasets from 8 cohort studies that enrolled more than 170,000 participants were included. The summary HR per 1-MET higher of eCRF was 0.83 (95% CI 0.80 to 0.86) for cardiovascular mortality (11 datasets) and 0.83 (95% CI 0.78 to 0.88) for all-cause mortality (14 datasets) in the general population. These associations showed no sex-difference and were all linearly shaped (all pnonlinearity ≥ 0.27). The performance of eCRF (assessed by the area under the curve) in discriminating future risk of cardiovascular and all-cause mortality was higher than all its components (such as physical activity, resting heart rate, and body mass index, all p < 0.05), but slightly lower than exercise testing-measured CRF. ConclusionsHigher eCRF was independently associated with lower risk of cardiovascular and all-cause mortality in the general population, indicating that eCRF might hold the potential as an effective and practical risk prediction tool in epidemiological or population research.

Prospective validation of an 11-gene mRNA host response score for mortality risk stratification in the intensive care unit

Several clinical calculators predict intensive care unit (ICU) mortality, however these are cumbersome and often require 24 h of data to calculate. Retrospective studies have demonstrated the utility of whole blood transcriptomic analysis in predicting mortality. In this study, we tested prospective validation of an 11-gene messenger RNA (mRNA) score in an ICU population. Whole blood mRNA from 70 subjects in the Stanford ICU Biobank with samples collected within 24 h of Emergency Department presentation were used to calculate an 11-gene mRNA score. We found that the 11-gene score was highly associated with 60-day mortality, with an area under the receiver operating characteristic curve of 0.68 in all patients, 0.77 in shock patients, and 0.98 in patients whose primary determinant of prognosis was acute illness. Subjects with the highest quartile of mRNA scores were more likely to die in hospital (40% vs 7%, p < 0.01) and within 60 days (40% vs 15%, p = 0.06). The 11-gene score improved prognostication with a categorical Net Reclassification Improvement index of 0.37 (p = 0.03) and an Integrated Discrimination Improvement index of 0.07 (p = 0.02) when combined with Simplified Acute Physiology Score 3 or Acute Physiology and Chronic Health Evaluation II score. The test performed poorly in the 95 independent samples collected > 24 h after emergency department presentation. Tests will target a 30-min turnaround time, allowing for rapid results early in admission. Moving forward, this test may provide valuable real-time prognostic information to improve triage decisions and allow for enrichment of clinical trials.

Prognostic Value of B-Type Natriuretic Peptide in Liver Transplant Patients: Implication in Posttransplant Mortality.

Despite frequent cirrhotic cardiomyopathy or subclinical heart failure (HF), the prognostic value of peri-liver transplant (LT) B-type natriuretic peptide (BNP) has been poorly studied in advanced liver disease. We examined the association between BNP and mortality in a large cohort of LT patients and identified risk factors for peri-LT BNP increase. Using prospectively collected data from the Asan LT Registry between 2008 and 2019, 3,811 patients who measured serial pretransplant BNP (preBNP) and peak BNP levels within the first 3 posttransplant days (postBNPPOD3 ) were analyzed. Thirty-day all-cause mortality predicted by adding preBNP and/or postBNPPOD3 to the traditional Revised Cardiac Risk Index (RCRI) was evaluated. PreBNP>400pg/mL (known cutoff of acute HF) was found in 298 (7.8%); however, postBNPPOD3 >400pg/mL was identified in 961 (25.2%) patients, specifically in 40.4% (531/1,315) of those with a Model for End-Liver Disease score (MELDs) > 20. Strong predictors of postBNPPOD3 >400pg/mL were preBNP, hyponatremia, and MELDs, whereas those of preBNP>400pg/mL were MELDs, kidney failure, and respiratory failure. Among 100 (2.6%) post-LT patients who died within 30days, patients with postBNPPOD3 ≤150pg/mL (43.1%, reference group), 150-400pg/mL (31.7%), 400-1,000pg/mL (18.5%), 1,000-2,000pg/mL (4.7%), and >2,000pg/mL (2.0%) had 30-day mortalities of 0.9%, 2.2%, 4.0%, 7.7%, and 22.4%, respectively. Adding preBNP, postBNPPOD3 , and both BNP to RCRI improved net reclassification index to 22.5%, 29.5%, and 33.1% of 30-day mortality, respectively. PostBNPPOD3 >400pg/mL after LT was markedly prevalent in advanced liver disease and mainly linked to elevated preBNP. Routine monitoring of peri-LT BNP provides incremental prognostic information; therefore, it could help risk stratification for mortality as a practical and useful biomarker in LT.

