Mortality Predictions Research Articles

Evaluation of firearm injuries by trauma scoring systems in a secondary health care institution.

Scoring systems are widely used in trauma patients and are very important in the care of trauma patients. The objective of this study was to investigate the effectiveness of scoring systems in evaluating the patients who were brought to a hospital without a trauma center due to firearm injuries (FIs) using scoring systems, and the efficacy of these systems in identifying patients who should be treated in a trauma center. Patients brought to Hakkari Yüksekova State Hospital due to FIs between January 2010 and December 2019 were retrospectively investigated. Patients aged 16 and over were included in the study. Patients who were brought to the hospital while deceased and those who did not respond to cardiopulmonary resuscitation in the hospital were excluded from the study. In addition, patients who underwent simple outpatient medical intervention were not included in the patient group. Patients were evaluated demographically. The mortality predictions of trauma scoring systems, which are widely used in the evaluation of trauma patients, were examined. In addition, the effectiveness of scoring systems in identifying patients who should be treated in trauma centers was investigated. In the study, 331 patients, 96.9% of whom were male, were evaluated. The patient group consisted of young patients and the median age was 27 (IQR, 24-29) years. A total of 74 (22.4%) patients were referred to trauma referral hospitals for diagnosis and treatment. Mortality occurred in 46 (13.9%) patients. Glasgow coma scale, injury severity score (ISS), revised trauma score, new trauma score, and trauma revised ISS were found effective for predicting mortality in patients, and scoring systems were correlated with each other. However, scoring systems were not found effective in distinguishing patients who should be treated in a trauma center. Scoring systems are significant in predicting mortality of patients with gunshot wounds. However, trauma scoring systems have not been found to be effective in distinguishing patients who require referral to a trauma center.

387. Markers for Mortality in COVID-19 Patients with Atrial Fibrillation or Flutter

BackgroundSevere Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection can lead to many different cardiovascular complications, we were interested in studying prognostic markers in patients with atrial fibrillation/flutter (A. Fib/Flutter).MethodsA retrospective cohort study of patients with confirmed COVID-19 and either with existing or new onset A. Fib/Flutter who were admitted to our hospital between March 15 and May 20, 2020. Demographic, outcome and laboratory data were extracted from the electronic medical record and compared between survivors and non-survivors. Univariate and multivariate logistic regression were employed to identify the prognostic markers associated with mortality in patients with A. Fib/FlutterResultsThe total number of confirmed COVID-19 patients during the study period was 350; 37 of them had existing or new onset A. Fib/Flutter. Twenty one (57%) expired, and 16 (43%) were discharged alive. The median age was 72 years old, ranged from 19 to 100 years old. Comorbidities were present in 33 (89%) patients, with hypertension (82%) being the most common, followed by diabetes (46%) and coronary artery disease (30%).New onset of atrial fibrillation was identified in 23 patients (70%), of whom 13 (57%) expired; 29 patients (78%) presented with atrial fibrillation with rapid ventricular response, and 2 patients (5%) with atrial flutter. Mechanical ventilation was required for 8 patients, of whom 6 expired.In univariate analysis, we found a significant difference in baseline ferritin (p=0.04), LDH (p=0.02), neutrophil-lymphocyte ratio (NLR) (p=0.05), neutrophil-monocyte ratio (NMR) (p=0.03) and platelet (p=0.015) between survivors and non-survivors. With multivariable logistic regression analysis, the only value that had an odds of survival was a low NLR (odds ratio 0.74; 95% confidence interval 0.53–0.93).ConclusionThis retrospective cohort study of hospitalized patients with COVID-19 demonstrated an association of increase NLR as risk factors for death in COVID-19 patients with A. Fib/Flutter. A high NLR has been associated with increased incidence, severity and risk for stroke in atrial fibrillation patients but to our knowledge, we are first to demonstrate the utilization in mortality predictions in COVID-19 patients with A. Fib/Flutter.Disclosures Jihad Slim, MD, Abbvie (Speaker’s Bureau)Gilead (Speaker’s Bureau)Jansen (Speaker’s Bureau)Merck (Speaker’s Bureau)ViiV (Speaker’s Bureau)

A framework for making predictive models useful in practice.

