Predicted Risk Of Death Research Articles

Chronic heart failure in intensive care unit: can we accurately predict the risk?

Abstract Background Severity of disease scoring systems, namely the Simplified Acute Physiology Score (SAPS) and Acute Physiology and Chronic Health Evaluation (APACHE), are widely used to predict mortality in Intensive Care Units (ICU). Yet, neither score includes chronic HF in their model. We aimed to evaluate whether these scores perform well in risk prediction of death of patients previously diagnosed with heart failure (HF). Methodology This is a single-center retrospective cohort of patients admitted to an ICU in 2019. Those whose admission lasted <24 hours were excluded from analysis. The SAPS II and APACHE II scores were calculated using data from the first 24 hours of ICU admission, imputing the worst variable obtained within this timeframe. HF was defined according to the ESC recommendations. In order to assess the performance of the scores, Receiver Operating Characteristic (ROC) Curves were used to predict the risk of death in ICU in HF compared to the non-HF population. Results A total of 267 patients were hospitalized in ICU for a period over 24 hours in 2019 (mean age 67±16 years; 58.8% males; 21.7% with chronic HF; 33.7% admitted for sepsis). Compared to patients without HF, those with chronic HF were older (74±13 vs. 65±16 years; p<0.001) and had higher risk scores (mean SAPS II: 43.2±21.7 vs. 56.5±20.7; p<0.001; mean APACHE II: 19.8±10.0 vs. 25.1±10.0; p<0.001). Moreover, these patients were at higher risk of meaningful events during hospitalization (e.g. acute kidney injury: 38.0 vs. 66.1%; p<0.001; shock at any time: 52.4 vs. 67.8%; p=0.036). Furthermore, patients with HF had a trend towards higher mortality rates in ICU (17.3 vs. 28.8%; p=0.051) and a significantly higher death in overall hospitalization (30.8 vs. 45.8%; p=0.032). ROC curves performed well in predicting the risk of ICU death regardless of HF (SAPS II – AUC 0.78 vs. 0.81; p=0.36; APACHE II – AUC 0.75 vs. 0.78; p=0.37). Conclusion Approximately 1 in every 4 patients admitted to the ICU had chronic HF. Traditional risk scoring systems (SAPS II and APACHE II) performed well regardless of HF. While these results are reassuring as far as risk stratification accuracy is concerned, HF patients remained at a higher risk for worse outcomes. Therefore, prognostic tools with a therapeutic clinical applicability are urgently needed to improve the outcome of this population. Funding Acknowledgement Type of funding sources: None.

Risk prediction of covid-19 related death and hospital admission in adults after covid-19 vaccination: national prospective cohort study

ObjectivesTo derive and validate risk prediction algorithms to estimate the risk of covid-19 related mortality and hospital admission in UK adults after one or two doses of covid-19 vaccination.DesignProspective, population...

Development and evaluation of an early death risk prediction model after acute type A aortic dissection.

BackgroundThe purpose of the study was to assess the relationship between preoperative laboratory examination, clinical imaging data, and postoperative death of patients with acute type A aortic dissection (ATAAD) and to establish a prediction model of hospital death risk after the operation.MethodsA total of 224 cases of acute Standford A aortic dissection were treated by total arch replacement using a tetrafurcate graft with stented elephant trunk implantation in Tianjin Chest Hospital. Based on preoperative laboratory examination and clinical imaging data of patients with ATAAD, the independent risk factors of postoperative hospital death were obtained using logistic analysis, and a risk prediction model of postoperative hospital death was developed.ResultsIndependent risk factors of postoperative death in patients with ATAAD were: body mass index (BMI), preoperative neutrophil to lymphocyte ratio (NLR), mean platelet volume (MPV), creatinine (Cr), D-dimer, high-sensitive cardiac troponin T (hs-CTnT), apolipoprotein A1, left subclavian artery involvement, and iliac artery involvement. The regression equation of postoperative death risk was: logitP1 = −9.584 + 1.060 × NLR + 1.586 × MPV + 1.009 × Cr + 1.067 × D-dimer + 2.023 × hs-CTnT; the regression equation of postoperative death risk was: logitP2 = −3.296 + 3.242 × left subclavian artery involved + 4.564 × iliac artery involved; the regression equation of postoperative death risk was: logitP3 = −12.864 + 1.149 × BMI + 4.731 × left subclavian artery involved + 4.150 × iliac artery involved + 1.064 × NLR + 1.011 × Cr + 1.084 × D-dimer + 2.242 × hs-CTnT + 3.233 × apolipoprotein A1.ConclusionsBMI, NLR, MPV, Cr, D-dimer, hs-CTnT, apolipoprotein A1, left subclavian artery involvement and iliac artery involvement can affect the hospital mortality rate of aortic dissection undergoing Sun’s operation to varying degrees, which may be helpful to guide the design of the perioperative treatment strategy.