Prognostic Value of Cortisol Index of Endobiogeny in Acute Myocardial Infarction Patients.

Background and Objectives: Serum cortisol has been extensively studied for its role during acute myocardial infarction (AMI). Reports have been inconsistent, with high and low serum cortisol associated with various clinical outcomes. Several publications claim to have developed methods to evaluate cortisol activity by using elements of complete blood count with its differential. This study aims to compare the prognostic value of the cortisol index of Endobiogeny with serum cortisol in AMI patients, and to identify if the risk of mortality in AMI patients can be more precisely assessed by using both troponin I and cortisol index than troponin I alone. Materials and methods: This prospective study included 123 consecutive patients diagnosed with AMI. Diagnostic coronary angiography and revascularization was performed for all patients. Cortisol index was measured on admission, on discharge, and after 6 months. Two year follow-up for all patients was obtained. Results: Our study shows cortisol index peaks at 7–12 h after the onset of AMI, while serum cortisol peaked within 3 h from the onset of AMI. The cortisol index is elevated at admission, then significantly decreases at discharge; furthermore, the decline to its bottom most at 6 months is observed with mean values being constantly elevated. The cortisol index on admission correlated with 24-month mortality. We established combined cut-off values of cortisol index on admission > 100 and troponin I > 1.56 μg/las a prognosticator of poor outcomes for the 24-month period. Conclusions: The cortisol index derived from the global living systems theory of Endobiogeny is more predictive of mortality than serum cortisol. Moreover, a combined assessment of cortisol index and Troponin I during AMI offers more accurate risk stratification of mortality risk than troponin alone.

Machine Learning for Predicting Mortality in Transcatheter Aortic Valve Implantation: An Inter-Center Cross Validation Study.

Current prognostic risk scores for transcatheter aortic valve implantation (TAVI) do not benefit yet from modern machine learning techniques, which can improve risk stratification of one-year mortality of patients before TAVI. Despite the advancement of machine learning in healthcare, data sharing regulations are very strict and typically prevent exchanging patient data, without the involvement of ethical committees. A very robust validation approach, including 1300 and 631 patients per center, was performed to validate a machine learning model of one center at the other external center with their data, in a mutual fashion. This was achieved without any data exchange but solely by exchanging the models and the data processing pipelines. A dedicated exchange protocol was designed to evaluate and quantify the model’s robustness on the data of the external center. Models developed with the larger dataset offered similar or higher prediction accuracy on the external validation. Logistic regression, random forest and CatBoost lead to areas under curve of the ROC of 0.65, 0.67 and 0.65 for the internal validation and of 0.62, 0.66, 0.68 for the external validation, respectively. We propose a scalable exchange protocol which can be further extended on other TAVI centers, but more generally to any other clinical scenario, that could benefit from this validation approach.

Effectiveness of sepsis bundle application and outcomes predictors to cirrhotic patients with septic shock

IntroductionCirrhotic patients with septic shock have a poorer prognosis compared with the general population. Our study aimed to investigate the survival benefit of the implementation of hour-1 bundle proposed by Surviving Sepsis Campaign, and to analyze the predictors associated with short-term mortality of these patients.MethodsA single-center, retrospective case-control study was conducted among adult patients who visited the emergency department between January 1, 2018 and December 31, 2019. All patients with a diagnosis of liver cirrhosis and septic shock were enrolled. Their baseline characteristics, laboratory results, source of sepsis, and sepsis bundle management were recorded. We further divided the patients into survivor and non-survivor groups to identify independent prognostic factors.ResultsA total of 88 patients were eligible for this study. The overall 30-day mortality rate was 53.4% (47/88). The proportion of hour-1 bundle achievement was 30.7% (27/88). There were no significant mortality differences between the hour-1 bundle achievement and non-achievement groups (44.4% vs. 57.4%, p = 0.35). Compared with the patients in the survivor group, patients in the non-survivor group had significantly more advanced stage of cirrhosis and a lower proportion of receiving source control (4.3% vs. 22.0%, p = 0.02). The chronic liver failure-sequential organ failure assessment (CLIF-SOFA) score (adjusted hazard ratio [AHR] =1.52, p < 0.01), serum lactate (AHR =1.03, p < 0.01), and source control (AHR =0.54, p = 0.02) were identified as independent prognostic factors in the multivariate regression model. Furthermore, the CLIF-SOFA score (area under curve [AUC]: 0.81) and lactate levels (AUC: 0.77) revealed good mortality discrimination ability in cirrhotic patients with septic shock.ConclusionsThe application of the hour-1 bundle did not reveal a significant survival benefit to cirrhotic patients with septic shock. Clinicians could utilize CLIF-SOFA scores and lactate levels for mortality risk stratification and put more emphasis on the feasibility of source control to improve their prognosis.

