Risk Prediction Model For Mortality Research Articles

Generalizability challenges of mortality risk prediction models: A retrospective analysis on a multi-center database.

Modern predictive models require large amounts of data for training and evaluation, absence of which may result in models that are specific to certain locations, populations in them and clinical practices. Yet, best practices for clinical risk prediction models have not yet considered such challenges to generalizability. Here we ask whether population- and group-level performance of mortality prediction models vary significantly when applied to hospitals or geographies different from the ones in which they are developed. Further, what characteristics of the datasets explain the performance variation? In this multi-center cross-sectional study, we analyzed electronic health records from 179 hospitals across the US with 70,126 hospitalizations from 2014 to 2015. Generalization gap, defined as difference between model performance metrics across hospitals, is computed for area under the receiver operating characteristic curve (AUC) and calibration slope. To assess model performance by the race variable, we report differences in false negative rates across groups. Data were also analyzed using a causal discovery algorithm "Fast Causal Inference" that infers paths of causal influence while identifying potential influences associated with unmeasured variables. When transferring models across hospitals, AUC at the test hospital ranged from 0.777 to 0.832 (1st-3rd quartile or IQR; median 0.801); calibration slope from 0.725 to 0.983 (IQR; median 0.853); and disparity in false negative rates from 0.046 to 0.168 (IQR; median 0.092). Distribution of all variable types (demography, vitals, and labs) differed significantly across hospitals and regions. The race variable also mediated differences in the relationship between clinical variables and mortality, by hospital/region. In conclusion, group-level performance should be assessed during generalizability checks to identify potential harms to the groups. Moreover, for developing methods to improve model performance in new environments, a better understanding and documentation of provenance of data and health processes are needed to identify and mitigate sources of variation.

Machine Learning-Based Personalized Risk Prediction Model for Mortality of Patients Undergoing Mitral Valve Surgery: The PRIME Score.

BackgroundMitral valve surgery (MVS) is an effective treatment for mitral valve diseases. There is a lack of reliable personalized risk prediction models for mortality in patients undergoing mitral valve surgery. Our aim was to develop a risk stratification system to predict all-cause mortality in patients after mitral valve surgery.MethodsDifferent machine learning models for the prediction of all-cause mortality were trained on a derivation cohort of 1,883 patients undergoing mitral valve surgery [split into a training cohort (70%) and internal validation cohort (30%)] to predict all-cause mortality. Forty-five clinical variables routinely evaluated at discharge were used to train the models. The best performance model (PRIME score) was tested in an externally validated cohort of 220 patients undergoing mitral valve surgery. The model performance was evaluated according to the area under the curve (AUC). Net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were compared with existing risk strategies.ResultsAfter a median follow-up of 2 years, there were 133 (7.063%) deaths in the derivation cohort and 17 (7.727%) deaths in the validation cohort. The PRIME score showed an AUC of 0.902 (95% confidence interval [CI], 0.849–0.956) in the internal validation cohort and 0.873 (95% CI: 0.769–0.977) in the external validation cohort. In the external validation cohort, the performance of the PRIME score was significantly improved compared with that of the existing EuroSCORE II (NRI = 0.550, [95% CI 0.001–1.099], P = 0.049, IDI = 0.485, [95% CI 0.230–0.741], P < 0.001).ConclusionMachine learning-based model (the PRIME score) that integrate clinical, demographic, imaging, and laboratory features demonstrated superior performance for the prediction of mortality patients after mitral valve surgery compared with the traditional risk model EuroSCORE II.Clinical Trial Registration[http://www.clinicaltrials.gov], identifier [NCT05141292].

Efficacy of the P-POSSUM scoring system in prediction of post-operative mortality and morbidity in patients undergoing emergency exploratory laparotomy in a tertiary institute in East India