ObjectiveTo analyze the impact of factors in healthcare delivery on the net benefit of triggering an Advanced Care Planning (ACP) workflow based on predictions of 12-month mortality.Materials and MethodsWe built a predictive model of 12-month mortality using electronic health record data and evaluated the impact of healthcare delivery factors on the net benefit of triggering an ACP workflow based on the models’ predictions. Factors included nonclinical reasons that make ACP inappropriate: limited capacity for ACP, inability to follow up due to patient discharge, and availability of an outpatient workflow to follow up on missed cases. We also quantified the relative benefits of increasing capacity for inpatient ACP versus outpatient ACP.ResultsWork capacity constraints and discharge timing can significantly reduce the net benefit of triggering the ACP workflow based on a model’s predictions. However, the reduction can be mitigated by creating an outpatient ACP workflow. Given limited resources to either add capacity for inpatient ACP versus developing outpatient ACP capability, the latter is likely to provide more benefit to patient care.DiscussionThe benefit of using a predictive model for identifying patients for interventions is highly dependent on the capacity to execute the workflow triggered by the model. We provide a framework for quantifying the impact of healthcare delivery factors and work capacity constraints on achieved benefit.ConclusionAn analysis of the sensitivity of the net benefit realized by a predictive model triggered clinical workflow to various healthcare delivery factors is necessary for making predictive models useful in practice.

Development and validation of two composite aging measures using clinical biomarkers in the Chinese population

Quantifying aging is crucial for addressing aging and related issues. This study aimed to: 1) develop two composite aging measures in the Chinese population using two recent advanced algorithms (the Klemera and Doubal method and Mahalanobis distance); and 2) validate the two measures by examining their associations with mortality and disease counts. Based on data from the China Nutrition and Health Survey 2009 wave (N=8,119, aged 20-79 years, 53.5% women), a nationwide prospective cohort study of the Chinese population, we developed Klemera and Doubal method-biological age (KDM-BA) and physiological dysregulation (PD, derived from Mahalanobis distance) using 12 routine clinical biomarkers. For the validation analysis, we used Cox proportional hazard regression models (for mortality) and linear, Poisson, and logistic regression models (for disease counts) to examine the associations. We replicated the validation analysis in the China Health and Retirement Longitudinal Study (CHARLS, N=9,304, aged 45-99 years, 53.4% women). We found that both aging measures were predictive of mortality after accounting for age and gender (KDM-BA, per one-year, HR=1.14, 95%CI=1.08, 1.19; PD, per one-SD, HR=1.50, 95%CI=1.33, 1.69). With few exceptions, these mortality predictions were robust across stratifications by age, gender, education, and health behaviors. The two aging measures were associated with disease counts both cross-sectionally and longitudinally. These results were generally replicable in CHARLS although four biomarkers were not available. In summary, we successfully developed and validated two composite aging measures‒‒KDM-BA and PD, which have great potentials for applications in early identifications and preventions of aging and aging related diseases in China.

Risk Prediction Models for Post-Operative Mortality in Patients With Cirrhosis.

Patients with cirrhosis are at increased risk of postoperative mortality. Currently available tools to predict postoperative risk are suboptimally calibrated and do not account for surgery type. Our objective was to use population-level data to derive and internally validate cirrhosis surgical risk models. We conducted a retrospective cohort study using data from the Veterans Outcomes and Costs Associated with Liver Disease (VOCAL) cohort, which contains granular data on patients with cirrhosis from 128 U.S. medical centers, merged with the Veterans Affairs Surgical Quality Improvement Program (VASQIP) to identify surgical procedures. We categorized surgeries as abdominal wall, vascular, abdominal, cardiac, chest, or orthopedic and used multivariable logistic regression to model 30-, 90-, and 180-day postoperative mortality (VOCAL-Penn models). We compared model discrimination and calibration of VOCAL-Penn to the Mayo Risk Score (MRS), Model for End-Stage Liver Disease (MELD), Model for End-Stage Liver Disease-Sodium MELD-Na, and Child-Turcotte-Pugh (CTP) scores. We identified 4,712 surgical procedures in 3,785 patients with cirrhosis. The VOCAL-Penn models were derived and internally validated with excellent discrimination (30-day postoperative mortality C-statistic=0.859; 95% confidence interval [CI], 0.809-0.909). Predictors included age, preoperative albumin, platelet count, bilirubin, surgery category, emergency indication, fatty liver disease, American Society of Anesthesiologists classification, and obesity. Model performance was superior to MELD, MELD-Na, CTP, and MRS at all time points (e.g., 30-day postoperative mortality C-statistic for MRS=0.766; 95% CI, 0.676-0.855) in terms of discrimination and calibration. The VOCAL-Penn models substantially improve postoperative mortality predictions in patients with cirrhosis. These models may be applied in practice to improve preoperative risk stratification and optimize patient selection for surgical procedures (www.vocalpennscore.com).