Clinical frailty scale as a point of care prognostic indicator of mortality in COVID-19: a systematic review and meta-analysis.

BackgroundCOVID-19 has resulted in the largest pandemic experienced since 1918, accounting for over 2 million deaths globally. Frail and older people are at the highest risk of mortality. The main objective of the present research was to quantify the impact of clinical frailty scale (CFS) by increasing severity of frailty and to identify other personal prognostic factors associated with increased mortality from COVID-19.MethodsThis study offers a contemporary systematic review and meta-analysis to analyse the stratified mortality risk by increasing CFS sub-categories (1–3, 4–5 and 6–9). Databases searched included EMBASE, MEDLINE, CAB s, PsychInfo, and Web of Science with end-search restriction the 18th December 2020. Publications identified via MedRevix were followed up on the 23rd March 2021 in peer-reviewed database search, and citations were updated as published. Prospective and retrospective cohort studies which reported the association between CFS and COVID-19 mortality were included. Thirty-four studies were eligible for systematic review and seventeen for meta-analysis, with 81–87% (I2) heterogeneity.FindingsAll studies [N: 34] included patients from a hospital setting, comprising a total of 18,042 patients with mean age 72.8 (Min: 56; Max: 86). The CFS 4–5 patient group had significantly increased mortality when compared to patients with CFS 1–3 [(RE) OR 1.95 (1.32 (95% CI), 2.87 (95% CI)); I2 81%; p = 0.0008]. Furthermore, CFS 6–9 patient group displayed an even more noticeable mortality increase when compared to patients with CFS 1–3 [(RE) OR 3.09 (2.03, 4.71); I2 87%; p<0.0001]. Generic inverse variance analysis of adjusted hazard ratio among included studies highlighted that CFS (p = 0.0001), male gender (p = 0.0009), National Early Warning Score (p = 0.0001), Ischaemic Heart Disease (IHD) (p = 0.07), Hypertension (HT) (p<0.0001), and Chronic Kidney Disease (CKD) (p = 0.0009) were associated with increased COVID-19 mortality.InterpretationOur findings suggest a differential stratification of CFS scores in the context of COVID-19 infection, in which CFS 1–3 patients may be considered at lower risk, CFS 4–5 at moderate risk, and CFS 6–9 at high risk of mortality regardless of age. Overall, our study not only aims to alert clinicians of the value of CFS scores, but also highlight the multiple dimensions to consider such as age, gender and co-morbidities, even among moderately frail patients in relation to COVID-19 mortality.FundingNone.

Effect of in-hospital glycemic variability on mortality in patients with diabetes

IntroductionThe aim of this study was to evaluate the impact of mean blood glucose (MBG) and glycaemic variability (GV) during hospitalisation on mortality after discharge. Material and methodsWe conducted a retrospective longitudinal analytical cohort study that included patients discharged form a department of internal medicine with a diabetes-related diagnosis. The evaluated prognosis was mortality. During hospitalisation, the patients’ clinical, laboratory and glycaemic control-related variables were recorded (MBG, GV and hypoglycaemia). The GV was measured with the coefficient of variation (CV).We calculated the mortality rates for every 1000 patient-years and compared them with Kaplan–Meier curves. We determined the predictors of mortality by performing a Cox regression. ResultsThe study included 276 patients with a mean age of 77.6 (SD, 10.2) years. The median outpatient follow-up duration was 2.7 years.In the multivariate analysis, an MBG > 140 mg/dL (HR = 1.72; 95% CI 1.14–2.61; p = .01) and a CV > 0.29 (HR = 1.52; 95% CI 1.12–2.06; p = .006), but not the presence of hypoglycaemia, were additively and independently associated with an increased risk of mortality. An MG > 140 mg/dL with a CV > 0.29 increased the mortality rates (123 vs. 317 per 1000 patient-year; p < .001) and the adjusted mortality risk (HR = 2.70; 95% CI 1.71–4.27; p < .001) compared with having an MBG ≤ 140 mg/dL. ConclusionThe simultaneous presence of a high MBG level and CV constitutes a powerful tool for stratifying mortality risk after hospital discharge.