Background: The tenet of any surgery is to reduce the morbidity and mortality. The outcome is not solely dependent on surgeon’s operative expertise alone, but also significantly influencedby a lot of patient factors. The Physiological and Operative Severity Score for the enUmeration of Morbidity and Mortality (POSSUM), a risk-adjusted scoring system, has already been used in various parts of India. There is scarcity of published research using this tool in the population of Eastern India. Aims and Objectives: The aim of the study was to determine the efficacy of modified version Portsmouth-POSSUM (P-POSSUM) scoring system in predicting the postoperative morbidity and mortality in population of East India. Materials and Methods: The study was carried out in 50 patients of either gender, of age group 13–80 years, undergoing emergency laparotomy and followed up to 30-day post-operative follow-up of all patients till the sample size of 50 was reached. The physiological scores were recorded during admission and the operative severity score was recorded according to the intra-operative findings of operating surgeon. Any post-operative morbidity or death was recorded during hospital stay or on follow-up for a period of 30 days. The Hosmer–Lemeshow test was used for goodness-of-fit in logistics regression for mortality risk prediction model. The ratio of observed and expected rates (O: E ratio) was calculated. Results: Using the scoring tool, no significant difference was observed between the expected and observed mortality (Chi-square 2.166; df=8; P=0.976, O: E ratio 1.4) but a considerable difference was observed between the observed between the expected morbidity rates (χ²=22.806, df=8, P=0.004, O: E ratio 1.137). Conclusion: The P-POSSUM score was able to accurately predict the mortality in Eastern India surgical settings. However, the model was not that accurate in prediction of morbidity.

Risk prediction models incorporating institutional case volume for mortality after hip fracture surgery in the elderly.

While higher institutional case volume is associated with better postoperative outcomes in various types of surgery, institutional case volume has been rarely included in risk prediction models for surgical patients. This study aimed to develop and validate the predictive models incorporating institutional case volume for predicting in-hospital mortality and 1-year mortality after hip fracture surgery in the elderly. Data for all patients (≥ 60years) who underwent surgery for femur neck fracture, pertrochanteric fracture, or subtrochanteric fracture between January 2008 and December 2016 were extracted from the Korean National Health Insurance Service database. Patients were randomly assigned into the derivation cohort or the validation cohort in a 1:1 ratio. Risk prediction models for in-hospital mortality and 1-year mortality were developed in the derivation cohort using the logistic regression model. Covariates included age, sex, type of fracture, type of anaesthesia, transfusion, and comorbidities such as hypertension, diabetes, coronary artery disease, chronic kidney disease, cerebrovascular disease, and dementia. Two separate models, one with and the other without institutional case volume as a covariate, were constructed, evaluated, and compared using the likelihood ratio test. Based on the models, scoring systems for predicting in-hospital mortality and 1-year mortality were developed. Analysis of 196,842 patients showed 3.6% in-hospital mortality (7084/196,842) and 15.42% 1-year mortality (30,345/196,842). The model for predicting in-hospital mortality incorporating the institutional case volume demonstrated better discrimination (c-statistics 0.692) compared to the model without the institutional case volume (c-statistics 0.688; likelihood ratio test p value < 0.001). The performance of the model for predicting 1-year mortality was also better when incorporating institutional case volume (c-statistics 0.675 vs. 0.674; likelihood ratio test p value < 0.001). The new institutional case volume incorporated scoring system may help to predict in-hospital mortality and 1-year mortality after hip fracture surgery in the elderly population.

Derivation and Validation of a Machine Learning-Based Risk Prediction Model for In-Hospital Mortality in Patients with Acute Heart Failure

Derivation and Validation of a Machine Learning-Based Risk Prediction Model for In-Hospital Mortality in Patients with Acute Heart Failure

Clinical risk prediction model for 30-day all-cause re-hospitalisation or mortality in patients hospitalised with heart failure

BackgroundThis study aimed to develop a risk prediction model (AUS-HF model) for 30-day all-cause re-hospitalisation or death among patients admitted with acute heart failure (HF) to inform follow-up after hospitalisation. The model uses routinely collected measures at point of care. MethodsWe analyzed pooled individual-level data from two cohort studies on acute HF patients followed for 30-days after discharge in 17 hospitals in Victoria, Australia (2014–2017). A set of 58 candidate predictors, commonly recorded in electronic medical records (EMR) including demographic, medical and social measures were considered. We used backward stepwise selection and LASSO for model development, bootstrap for internal validation, C-statistic for discrimination, and calibration slopes and plots for model calibration. ResultsThe analysis included 1380 patients, 42.1% female, median age 78.7 years (interquartile range = 16.2), 60.0% experienced previous hospitalisation for HF and 333 (24.1%) were re-hospitalised or died within 30 days post-discharge. The final risk model included 10 variables (admission: eGFR, and prescription of anticoagulants and thiazide diuretics; discharge: length of stay>3 days, systolic BP, heart rate, sodium level (<135 mmol/L), >10 prescribed medications, prescription of angiotensin converting enzyme inhibitors or angiotensin receptor blockers, and anticoagulants prescription. The discrimination of the model was moderate (C-statistic = 0.684, 95%CI 0.653, 0.716; optimism estimate = 0.062) with good calibration. ConclusionsThe AUS-HF model incorporating routinely collected point-of-care data from EMRs enables real-time risk estimation and can be easily implemented by clinicians. It can predict with moderate accuracy risk of 30-day hospitalisation or mortality and inform decisions around the intensity of follow-up after hospital discharge.