O12 Artificial intelligence-based risk prediction for death after emergency laparotomy using multi slice contrast enhanced computerised tomography

Abstract Introduction Emergency laparotomy has a considerable mortality risk, with more than one in ten patients not surviving to discharge. Preoperative risk prediction using clinical tools is well established, however implemented variably. Preoperative CT is undertaken almost universally and contains granular data beyond diagnostics, including body composition, disease severity and other abstract features with the potential to enhance risk prediction. In this study we established the value of features extracted in an automated fashion from pre-operative CT in predicting 90-day post-surgery mortality. Method Anonymised CTs were collated from patients undergoing emergency laparotomy at ten hospitals in Southern England (2016–2017). For each case, axial portal venous abdominal/pelvic series were analysed using a pre-trained neural network, with each image converted into a matrix of numerical features. An elastic-net regression model to predict 90-day mortality was trained using these features and evaluated by bootstrapping with 1000 resampled datasets. Result A total of 136,709 images from 274 cases were available for analysis with a mean of 503 per case. Mortality within 90 days occurred in 34 cases (12.4%) with an average NELA mortality prediction of 8.5%. On internal (bootstrap) validation, the elastic net model derived from CT yielded excellent performance (AUC 0.903 95%CI 0.897–0.909), significantly in excess of the NELA risk calculator (AUC 0.809 95%CI 0.736–0.875), with a broader prediction range (0.01%-89.71%). Conclusion Artificial intelligence techniques applied to routinely performed cross-sectional imaging predicts emergency laparotomy mortality with greater accuracy than clinical data alone. Integration of these automated tools may be possible in the future. Take-home Message Automated analysis of CT can accurately predict risk of mortality after emergency laparotomy.

Incremental Values of T1 Mapping in the Prediction of Sudden Cardiac Death Risk in Hypertrophic Cardiomyopathy: A Comparison With Two Guidelines.

Background: MRI native T1 mapping and extracellular volume fraction (ECV) are quantitative values that could reflect various myocardial tissue characterization. The role of these parameters in predicting the risk of sudden cardiac death (SCD) in hypertrophic cardiomyopathy (HCM) is still poorly understood.Aim: This study aims to investigate the ability of native T1 mapping and ECV values to predict major adverse cardiovascular events (MACE) in HCM, and its incremental values over the 2014 European Society of Cardiology (ESC) and enhanced American College of Cardiology/American Heart Association (ACC/AHA) guidelines.Methods: Between July 2016 and October 2020, HCM patients and healthy individuals with sex and age matched who underwent cardiac MRI were prospectively enrolled. The native T1 and ECV parameters were measured. The SCD risk was evaluated by the 2014 ESC guidelines and enhanced ACC/AHA guidelines. MACE included cardiac death, transplantation, heart failure admission, and implantable cardioverter-defibrillator implantation.Results: A total of 203 HCM patients (54.2 ± 14.9 years) and 101 healthy individuals (53.2 ± 14.7 years) were evaluated. During a median follow-up of 15 months, 25 patients (12.3%) had MACE. In multivariate Cox regression analysis, global native T1 mapping (hazard ratio (HR): 1.446; 95% confidence interval (CI): 1.195–1.749; P < 0.001) and non-sustained ventricular tachycardia (NSVT) (HR: 4.949; 95% CI, 2.033–12.047; P < 0.001) were independently associated with MACE. Ten of 86 patients (11.6%) with low SCD risk assessed by the two guidelines had MACE. In this subgroup of patients, multivariate Cox regression analysis showed that global native T1 mapping was independently associated with MACE (HR: 1.532; 95% CI: 1.221–1.922; P < 0.001). In 85 patients with conflicting results assessed by the two guidelines, end-stage systolic dysfunction was independently associated with MACE (HR: 7.942, 95% CI: 1.322–47.707, P = 0.023). In 32 patients with high SCD risk assessed by the two guidelines, NSVT was independently associated with MACE (HR: 9.779, 95% CI: 1.953–48.964, P = 0.006).Conclusion: The global native T1 mapping could provide incremental values and serve as potential supplements to the current guidelines in the prediction of MACE.