Predictive Value of Rotterdam Score and Marshall Score in Traumatic Brain Injury: A Contemporary Review

AbstractThis article conducts a contemporary comparative review of the medical literature to update and establish evidence as to which framework among Rotterdam and Marshall computed tomography (CT)-based scoring systems predicts traumatic brain injury (TBI) outcomes better. The scheme followed was following the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines for literature search. The search started on August 15, 2020 and ended on December 31, 2020. The combination terms used were Medical Subject Headings terms, combination keywords, and specific words used for describing various pathologies of TBI to identify the most relevant article in each database. PICO question to guide the search strategy was: “what is the use of Marshall (I) versus Rotterdam score (C) in TBI patients (P) for mortality risk stratification (O).” The review is based on 46 references which included a full review of 14 articles for adult TBI patients and 6 articles for pediatric TBI articles comparing Rotterdam and Marshall CT scores. The review includes 8,243 patients, of which 2,365 were pediatric and 5,878 were adult TBI patients. Marshall CT classification is not ordinal, is more descriptive, has better inter-rater reliability, and poor performance in a specific group of TBI patients requiring decompressive craniectomy. Rotterdam CT classification is ordinal, has better discriminatory power, and a better description of the dynamics of intracranial changes. The two scoring systems are complimentary. A combination of clinical parameters, severity, ischemic and hemodynamic parameters, and CT scoring system could predict the prognosis of TBI patients with significant accuracy. None of the classifications has good evidence for use in pediatric patients.

Lipid and Lipoprotein Dysregulation in Sepsis: Clinical and Mechanistic Insights into Chronic Critical Illness.

In addition to their well-characterized roles in metabolism, lipids and lipoproteins have pleiotropic effects on the innate immune system. These undergo clinically relevant alterations during sepsis and acute inflammatory responses. High-density lipoprotein (HDL) plays an important role in regulating the immune response by clearing bacterial toxins, supporting corticosteroid release, decreasing platelet aggregation, inhibiting endothelial cell apoptosis, reducing the monocyte inflammatory response, and inhibiting expression of endothelial cell adhesion molecules. It undergoes quantitative as well as qualitative changes which can be measured using the HDL inflammatory index (HII). Pro-inflammatory, or dysfunctional HDL (dysHDL) lacks the ability to perform these functions, and we have also found it to independently predict adverse outcomes and organ failure in sepsis. Another important class of lipids known as specialized pro-resolving mediators (SPMs) positively affect the escalation and resolution of inflammation in a temporal fashion. These undergo phenotypic changes in sepsis and differ significantly between survivors and non-survivors. Certain subsets of sepsis survivors go on to have perilous post-hospitalization courses where this inflammation continues in a low grade fashion. This is associated with immunosuppression in a syndrome of persistent inflammation, immunosuppression, and catabolism syndrome (PICS). The continuous release of tissue damage-related patterns and viral reactivation secondary to immunosuppression feed this chronic cycle of inflammation. Animal data indicate that dysregulation of endogenous lipids and SPMs play important roles in this process. Lipids and their associated pathways have been the target of many clinical trials in recent years which have not shown mortality benefit. These results are limited by patient heterogeneity and poor animal models. Considerations of sepsis phenotypes and novel biomarkers in future trials are important factors to be considered in future research. Further characterization of lipid dysregulation and chronic inflammation during sepsis will aid mortality risk stratification, detection of sepsis, and inform individualized pharmacologic therapies.

Long-term mortality risk stratification of liver transplant recipients: real-time application of deep learning algorithms on longitudinal data