Mortality prediction in intensive care units including premorbid functional status improved performance and internal validity

ObjectivePrognostic models are key for benchmarking intensive care units (ICUs). They require up–to–date predictors and should report transportability properties for reliable predictions. We developed and validated an in–hospital mortality risk prediction model to facilitate benchmarking, quality assurance, and health economics evaluation. Study Design and SettingWe retrieved data from the database of an international (Finland, Estonia, Switzerland) multicenter ICU cohort study from 2015 to 2017. We used a hierarchical logistic regression model that included age, a modified Simplified Acute Physiology Score–II, admission type, premorbid functional status, and diagnosis as grouping variable. We used pooled and meta–analytic cross–validation approaches to assess temporal and geographical transportability. ResultsWe included 61,224 patients treated in the ICU (hospital mortality 10.6%). The developed prediction model had an area under the receiver operating characteristic curve 0.886, 95% confidence interval (CI) 0.882–0.890; a calibration slope 1.01, 95% CI (0.99–1.03); a mean calibration –0.004, 95% CI (–0.035 to 0.027). Although the model showed very good internal validity and geographic discrimination transportability, we found substantial heterogeneity of performance measures between ICUs (I–squared: 53.4–84.7%). ConclusionA novel framework evaluating the performance of our prediction model provided key information to judge the validity of our model and its adaptation for future use.

Abstract 9202: Sudden Cardiac Death and Icd Use in Rasopathy-Associated Hypertrophic Cardiomyopathy: An International Multi-Centre Study

Introduction: Sudden cardiac death (SCD) burden in RASopathy patients (pts) with hypertrophic cardiomyopathy (RAS-HCM) is not known. We compared SCD events and transplant-free survival in RAS versus primary HCM (P-HCM) pts. Methods: The international 22-center retrospective cohort study (PRIMaCY) included 190 pediatric RAS-HCM and 568 P-HCM pts. The primary outcome was 10-year freedom from SCD events (composite of SCD, resuscitated sudden cardiac arrest, and appropriate shock from a prophylactic ICD), estimated using Kaplan Meier method. Results: Of 190 RAS-HCM pts, 82% had Noonan syndrome. Age at diagnosis was lower in RAS-HCM vs P-HCM (0.9 vs 9.7 yrs, p&lt;0.001). There was no difference in baseline LV posterior wall or septal thickness, but LVOT gradient was higher in RAS-HCM vs P-HCM (30 vs 13mmHg, p&lt;0.001). However, frequency of myectomy was not different in RAS-HCM vs P-HCM (15.2% vs 16.7%, p=0.55). Eighteen RAS-HCM pts died (4 SCD, 2 bradycardic arrests, 5 heart failure, 3 respiratory failure, 2 other) with lower 10-year transplant-free survival in RAS-HCM vs P-HCM (84.2% vs 91.6%, p=0.026). Although 10-year cumulative proportion of SCD events was lower in RAS-HCM than P-HCM (5.4% vs 15.1% respectively, p&lt;0.001) (Figure 1), the median age at event was not different (12yrs in RAS-HCM vs 15.2yrs in P-HCM). Of 5 RAS-HCM pts with a SCD event, 4 died and 1 received a secondary prevention ICD. An additional 10 pts received a prophylactic ICD with no appropriate shock during follow-up. The 10-year cumulative proportion of prophylactic ICD was lower at 6.3% in RAS-HCM vs 27.2% in P-HCM (p&lt;0.001). Conclusion: Transplant-free survival was lower in RAS-HCM than P-HCM with mortality related primarily to non-SCD events. Of the RAS-HCM patients with SCD events, 4 died without an ICD highlighting the need for better risk prediction models for both SCD and heart failure mortality in this population.