Down-regulation of miR-219-5p increase the risk of cancer-related mortality in patients with prostate cancer

IntroductionProstate cancer is a common malignancy in men that is difficult to treat and carries a high risk of death. miR-219-5p is expressed in reduced amounts in many malignancies. However,...

Prediction of neonatal deaths in NICUs: development and validation of machine learning models

BackgroundPrediction of neonatal deaths in NICUs is important for benchmarking and evaluating healthcare services in NICUs. Application of machine learning techniques can improve physicians’ ability to predict the neonatal deaths. The aim of this study was to present a neonatal death risk prediction model using machine learning techniques.MethodsThis study was conducted in Tehran, Iran in two phases. Initially, important risk factors in neonatal death were identified and then several machine learning models including Artificial Neural Network (ANN), decision tree (Random Forest (RF), C5.0 and CHART tree), Support Vector Machine (SVM), Bayesian Network and Ensemble models were developed. Finally, we prospectively applied these models to predict neonatal death in a NICU and followed up the neonates to compare the outcomes of these neonates with real outcomes.Results17 factors were considered important in neonatal mortality prediction. The highest Area Under the Curve (AUC) was achieved for the SVM and Ensemble models with 0.98. The best precision and specificity were 0.98 and 0.94, respectively for the RF model. The highest accuracy, sensitivity and F-score were achieved for the SVM model with 0.94, 0.95 and 0.96, respectively. The best performance of models in prospective evaluation was for the ANN, C5.0 and CHAID tree models.ConclusionUsing the developed machine learning models can help physicians predict the neonatal deaths in NICUs.

Mechanical dispersion: are we one step closer to a more reliable risk prediction of sudden cardiac death

Mechanical dispersion: are we one step closer to a more reliable risk prediction of sudden cardiac death

A novel nomogram for early prediction of death in severe neurological disease patients with electroencephalographic periodic discharges

ObjectiveTo investigate death-related factors in patients with electroencephalographic (EEG) periodic discharges (PDs) and to construct a model for death prediction. MethodsThis case-control study enrolled a total of 80 severe neurological disease patients with EEG PDs within 72 h of admission to the neuroscience intensive care unit (NICU). According to modified Rankin scale (mRS) scores half a year after discharge, patients were divided into a survival group (<6 points) and a death group (6 points). Their relevant clinical and biochemical indicators as well as EEG characteristics were retrospectively analyzed. Logistic regression analysis was used to identify the risk factors associated with the death of patients with EEG PDs. A death risk prediction model and an individualized nomogram prediction model were constructed, and the prediction performance and concordance of the models were evaluated. ResultsMultivariate logistic regression analysis showed that the involvement of both gray and white matter in imaging, disappearance of EEG reactivity, occurrence of stimulus-induced rhythmic, periodic, or ictal discharges (SIRPIDs), and an interval time of 0.5–4 s were independent risk factors for death. A regression model was established according to the multivariate logistic regression analysis, and the area under the curve of this model was 0.9135. The accuracy of the model was 87.01%, the sensitivity was 87.17%, and the specificity was 89.17%. A nomogram model was constructed, and a concordance index of 0.914 was obtained after internal validation. ConclusionThe regression model based on risk factors has high accuracy in predicting the risk of death of patients with EEG PDs. SignificanceThis model can help clinicians in the early assessment of the prognosis of severe neurological disease patients with EEG PDs.

Risk prediction for death and end-stage renal disease does not parallel the real-life trajectory of older patients with advanced chronic kidney disease – a Romanian center experience.