Survival of liver transplant recipients beyond 1 year since transplantation is compromised by an increased risk of cancer, cardiovascular events, infection, and graft failure. Few clinical tools are available to identify patients at risk of these complications, which would flag them for screening tests and potentially life-saving interventions. In this retrospective analysis, we aimed to assess the ability of deep learning algorithms of longitudinal data from two prospective cohorts to predict complications resulting in death after liver transplantation over multiple timeframes, compared with logistic regression models. In this machine learning analysis, model development was done on a set of 42 146 liver transplant recipients (mean age 48·6 years [SD 17·3]; 17 196 [40·8%] women) from the Scientific Registry of Transplant Recipients (SRTR) in the USA. Transferability of the model was further evaluated by fine-tuning on a dataset from the University Health Network (UHN) in Canada (n=3269; mean age 52·5 years [11·1]; 1079 [33·0%] women). The primary outcome was cause of death, as recorded in the databases, due to cardiovascular causes, infection, graft failure, or cancer, within 1 year and 5 years of each follow-up examination after transplantation. We compared the performance of four deep learning models against logistic regression, assessing performance using the area under the receiver operating characteristic curve (AUROC). In both datasets, deep learning models outperformed logistic regression, with the Transformer model achieving the highest AUROCs in both datasets (p<0·0001). The AUROC for the Transformer model across all outcomes in the SRTR dataset was 0·804 (99% CI 0·795-0·854) for 1-year predictions and 0·733 (0·729-0·769) for 5-year predictions. In the UHN dataset, the AUROC for the top-performing deep learning model was 0·807 (0·795-0·842) for 1-year predictions and 0·722 (0·705-0·764) for 5-year predictions. AUROCs ranged from 0·695 (0·680-0·713) for prediction of death from infection within 5 years to 0·859 (0·847-0·871) for prediction of death by graft failure within 1 year. Deep learning algorithms can incorporate longitudinal information to continuously predict long-term outcomes after liver transplantation, outperforming logistic regression models. Physicians could use these algorithms at routine follow-up visits to identify liver transplant recipients at risk for adverse outcomes and prevent these complications by modifying management based on ranked features. Canadian Donation and Transplant Research Program, CIFAR AI Chairs Program.

The Range of Cardiogenic Shock Survival by Clinical Stage: Data From the Critical Care Cardiology Trials Network Registry.

Cardiogenic shock presents with variable severity. Categorizing cardiogenic shock into clinical stages may improve risk stratification and patient selection for therapies. We sought to determine whether a structured implementation of the 2019 Society for Cardiovascular Angiography and Interventions clinical cardiogenic shock staging criteria that is ascertainable in clinical registries discriminates mortality in a contemporary population with or at-risk for cardiogenic shock. We developed a pragmatic application of the Society for Cardiovascular Angiography and Interventions cardiogenic shock staging criteria-A (at-risk), B (beginning), C (classic cardiogenic shock), D (deteriorating), or E (extremis)-and examined outcomes by stage. The Critical Care Cardiology Trials Network is an investigator-initiated multicenter research collaboration coordinated by the TIMI Study Group (Boston, MA). Consecutive admissions with or at-risk for cardiogenic shock during two annual 2-month collection periods (2017-2019) were analyzed. Patients with or at-risk for cardiogenic shock. Of 8,240 CICU admissions reviewed, 1,991 (24%) had or were at-risk for cardiogenic shock. Distributions across the five stages were as follows: A: 33%; B: 7%; C: 16%; D: 23%; and E: 21%. Overall in-hospital mortality among patients with established cardiogenic shock was 39%; however, mortality varied from only 15.8% to 32.1% to 62.5% across stages C, D, and E (Cochran-Armitage ptrend < 0.0001). The Society for Cardiovascular Angiography and Interventions stages improved mortality prediction beyond the Sequential Organ Failure Assessment and Intra-Aortic Balloon Pumpin Cardiogenic Shock II scores. Although overall mortality in cardiogenic shock remains high, it varies considerably based on clinical stage, identifying stage C as relatively lower risk. We demonstrate a pragmatic adaptation of the Society for Cardiovascular Angiography and Interventions cardiogenic shock stages that effectively stratifies mortality risk and could be leveraged for future clinical research.

Community-acquired pneumonia severity assessment tools in patients hospitalized with COVID-19: a validation and clinical applicability study