A Review of Mortality Risk Prediction Models in Smartphone Applications.

Healthcare professionals in healthcare systems need access to freely available, real-time, evidence-based mortality risk prediction smartphone applications to facilitate resource allocation. The objective of this study is to evaluate the quality of smartphone mobile health applications that include mortality prediction models, and corresponding information quality. We conducted a systematic review of commercially available smartphone applications in Google Play for Android, and iTunes for iOS smartphone applications. We performed initial screening, data extraction, and rated smartphone application quality using the Mobile Application Rating Scale: user version (uMARS). The information quality of smartphone applications was evaluated using two patient vignettes, representing low and high risk of mortality, based on critical care data from the Medical Information Mart for Intensive Care (MIMIC) III database. Out of 3051 evaluated smartphone applications, 33 met our final inclusion criteria. We identified 21 discrete mortality risk prediction models in smartphone applications. The most common mortality predicting models were Sequential Organ Failure Assessment (SOFA) (n = 15) and Acute Physiology and Clinical Health Assessment II (n = 13). The smartphone applications with the highest quality uMARS scores were Observation—NEWS 2 (4.64) for iOS smartphones, and MDCalc Medical Calculator (4.75) for Android smartphones. All SOFA-based smartphone applications provided consistent information quality with the original SOFA model for both the low and high-risk patient vignettes. We identified freely available, high-quality mortality risk prediction smartphone applications that can be used by healthcare professionals to make evidence-based decisions in critical care environments.

Developing a Prediction Model for 7-Year and 10-Year All-Cause Mortality Risk in Type 2 Diabetes Using a Hospital-Based Prospective Cohort Study.

Leveraging easily accessible data from hospitals to identify high-risk mortality rates for clinical diabetes care adjustment is a convenient method for the future of precision healthcare. We aimed to develop risk prediction models for all-cause mortality based on 7-year and 10-year follow-ups for type 2 diabetes. A total of Taiwanese subjects aged ≥18 with outpatient data were ascertained during 2007–2013 and followed up to the end of 2016 using a hospital-based prospective cohort. Both traditional model selection with stepwise approach and LASSO method were conducted for parsimonious models’ selection and comparison. Multivariable Cox regression was performed for selected variables, and a time-dependent ROC curve with an integrated AUC and cumulative mortality by risk score levels was employed to evaluate the time-related predictive performance. The prediction model, which was composed of eight influential variables (age, sex, history of cancers, history of hypertension, antihyperlipidemic drug use, HbA1c level, creatinine level, and the LDL /HDL ratio), was the same for the 7-year and 10-year models. Harrell’s C-statistic was 0.7955 and 0.7775, and the integrated AUCs were 0.8136 and 0.8045 for the 7-year and 10-year models, respectively. The predictive performance of the AUCs was consistent with time. Our study developed and validated all-cause mortality prediction models with 7-year and 10-year follow-ups that were composed of the same contributing factors, though the model with 10-year follow-up had slightly greater risk coefficients. Both prediction models were consistent with time.

1657 Systematic Review and Meta-Analysis of Mortality Risk Prediction Models in Adult Cardiac Surgery

Abstract Background The most used mortality risk prediction models in cardiac surgery are the European System for Cardiac Operative Risk Evaluation(EuroSCORE)(ES) and Society of Thoracic Surgeons(STS) score. There is no agreement on which score should be considered more accurate nor which score should be utilised in each population sub-group. We sought to provide a thorough quantitative assessment of these 2 models. Method We performed a systematic literature review and captured information on discrimination, as quantified by the area under the receiver operator curve(AUC), and calibration, as quantified by the ratio of observed-to-expected mortality(O:E). We performed random effects meta-analysis of the performance of the individual models as well as pairwise comparisons and sub-group analysis by procedure type, time and continent. Results The ES2(AUC 0.783[95%CI 0.765-0.800];O:E 1.102[95%CI 0.943-1.289]) and STS(AUC 0.757[95%CI 0.727-0.785];O:E 1.111[95%CI 0.853-1.447]) both showed good overall discrimination and calibration. There was no significant difference in the discrimination of the two models(Difference in AUC -0.016; 95%CI -0.034 to -0.002;p0.09). However, the calibration of ES2 showed significant geographical variations(p &amp;lt; 0.001) and a trend towards miscalibration with time(p0.0057). This was not seen with STS. Conclusions ES2 and STS are both reliable predictors of short-term mortality following adult cardiac surgery in the populations from which they were derived. STS may have broader applications when comparing outcomes across continents and time periods as compared to ES2.