Low volume ECMO results study

Objectives: To report extracorporeal membrane oxygenation (ECMO) experience at Princess Alexandra and Gold Coast University hospitals and compare mortality with benchmarks. Design: Case series of patients treated with ECMO. Setting: Two adult tertiary Australian intensive care units with low ECMO case volumes. Participants: Patients treated with ECMO, aged > 18 years. Main outcome measures: Patients were categorised into respiratory, cardiac, and extracorporeal cardiopulmonary resuscitation (eCPR) groups. Observed mortality was compared with mortality predicted using individual risk of death predictions from the Survival after Veno-arterial ECMO (SAVE) and Respiratory ECMO Survival Prediction (RESP) scores; mortality predicted when mortality predictions of the SAVE score were modified to be consistent with the validation cohort in the SAVE study (Alfred Hospital); and with mortality predicted when eCPR patients were all assigned a risk of death equal to Extracorporeal Life Support Organization (ELSO) Registry eCPR mortality. Results: Over 10 years, 86 patients were treated with ECMO. Eight deaths were observed in 49 patients with respiratory failure, below the 95% CI (13-24) for the deaths predicted by the RESP score (P < 0.001). Nine deaths were observed in 27 patients with cardiac failure, below the 95% CI (14-23) for the deaths predicted by the SAVE score (P < 0.001), but within the 95% CI (9-17) for the deaths predicted by the SAVE score modified to be consistent with the Alfred Hospital cohort (P > 0.05). Seven deaths were observed in the ten eCPR patients, within the 95% CI (4-10) predicted using the risk of death derived from the ELSO Registry. Conclusions: Mortality in two low volume ECMO centres was not inferior to benchmarks.

Understanding and predicting forest mortality in the western United States using long-term forest inventory data and modeled hydraulic damage.

Global warming is expected to exacerbate the duration and intensity of droughts in the western United States, which may lead to increased tree mortality. A prevailing proximal mechanism of drought-induced tree mortality is hydraulic damage, but predicting tree mortality from hydraulic theory and climate data still remains a major scientific challenge. We used forest inventory data and a plant hydraulic model (HM) to address three questions: can we capture regional patterns of drought-induced tree mortality with HM-predicted damage thresholds; do HM metrics improve predictions of mortality across broad spatial areas; and what are the dominant controls of forest mortality when considering stand characteristics, climate metrics, and simulated hydraulic stress? We found that the amount of variance explained by models predicting mortality was limited (R2 median=0.10, R2 range: 0.00-0.52). HM outputs, including hydraulic damage and carbon assimilation diagnostics, moderately improve mortality prediction across the western US compared with models using stand and climate predictors alone. Among factors considered, metrics of stand density and tree size tended to be some of the most critical factors explaining mortality, probably highlighting the important roles of structural overshoot, stand development, and biotic agent host selection and outbreaks in mortality patterns.

Abstract 13611: Prognostic Value of Shock Index in Patients Admitted With Non ST-Segment Elevation Myocardial Infarction: The ARIC Study Community Surveillance

Introduction: Shock index (SI), defined as the ratio of heart rate (HR) to systolic blood pressure (SBP), is easily obtained and has been reported to predict all-cause mortality in acute myocardial infarction (AMI) hospitalizations, primarily among patients with STEMI. Large-scale investigations examining the prognostic value of SI in patients with NSTEMI are lacking. Methods: Hospitalizations for acute myocardial infarction were sampled from 4 US areas by the ARIC Study and classified by physician review. HR and SBP at first presentation were abstracted from the medical record. Pulseless patients, those with unobtainable SBP, and transfer patients were excluded. An SI value ≥0.7 was considered high. The TIMI risk score for NSTE-ACS was calculated using the established algorithm. Predictions of all-cause 28-day mortality were analyzed using receiver operating characteristics (ROC). Results: From 2000-2014, 18,301 weighted hospitalizations for NSTEMI patients were sampled and had vitals successfully obtained. Of these, 5753 (31%) had high SI (≥0.7). Patients with high SI were more often female (46% vs 39%) and older (65 vs 63 years), with more prevalent chronic kidney disease (40% vs. 32%), pneumonia (16% vs 8%), and complication by acute heart failure (42% vs 30%). However, the TIMI risk score was comparable for the 2 groups (4.3 vs. 4.2). Angiography (36% vs 58%), revascularization (19% vs 41%), and guideline-directed medications were less often administered to patients with high SI, and the 28-day mortality was nearly 3 times higher (13% vs 5%). Continuous values of SI were more predictive of 28-day mortality than the TIMI risk score, both overall (ROC-AUC: 0.68 vs 0.54; P &lt;0.0001) and among the subset of patients who were revascularized (ROC-AUC: 0.72 vs. 0.56; P &lt;0.0001). Conclusion: The SI outperformed the TIMI risk score for prediction of short-term mortality in patients hospitalized with NSTEMI, and may be useful for risk stratification in emergency settings.