ObjectiveTo externally validate community-acquired pneumonia (CAP) tools on patients hospitalized with coronavirus disease 2019 (COVID-19) pneumonia from two distinct countries, and compare their performance with recently developed COVID-19 mortality risk stratification tools. MethodsWe evaluated 11 risk stratification scores in a binational retrospective cohort of patients hospitalized with COVID-19 pneumonia in São Paulo and Barcelona: Pneumonia Severity Index (PSI), CURB, CURB-65, qSOFA, Infectious Disease Society of America and American Thoracic Society Minor Criteria, REA-ICU, SCAP, SMART-COP, CALL, COVID GRAM and 4C. The primary and secondary outcomes were 30-day in-hospital mortality and 7-day intensive care unit (ICU) admission, respectively. We compared their predictive performance using the area under the receiver operating characteristics curve (AUC), sensitivity, specificity, likelihood ratios, calibration plots and decision curve analysis. ResultsOf 1363 patients, the mean (SD) age was 61 (16) years. The 30-day in-hospital mortality rate was 24.6% (228/925) in São Paulo and 21.0% (92/438) in Barcelona. For in-hospital mortality, we found higher AUCs for PSI (0.79, 95% CI 0.77–0.82), 4C (0.78, 95% CI 0.75–0.81), COVID GRAM (0.77, 95% CI 0.75–0.80) and CURB-65 (0.74, 95% CI 0.72–0.77). Results were similar for both countries. For the 1%–20% threshold range in decision curve analysis, PSI would avoid a higher number of unnecessary interventions, followed by the 4C score. All scores had poor performance (AUC <0.65) for 7-day ICU admission. ConclusionsRecent clinical COVID-19 assessment scores had comparable performance to standard pneumonia prognostic tools. Because it is expected that new scores outperform older ones during development, external validation studies are needed before recommending their use.

Risk Prediction in Patients With Heart Failure With Preserved Ejection Fraction Using Gene Expression Data and Machine Learning.

Heart failure with preserved ejection fraction (HFpEF) has become a major health issue because of its high mortality, high heterogeneity, and poor prognosis. Using genomic data to classify patients into different risk groups is a promising method to facilitate the identification of high-risk groups for further precision treatment. Here, we applied six machine learning models, namely kernel partial least squares with the genetic algorithm (GA-KPLS), the least absolute shrinkage and selection operator (LASSO), random forest, ridge regression, support vector machine, and the conventional logistic regression model, to predict HFpEF risk and to identify subgroups at high risk of death based on gene expression data. The model performance was evaluated using various criteria. Our analysis was focused on 149 HFpEF patients from the Framingham Heart Study cohort who were classified into good-outcome and poor-outcome groups based on their 3-year survival outcome. The results showed that the GA-KPLS model exhibited the best performance in predicting patient risk. We further identified 116 differentially expressed genes (DEGs) between the two groups, thus providing novel therapeutic targets for HFpEF. Additionally, the DEGs were enriched in Gene Ontology terms and Kyoto Encyclopedia of Genes and Genomes pathways related to HFpEF. The GA-KPLS-based HFpEF model is a powerful method for risk stratification of 3-year mortality in HFpEF patients.

Prognosis of Severe Low-Flow, Low-Gradient Aortic Stenosis by Stroke Volume Index and Transvalvular Flow Rate

ObjectivesThis study determined whether flow state classified by stroke volume index (SVi) or transvalvular flow rate (FR) improved risk stratification of all-cause mortality, hospitalization due to heart failure, and aortic valvular interventions for patients with severe aortic stenosis (AS). BackgroundSVi is a widely accepted classification for flow state in severe low-flow, low-gradient (LFLG) AS. Recent studies suggest that FR more closely approximates true AS severity and provides more useful prognostication than SVi. MethodsPatients with severe AS over a 7-year period were subclassified by echocardiographic parameters. LFLG-AS was defined as severe AS (aortic valve area index [AVAi]: <0.6 cm2/m2), with a mean transvalvular pressure gradient of <40 mm Hg in the setting of low flow state: SVi of <35 ml/m2 and/or FR of <200 ml/s and subclassified into preserved (≥50%; paradoxical) or reduced (<50%; classical) left ventricular ejection fraction (LVEF). ResultsAmong 621 consecutive patients with severe AS, the proportions of patients classified as LFLG-AS were different between SVi and FR (p < 0.001). Classification using SVi, FR, and LVEF was a strong predictor of the composite endpoint at the 2-year follow-up. The addition of SVi to the echocardiographic and clinical model provided significant improvement in reclassification (net reclassification improvement: 0.089; 95% confidence interval [CI]: 0.045 to 0.133; p = 0.04), whereas addition of FR did not (net reclassification improvement: 0.061; 95% CI: 0.016 to 0.106; p = 0.17). C-statistics indicated improved risk discrimination when AVAi, LVEF, and SVi or FR were added as predictive variables to the clinical model (p = 0.006). ConclusionsLow SVi or FR was associated with adverse cardiovascular events and showed improvement in discrimination, but only SVi, not FR, significantly improved risk reclassification compared to other conventional clinical and echocardiographic predictors. This suggests that FR is not superior to SVi in distinguishing true severe from pseudosevere forms of AS and identification of patients with LFLG-AS who have worse outcomes.