Mortality-Risk Prediction Model from Road-Traffic Injury in Drunk Drivers: Machine Learning Approach.

Background: Alcohol-related road-traffic injury is the leading cause of premature death in middle- and lower-income countries, including Thailand. Applying machine-learning algorithms can improve the effectiveness of driver-impairment screening strategies by legal limits. Methods: Using 4794 RTI drivers from secondary cross-sectional data from the Thai Governmental Road Safety Evaluation project in 2002–2004, the machine-learning models (Gradient Boosting Classifier: GBC, Multi-Layers Perceptrons: MLP, Random Forest: RF, K-Nearest Neighbor: KNN) and a parsimonious logistic regression (Logit) were developed for predicting the mortality risk from road-traffic injury in drunk drivers. The predictors included alcohol concentration level in blood or breath, driver characteristics and environmental factors. Results: Of 4974 drivers in the derived dataset, 4365 (92%) were surviving drivers and 429 (8%) were dead drivers. The class imbalance was rebalanced by the Synthetic Minority Oversampling Technique (SMOTE) into a 1:1 ratio. All models obtained good-to-excellent discrimination performance. The AUC of GBC, RF, KNN, MLP, and Logit models were 0.95 (95% CI 0.90 to 1.00), 0.92 (95% CI 0.87 to 0.97), 0.86 (95% CI 0.83 to 0.89), 0.83 (95% CI 0.78 to 0.88), and 0.81 (95% CI 0.75 to 0.87), respectively. MLP and GBC also had a good model calibration, visualized by the calibration plot. Conclusions: Our machine-learning models can predict road-traffic mortality risk with good model discrimination and calibration. External validation using current data is recommended for future implementation.

Financial Estimation of the Uncertainty in Medicine Using Present Value of Medical Fees and a Mortality Risk Prediction Model: a Retrospective Cohort Study.

This study aimed to develop a method to enable the financial estimation of each patient’s uncertainty without focusing on healthcare technology. We define financial uncertainty (FU) as the difference between an actual amount of claim (AC) and the discounted present value of the AC (DAC). DAC can be calculated based on a discounted present value calculated using a cash flow, a period of investment, and a discount rate. The present study considered these three items as AC, the length of hospital stay, and the predicted mortality rate. The mortality prediction model was built using typical data items in standard level electronic medical records such as sex, age, and disease information. The performance of the prediction model was moderate because an area under curve was approximately 85%. The empirical analysis primarily compares the FU of the top 20 diseases with the actual AC using a retrospective cohort in the University of Miyazaki Hospital. The observational period is 5 years, from April 1, 2013, to March 31, 2018. The analysis demonstrates that the proportion of FU to actual AC is higher than 20% in low-weight children, patients with leukemia, brain tumor, myeloid leukemia, or non-Hodgkin’s lymphoma. For these diseases, patients cannot avoid long hospitalization; therefore, the medical fee payment system should be designed based on uncertainty. Our method is both practical and generalizable because it uses a small number of data items that are required in standard electronic medical records. This method contributes to the decision-making processes of health policymakers.

Development and Validation of a Multivariable Risk Prediction Model for COVID-19 Mortality in the Southern United States

ObjectiveTo evaluate clinical characteristics of patients admitted to the hospital with coronavirus disease 2019 (COVID-19) in Southern United States and development as well as validation of a mortality risk prediction model. Patients and MethodsSouthern Louisiana was an early hotspot during the pandemic, which provided a large collection of clinical data on inpatients with COVID-19. We designed a risk stratification model to assess the mortality risk for patients admitted to the hospital with COVID-19. Data from 1673 consecutive patients diagnosed with COVID-19 infection and hospitalized between March 1, 2020, and April 30, 2020, was used to create an 11-factor mortality risk model based on baseline comorbidity, organ injury, and laboratory results. The risk model was validated using a subsequent cohort of 2067 consecutive hospitalized patients admitted between June 1, 2020, and December 31, 2020. ResultsThe resultant model has an area under the curve of 0.783 (95% CI, 0.76 to 0.81), with an optimal sensitivity of 0.74 and specificity of 0.69 for predicting mortality. Validation of this model in a subsequent cohort of 2067 consecutively hospitalized patients yielded comparable prognostic performance. ConclusionWe have developed an easy-to-use, robust model for systematically evaluating patients presenting to acute care settings with COVID-19 infection.