IntroductionIn 2016, the European Renal Best Practice (ERBP) group published a guideline on the management of older adults with advanced chronic kidney disease (CKD). Two risk scores were highlighted: Bansal score for mortality, and Kidney Failure Risk Equation (KFRE) for estimating progression to end-stage kidney disease (ESRD). Our group, as part of the ERBP team, aimed to apply these risk prediction tools in a cohort of older adults with eGFR &lt;45 ml/min/1.73 m2.Material and methodsThis retrospective study included adults aged ≥65 years with CKD stage 3b-4, evaluated at a Romanian Outpatient Nephrology Department between October 2016 – October 2018. Bansal score was calculated for all subjects and then KFRE was used in the low mortality risk group. Outcomes were death or reaching ESRD. These outcomes were used to compare the difference between the estimated trajectory and real-life trajectory of patients. They were followed up until September 2019.ResultsFrom the total population (N=958 patients), more than half (N1=548, 57.2%) had a high mortality risk. In the remaining group with low mortality risk (N2=410, 42.8%), a significant percentage (75.4%) presented a low risk of progression to ESRD. Real-life events consisted of 164 deaths and 31 dialysis initiations. We found similar death rates in the two groups (high versus low risk of mortality). There was no difference in the rate of dialysis initiation between subjects with a high or low risk of progression to ESRD.ConclusionsRisk prognostication for death and ESRD did not parallel the real-life trajectory of our older patients with advanced CKD.

29 Predicting Unplanned Readmission and Death After Hospital Discharge: How Do Frailty Tools Compare to Electronic Health Record Frailty Markers?

Abstract Introduction Numerous frailty tools and definitions have been described. Amongst hospitalised patients, the validity of face-to-face instruments may be confounded by acute illness. However, patient assessment after recovery at the point of hospital discharge, or recognition of electronic health record (EHR) frailty markers, may overcome this issuep. Methods In a consented, prospective observational cohort study, we recruited patients ≥70 years old within 24 hours of expected discharge from the cardiology ward of the Royal Infirmary of Edinburgh. Three established frailty instruments were tested: the Fried phenotype, Short Physical Performance Battery and nurse-administered Clinical Frailty Scale (CFS). An unweighted 32-item EHR score was generated using frailty markers (e.g. falls risk, continence, cognition) recorded within mandated admission documentation. Comorbidity was assessed by count of chronic health conditions. Outcomes were a 90-day composite of unplanned readmission or death and 12-month mortality. Adjusted Cox modelling determined the hazard ratio (HR) per standard deviation increase in each frailty score. Results 186 patients (mean age 79 ± 6 years, 64% male) were included, of whom 55 (30%) had a 90-day composite outcome, and 21 (11%) died within 12 months. All four frailty tools were moderately correlated with age and comorbidity (Pearson’s r 0.21 to 0.43, all p &amp;lt; 0.05). The Fried phenotype (HR 1.47, 95% CI 1.18–1.81), CFS (HR 1.24, 95% CI 1.01–1.51) and EHR score (HR 1.26, 95% CI 1.03–1.55) independently predicted 90-day readmission or death, after adjustment for age, sex and comorbidity. All frailty instruments were independent predictors of 12-month mortality, with age, sex and comorbidity losing predictive power (p &amp;gt; 0.05) once frailty was included in modelling. Conclusions At hospital discharge, the Fried phenotype and CFS added to age and comorbidity in risk prediction for future unplanned readmission or death. EHR frailty markers appeared comparable to face-to-face assessment. An automated trigger for high-risk patients using routine EHR data merits prospective evaluation.

Association of P-Wave Abnormalities With Sudden Cardiac and Cardiovascular Death: The ARIC Study.

Association of P-Wave Abnormalities With Sudden Cardiac and Cardiovascular Death: The ARIC Study.

Clinical and inflammatory features based machine learning model for fatal risk prediction of hospitalized COVID-19 patients: results from a retrospective cohort study