Prediction of mortality in COVID-19 through combing CT severity score with NEWS, qSOFA, or peripheral perfusion index.

IntroductionThe assessment of disease severity and the prediction of clinical outcomes at early disease stages can contribute to decreased mortality in patients with Coronavirus disease 2019 (COVID-19). This study was conducted to develop and validate a multivariable risk prediction model for mortality with using a combination of computed tomography severity score (CT-SS), national early warning score (NEWS), and quick sequential (sepsis-related) organ failure assessment (qSOFA) in COVID-19 patients.MethodsWe retrospectively collected medical data from 655 adult COVID-19 patients admitted to our hospital between July and November 2020. Data on demographics, clinical characteristics, and laboratory and radiological findings measured as part of standard care at admission were used to calculate NEWS, qSOFA score, CT-SS, peripheral perfusion index (PPI) and shock index (SI). Logistic regression and Cox proportional hazard models were used to predict mortality, which was our primary outcome. The predictive accuracy of distinct scoring systems was evaluated by the receiver-operating characteristic (ROC) curve analysis.ResultsThe median age was 50.0 years [333 males (50.8%), 322 females (49.2%)]. Higher NEWS and SI was associated with time-to-death within 90-days, whereas higher age, CT-SS and lower PPI were significantly associated with time-to-death within both 14 days and 90 days in the adjusted Cox regression model. The CT-SS predicted different mortality risk levels within each stratum of NEWS and qSOFA and improved the discrimination of mortality prediction models. Combining CT-SS with NEWS score yielded more accurate 14 days (DBA: −0.048, p = 0.002) and 90 days (DBA: −0.066, p < 0.001) mortality prediction.ConclusionCombining severity tools such as CT-SS, NEWS and qSOFA improves the accuracy of predicting mortality in patients with COVID-19. Inclusion of these tools in decision strategies might provide early detection of high-risk groups, avoid delayed medical attention, and improve patient outcomes.

Impact of Intraoperative Data on Risk Prediction for Mortality After Intra-Abdominal Surgery.

Risk prediction models for postoperative mortality after intra-abdominal surgery have typically been developed using preoperative variables. It is unclear if intraoperative data add significant value to these risk prediction models. With IRB approval, an institutional retrospective cohort of intra-abdominal surgery patients in the 2005 to 2015 American College of Surgeons National Surgical Quality Improvement Program was identified. Intraoperative data were obtained from the electronic health record. The primary outcome was 30-day mortality. We evaluated the performance of machine learning algorithms to predict 30-day mortality using: 1) baseline variables and 2) baseline + intraoperative variables. Algorithms evaluated were: 1) logistic regression with elastic net selection, 2) random forest (RF), 3) gradient boosting machine (GBM), 4) support vector machine (SVM), and 5) convolutional neural networks (CNNs). Model performance was evaluated using the area under the receiver operator characteristic curve (AUROC). The sample was randomly divided into a training/testing split with 80%/20% probabilities. Repeated 10-fold cross-validation identified the optimal model hyperparameters in the training dataset for each model, which were then applied to the entire training dataset to train the model. Trained models were applied to the test cohort to evaluate model performance. Statistical significance was evaluated using P < .05. The training and testing cohorts contained 4322 and 1079 patients, respectively, with 62 (1.4%) and 15 (1.4%) experiencing 30-day mortality, respectively. When using only baseline variables to predict mortality, all algorithms except SVM (area under the receiver operator characteristic curve [AUROC], 0.83 [95% confidence interval {CI}, 0.69-0.97]) had AUROC >0.9: GBM (AUROC, 0.96 [0.94-1.0]), RF (AUROC, 0.96 [0.92-1.0]), CNN (AUROC, 0.96 [0.92-0.99]), and logistic regression (AUROC, 0.95 [0.91-0.99]). AUROC significantly increased with intraoperative variables with CNN (AUROC, 0.97 [0.96-0.99]; P = .047 versus baseline), but there was no improvement with GBM (AUROC, 0.97 [0.95-0.99]; P = .3 versus baseline), RF (AUROC, 0.96 [0.93-1.0]; P = .5 versus baseline), and logistic regression (AUROC, 0.94 [0.90-0.99]; P = .6 versus baseline). Postoperative mortality is predicted with excellent discrimination in intra-abdominal surgery patients using only preoperative variables in various machine learning algorithms. The addition of intraoperative data to preoperative data also resulted in models with excellent discrimination, but model performance did not improve.

Nomogram for the Prediction of Intrahospital Mortality Risk of Patients with ST-Segment Elevation Myocardial Infarction Complicated with Hyperuricemia: A Multicenter Retrospective Study.

PurposeThis study aimed to establish an accurate and easy predictive model for ST-segment elevation myocardial infarction (STEMI) patients with hyperuricemia, using readily available features to estimate intrahospital mortality risk.Patients and MethodsThis was a multicenter retrospective study involving the development of risk prediction models for intrahospital mortality among all STEMI patients with hyperuricemia from Zunyi Medical University Chest Pain Center’s specialized alliance between January 1, 2016 and June 30, 2020. The primary outcome was intrahospital mortality. A total of 48 candidate variables were considered from demographic and clinical data. The least absolute shrinkage and selection operator (LASSO) was used to develop a nomogram. Concordance index values, decision curve analysis, the area under the curve (AUC), and clinical impact curves were examined. In this study, 489 patients with STEMI were included in the training dataset and an additional 209 patients from the 44 chest pain centers were included in the test cohort. B-type natriuretic peptides, α-hydroxybutyrate dehydrogenase (α-HBDH), cystatin C, out-of-hospital cardiac arrest (OHCA), shock index, and neutrophil-to-lymphocyte ratio were associated with intrahospital mortality and included in the nomogram.ResultsThe model showed good discrimination power, and the AUC generated to predict survival in the training set was 0.875 (95% confidence interval, 0.825–0.925). In the validation set, the AUC of survival predictions was 0.87 (95% confidence interval, 0.792–0.947). Calibration plots and decision curve analysis showed good model performance in both datasets. A web-based calculator (https://bzxzmu.shinyapps.io/STEMI-with-Hyperuricemia-intrahospital-mortality/) was established based on the nomogram model, which was used to measure the levels of OHCA, neutrophil-to-lymphocyte ratio, shock index, α-HBDH, cystatin C, and B-type natriuretic peptides.ConclusionFor practical applications, this model may prove clinically useful for personalized therapy management in patients with STEMI with hyperuricemia.

Cardiorespiratory optimal point during exercise testing is related to cardiovascular and all-cause mortality.

Cardiorespiratory optimal point (COP) during exercise may be a potentially clinically useful cardiopulmonary exercise testing (CPET) variable, but its prognostic relevance for adverse cardiovascular disease (CVD) outcomes is unknown. We aimed to assess the association of COP during exercise with fatal mortality outcomes and the extent to which COP could improve the prediction of CVD mortality. Cardiorespiratory optimal point, the minimum value of the ventilatory equivalent for oxygen (VE/VO2) in a given minute of a CPET, was defined in 2,205men who underwent CPET. Hazard ratios (HRs) (95% confidence intervals [CIs]) for outcomes and measures of risk discrimination for CVD mortality were calculated. During a median follow-up of 28.8years, 402 fatal CHDs, 607 fatal CVDs, and 1,348 all-cause mortality events occurred. COP was continually associated with each outcome in a dose-response manner. On adjustment for established and emerging risk factors, the HRs (95% CIs) for fatal CHD, fatal CVD, and all-cause mortality were 3.05 (1.94-4.81), 2.82 (1.91-4.18) and 2.46 (1.85-3.27), respectively, per standard deviation increase in COP. After further adjustment for high sensitivity C-reactive protein, the HRs were 2.82 (1.78-4.46), 2.57 (1.73-3.81), and 2.27 (1.70-3.02), respectively. Addition of COP to a CVD mortality risk prediction model containing established risk factors was associated with a C-index change of 0.0139 (0.0040 to 0.0238; p=0.006) at 25years. COP during exercise is directly associated with fatal cardiovascular and all-cause mortality events in dose-response fashions. COP during exercise may improve the prediction of the long-term risk for CVD mortality.