Objectives To appraise effective predictors for COVID-19 mortality in a retrospective cohort study. Methods A total of 1270 COVID-19 patients, including 984 admitted in Sino French New City Branch (training and internal validation sets randomly split at 7:3 ratio) and 286 admitted in Optical Valley Branch (external validation set) of Wuhan Tongji hospital, were included in this study. Forty-eight clinical and laboratory features were screened with LASSO method. Further multi-tree extreme gradient boosting (XGBoost) machine learning-based model was used to rank importance of features selected from LASSO and subsequently constructed death risk prediction model with simple-tree XGBoost model. Performances of models were evaluated by AUC, prediction accuracy, precision, and F1 scores. Results Six features, including disease severity, age, levels of high-sensitivity C-reactive protein (hs-CRP), lactate dehydrogenase (LDH), ferritin, and interleukin-10 (IL-10), were selected as predictors for COVID-19 mortality. Simple-tree XGBoost model conducted by these features can predict death risk accurately with >90% precision and >85% sensitivity, as well as F1 scores >0.90 in training and validation sets. Conclusion We proposed the disease severity, age, serum levels of hs-CRP, LDH, ferritin, and IL-10 as significant predictors for death risk of COVID-19, which may help to identify the high-risk COVID-19 cases. KEY MESSAGES A machine learning method is used to build death risk model for COVID-19 patients. Disease severity, age, hs-CRP, LDH, ferritin, and IL-10 are death risk factors. These findings may help to identify the high-risk COVID-19 cases.

Nitrogen oxide and its circulating NOx metabolites, their role in human body functioning and cardiovascular death risk prediction (Part II)

Nitrogen oxide and its circulating NOx metabolites, their role in human body functioning and cardiovascular death risk prediction (Part II)

Nitric oxide and its circulating NOx metabolites, their role in human body functioning and cardiovascular death risk prediction (part I)

Nitric oxide and its circulating NOx metabolites, their role in human body functioning and cardiovascular death risk prediction (part I)

Is positivity for hepatitis C virus antibody predictive of lower risk of death in COVID-19 patients with cirrhosis?

Liver injury has been reported in coronavirus disease 2019 (COVID-19) cases but the impact of pre-existing liver damage and related etiology have not been completely elucidated. Our research interests include the potential reciprocal influence of COVID-19 and pre-existing liver damage related to hepatitis C virus (HCV) infection, in particular. To this end, we have evaluated three cohorts of patients admitted at three Italian hospitals during the coronavirus pandemic; these included 332 patients with COVID-19 and 1527 patients with HCV who were from established real-world antiviral treatment study cohorts (sofosbuvir/velpatasvir), with either liver disease (various severities; n = 1319) or cirrhosis (n = 208). Among the COVID-19 patients, 10 had cirrhosis (3%), including 7 of metabolic origin and 3 of viral origin. Mortality among the COVID-19 patients was 27.1%, with 70% of those with cirrhosis of metabolic etiology having died. Cirrhosis, older age, low white blood cell count and lymphocyte count being identified as risk predictors of death [odds ratio (OR) = 13.7, 95% confidence interval (CI): 2.59-83.01, P = 0.006; OR = 1.05, 95%CI: 1.03-1.08, P = 0.0001; OR = 1.09, 95%CI: 1.36-1.16, P = 0.001; OR = 0.61, 95%CI: 0.39-0.93, P = 0.023, respectively]. In the two cohorts of HCV patients, COVID-19 diagnosis was made in 0.07% of those with liver disease and 1% of those with cirrhosis. Thus, the prevalence of HCV antibodies among COVID-19-infected patients was comparable to that currently reported for the general population in Italy. Amongst the COVID-19 patients, pre-existing metabolic cirrhosis appears to be associated with higher mortality, while HCV antibodies may be suggestive of “protection” against COVID-19.