Percentage of Age-Predicted Cardiorespiratory Fitness Is Inversely Associated with Cardiovascular Disease Mortality: A Prospective Cohort Study

Introduction: Percentage of age-predicted cardiorespiratory fitness (% age-predicted CRF) is a potentially useful cardiopulmonary exercise testing (CPX) parameter, but there are limited data on its prognostic relevance for adverse cardiovascular disease (CVD) outcomes. We aimed to assess the association of % age-predicted CRF with CVD mortality and the extent to which % age-predicted CRF measurements could improve the prediction of CVD mortality. Methods: Peak oxygen uptake, used as the measure of CRF, was directly assessed in 2,276 men who underwent CPX. The age-predicted CRF estimated from a regression equation for age was transformed to % age-predicted CRF with the following formula: (achieved CRF/age-predicted CRF) × 100. Hazard ratios (HRs) (95% confidence intervals [CIs]) and measures of risk discrimination for CVD mortality were calculated. Results: During a median follow-up of 28.5 years, 643 fatal CVDs were recorded. The relationship between % age-predicted CRF and CVD mortality was dose response in nature. In analysis adjusted for conventional risk factors, one standard deviation increase in % age-predicted CRF was associated with a reduced risk of CVD mortality (HR 0.61; 95% CI: 0.56–0.67), which was minimally attenuated on further adjustment for several other confounders (HR 0.71; 95% CI: 0.64–0.78). Addition of % age-predicted CRF to a CVD mortality risk prediction model containing established risk factors significantly improved risk discrimination and reclassification. Conclusion: Percentage of age-predicted CRF is inversely and independently associated with CVD mortality in a graded fashion and significantly improves the prediction and classification of the long-term risk for CVD mortality beyond established risk factors.

Development and Validation of a Risk Prediction Model for Mortality Using Nomogram Among the Chinese Older Adults

Background: Rapid population ageing is a worldwide trend, highlighting an urgent need to have a reliable prediction tool for estimating a general older adult mortality risk. Previous predictive models have limited by the cause mortality, clinical indexes, participants as young as 50 years old, and western countries. We aimed to develop and validate a all-cause mortality prediction model based on multidimensional risk factors in Chinese older adults. Methods: In three prospective cohorts, we used the Chinese Longitudinal Healthy Longevity Survey (CLHLS) (mean age 88.08, n=37074) as the derivation cohort and the Healthy Ageing and Biomarkers Cohort Study (HABCS) (mean age 86.74, n=2552) and the China Health and Retirement Longitudinal Study (CHARLS) (mean age 72.48, n=4794) as validation cohorts. 42 risk factors were filtered by Lasso regression. The predictive factors of final models were decided by both improved net reclassification improvement (NRI) and public health significance. Cox proportional hazards models were established the mortality risk prediction equations. Model fit was evaluated by discrimination C-statistic. We conducted the 10x10 cross-validation technique to evaluate the internal consistency of the discrimination and calibration performance. Calibration plots were shown the results by comparing the predicted probabilities with the observed probabilities. The prediction ability of the equation was illustrated by using Nomogram. Findings: During 20-year follow-up, 28158 older adults (76%) died, leading to a total of 101429 person-years of observation in the CLHLS. The final model for all-cause mortality included age, sex, cognitive status, ADL status, current spouse status, past smoke status, frequency of doing housework, frequency of reading and watching TV, and frequency of doing garden work. The C statistic of derivation Cohort was 0.728 (96%CI, 0.724,0.732), and external validation were 0.761 (96%CI, 0.749,0.773) in HABCS and 0.713 (96%CI, 0.697,0.729) in CHARS. Interpretation: This study provides an objective, validated, and acceptable predictive factors to predict mortality risk in Chinese older adults. Our finding suggest that these predictive factors could be useful for policy, epidemiological and clinical applications. Funding Information: This work was supported by the fund of “San-ming” Project of Medicine in Shenzhen (NO.SZSM201812088) for data collection and analysis. Declaration of Interests: None. Ethics Approval Statement: The Protection of Human Subjects for the CLHLS, HACBS and CHARLS were approved by the biomedical ethics committee of Peking university. The written informed consent was provided by survey respondents before investigation.