ICD implantation and device therapy: Fabry vs hypertrophic cardiomyopathy

Abstract Background Cardiac involvement of Fabry disease (FD) includes left ventricular hypertrophy (LVH), inflammation, arrhythmia and sudden death. There are limited data regarding potential risk predictors and no definitive criteria exist to guide implantation of cardiac devices for primary prevention. Despite phenotypic similarities between FD and hypertrophic cardiomyopathy (HCM), FD is specifically excluded from the ESC sudden cardiac death risk prediction tool used for HCM. Purpose To evaluate differences in device implantation and arrhythmic burden between advanced Fabry cardiomyopathy and HCM. Methods This multi-centre, retrospective study including genetically confirmed FD and age/gender-matched HCM patients all of whom had an implantable cardioverter defibrillator (ICD). Data was collected prior to device implantation on disease characteristics, biomarkers and CMR imaging. Arrhythmia burden and changes in therapy were captured following implantation. Results Of the UK FD population under follow-up, 50/880 had an ICD implanted (80% males, 51% non-classical mutation, mean age at device implantation 57±12 years). A comparator group included 64 age- and gender-matched HCM patients (67% males, mean age at implant 46±39 years). Baseline LV mass was greater in FD (291±97g/m2 vs 218±78g/m2, p=0.012). FD patients had higher troponin (95 vs 19ng/l, p&amp;lt;0.001) but similar NT-pro-BNP (1687 vs 888ng/l, p=0.086) levels. Indications for ICD insertion in FD included: presumed HCM dual pathology 14%, symptomatic NSVT 18%, asymptomatic NSVT 24%, co-existing long QT 2%, CRT-D 14%, no clear indication (primary prevention in the presence of multiple potential arrhythmic risk factors) 28%. All HCM patients were risk stratified and underwent device implantation based on an estimated 5-year SCD risk &amp;gt;4%. All FD patients had a SCD risk &amp;gt;4% using this risk calculator. Arrhythmia were more common in FD over shorter follow-up (37/50, 74% over 4.3±3.0 years vs 28/64 in HCM, 44% over 6.5±2.9 years, p=0.001). Notably, VT requiring anti-tachycardia pacing (ATP) +/− defibrillation therapy from the ICD was more frequent (FD: 14/50, 28% vs HCM: 8/64, 12%, p=0.055), as was immediate shock therapy for sustained VT (p=0.009, figure panel B). FD patients with arrhythmia were often older, had greater LV mass, a larger left atrium and a broader QRS duration. These clinical features tended to occur more frequently in FD than in the HCM group. Conclusion This study has shown that delivery of device therapy in Fabry cardiomyopathy is higher than in HCM. Despite similar rates of asymptomatic NSVT, a higher rate of ventricular arrhythmia requiring ATP/defibrillation therapy occurred in FD. Although both FD and HCM had similar risk percentages according to the ESC calculator, active troponitis in FD was double that of HCM. Funding Acknowledgement Type of funding source: None

Prognostic Value of Neutrophil-to-Lymphocyte Ratio in Predicting Death Risk in Patients with Severe Hand, Foot and Mouth Disease.

IntroductionSevere hand, foot, and mouth disease (HFMD) may lead to serious complications, which cause child mortality during outbreaks. The aim of this study was to determine whether neutrophil-to-lymphocyte ratio (NLR) can predict death risk in severe HFMD.MethodsMedical records for 664 severe HFMD patients were retrospectively examined, and NLR was calculated from blood counts. Youden’s index was calculated to determine the optimal NLR cutoff. Uni- and multivariate logistic regression were used to determine death risk factors associated with severe HFMD.ResultsAn NLR cutoff value of 2.01 and 2.50 respectively predicted mortality among all 664 severe HFMD and 137 critical HFMD. Among all 664 patients, the multivariate model identified the following as independently associated with death risk: high fever (OR 3.342, 95% CI 1.736–6.432), EV71 infection (OR 3.200, 95% CI 1.529–6.698), fasting glucose (OR 37.343, 95% CI 18.616–74.909), and NLR (>2.01) (OR 2.142, 95% CI 1.125–4.079). Among 137 critical HFMD, EV71 infection (OR 3.441, 95% CI 1.132–10.462), fasting glucose (OR 14.173, 95% CI 4.920–40.827), and NLR (>2.50) (OR 4.166, 95% CI 1.570–11.051) were associated with death risk.ConclusionIn conclusion, NLR (>2.01) in severe HFMD and NLR (>2.50) in critical HFMD patients may be associated with increased death risk.

Prediction of Sudden Cardiac Death Risk with a Support Vector Machine Based on Heart Rate Variability and Heartprint Indices.

Most methods for sudden cardiac death (SCD) prediction require long-term (24 h) electrocardiogram recordings to measure heart rate variability (HRV) indices or premature ventricular complex indices (with the heartprint method). This work aimed to identify the best combinations of HRV and heartprint indices for predicting SCD based on short-term recordings (1000 heartbeats) through a support vector machine (SVM). Eleven HRV indices and five heartprint indices were measured in 135 pairs of recordings (one before an SCD episode and another without SCD as control). SVMs (defined with a radial basis function kernel with hyperparameter optimization) were trained with this dataset to identify the 13 best combinations of indices systematically. Through 10-fold cross-validation, the best area under the curve (AUC) value as a function of γ (gamma) and cost was identified. The predictive value of the identified combinations had AUCs between 0.80 and 0.86 and accuracies between 80 and 86%. Further SVM performance tests on a different dataset of 68 recordings (33 before SCD and 35 as control) showed AUC = 0.68 and accuracy = 67% for the best combination. The developed SVM may be useful for preventing imminent SCD through early warning based on electrocardiogram (ECG) or heart rate monitoring.

Establishment of nomogram predicting model for the death risk of extremely severe burn patients and the predictive value

Objective: To explore the death risk factors of extremely severe burn patients, establish a death risk nomogram predicting model, and investigate the predictive value for death risk of extremely severe burn patients. Methods: The medical records of 231 extremely severe burn patients (190 males and 41 females, aged 18-60 years) who were admitted to the Institute of Burn Research of the First Affiliated Hospital of Army Medical University from January 2010 to October 2018 and met the inclusion criteria were analyzed retrospectively. According to the final outcome, the patients were divided into survival group of 173 patients and death group of 58 patients. The sex, age, severity of inhalation injury, total burn area, full-thickness burn area, burn index, rehydration coefficient and urine volume coefficient of the first and second 24 h after injury, the first base excess, shock index, and hematocrit (HCT) after admission, whether to have pre-hospital fluid infusion, use of ventilator, and use of continuous renal replacement therapy (CRRT), and abbreviated burn severity index (ABSI ) and Baux score on admission of patients in the two groups were recorded or calculated. According to the use of ventilator, the patients were divided into with ventilator group of 131 patients and without ventilator group of 100 patients, and the death, total burn surface area, burn index, incidence and severity of inhalation injury were recorded. According to the use of CRRT, the patients were divided into with CRRT group of 59 patients and without CRRT group of 172 patients, and the death, total burn surface area, and burn index were recorded. Data were statistically analyzed with t test, chi-square test, and Mann-Whitney U test to screen the death related factors of patients. The indexes with statistically significant differences between survival group and death group were included in the multivariate logistic regression analysis to screen the independent death risk factors of patients, and the death risk nomogram predicting model was constructed based on the results.The Bootstrap method was used to validate the death risk nomogram predicting model internally. The predictive value of the nomogram model for predicting death risk of patients was detected by drawing calibration graph and calculating concordance index, and the death risk scores of 231 patients were acquired according to the death risk nomogram model. The receiver's operating characteristic (ROC) curve was drawn, and the optimal threshold and the sensitivity and specificity of optimal threshold in the ROC curve and the area under the curve were calculated. Results: (1) There were statistically significant differences in burn index, ABSI on admission, severity of inhalation injury, total burn area, full-thickness burn area, rehydration coefficient at the first 24 h after injury, use of ventilator, use of CRRT, and Baux score on admission of patients between the two groups (Z=-7.696, -7.031, χ(2)=18.304, 63.065, 23.300, 13.073, 34.240, 59.586, t=-7.536, P<0.01). (2) There were statistically significant differences in death, incidence and severity of inhalation injury, total burn area, and burn index of patients between with ventilator group and without ventilator group (χ(2)=34.240, 17.394, 25.479, Z=-6.557, -7.049, P<0.01). (3) There were statistically significant differences in death, total burn area, and burn index of patients between with CRRT group and without CRRT group (χ(2)=62.982, Z= -47.421, -6.678, P<0.01). (4) The use of ventilator, use of CRRT, and burn index were independent risk factors for the death of extremely severe burn patients (odds ratio=3.277, 5.587, 1.067, 95% confidence interval=1.073-10.008, 2.384-13.093, 1.038-1.096, P<0.05 or P<0.01). (5) The initial concordance index of nomogram predicting model was 0.90 and the corrected concordance index was 0.89. The concordance indexes before and after correction were higher and similar, which showed that the nomogram had good concordance and predictive effect. The optimum threshold of ROC curve was 0.23, the sensitivity and specificity of optimum threshold were 86.0% and 80.0%, respectively, and the area under ROC curve was 0.90 (95% confidence interval=0.86-0.94, P<0.01). Conclusions: Severe burns and damage and/or failure of organ are the main death causes of extremely severe burn patients. The death risk nomogram predicting model established on the basis of use of ventilator, use of CRRT, and burn index have good predictive ability for death of extremely severe burn patients.