All-cause mortality supports the COVID-19 mortality in Belgium and comparison with major fatal events of the last century

BackgroundThe COVID-19 mortality rate in Belgium has been ranked among the highest in the world. To assess the appropriateness of the country’s COVID-19 mortality surveillance, that includes long-term care facilities deaths and deaths in possible cases, the number of COVID-19 deaths was compared with the number of deaths from all-cause mortality. Mortality during the COVID-19 pandemic was also compared with historical mortality rates from the last century including those of the Spanish influenza pandemic.MethodsExcess mortality predictions and COVID-19 mortality data were analysed for the period March 10th to June 21st 2020. The number of COVID-19 deaths and the COVID-19 mortality rate per million were calculated for hospitals, nursing homes and other places of death, according to diagnostic status (confirmed/possible infection). To evaluate historical mortality, monthly mortality rates were calculated from January 1900 to June 2020.ResultsNine thousand five hundred ninety-one COVID-19 deaths and 39,076 deaths from all-causes were recorded, with a correlation of 94% (Spearman’s rho, p < 0,01). During the period with statistically significant excess mortality (March 20th to April 28th; total excess mortality 64.7%), 7917 excess deaths were observed among the 20,159 deaths from all-causes. In the same period, 7576 COVID-19 deaths were notified, indicating that 96% of the excess mortality were likely attributable to COVID-19. The inclusion of deaths in nursing homes doubled the COVID-19 mortality rate, while adding deaths in possible cases increased it by 27%. Deaths in laboratory-confirmed cases accounted for 69% of total COVID-19-related deaths and 43% of in-hospital deaths. Although the number of deaths was historically high, the monthly mortality rate was lower in April 2020 compared to the major fatal events of the last century.ConclusionsTrends in all-cause mortality during the first wave of the epidemic was a key indicator to validate the Belgium’s high COVID-19 mortality figures. A COVID-19 mortality surveillance limited to deaths from hospitalised and selected laboratory-confirmed cases would have underestimated the magnitude of the epidemic. Excess mortality, daily and monthly number of deaths in Belgium were historically high classifying undeniably the first wave of the COVID-19 epidemic as a fatal event.

Usefulness of updated logistic clinical SYNTAX score based on MI-SYNTAX score in patients with ST-elevation myocardial infarction.

To compare the predictive performances of the prewiring, postwiring MI-SYNTAX scores, prewiring, and postwiring Updated Logistic Clinical SYNTAX score (LCSS) for 2-year all-cause mortality post percutaneous coronary intervention (PCI) in ST-elevation myocardial infarction (STEMI) patients. In patients with STEMI and undergoing primary PCI, coronary stenosis(es) distal to the culprit lesion is often observed after the restoration of coronary flow. To address comprehensively the complex coronary anatomy in these patients, prewiring and postwiring MI-SYNTAX scores have been reported in the literature. Furthermore, to enable individualized risk estimation for long-term all-cause mortality, the Updated LCSS has been developed by combining the anatomical SYNTAX score and clinical factors. In the randomized GLOBAL LEADERS trial, anatomical SYNTAX score analysis was performed by an independent angiographic corelab for the first 4,000 consecutive patients as a prespecified analysis; of these, 545 presented with STEMI. The efficacy of the mortality predictions of the four scores at 2 years were evaluated based on their discrimination and calibration abilities. Complete data was available in 512 patients (93.9%). When the patients were stratified into two groups based on the median of the scores, the prewiring and postwiring Updated LCSSs demonstrated that the high-score groups were associated with higher rates of 2-year all-cause mortality compared to the low-score groups (6.6 vs. 1.2%; log-rank p=.001 and 6.6 vs. 1.2%; log-rank p=.001, respectively). There were no statistically significant differences for predicting the mortality between the prewiring (area under the curve [AUC] 0.625), postwiring MI-SYNTAX score (AUC 0.614), prewiring (AUC 0.755), and postwiring Updated LCSS (AUC 0.757). In the integrated discrimination improvement (IDI), the prewiring MI-SYNTAX score had a better discrimination for the mortality than the postwiring MI-SYNTAX score (IDI -0.0082; p=.029). The four scores had acceptable calibration abilities for 2-year all-cause mortality. The prewiring Updated LCSS predicts long-term all-cause mortality with clearly useful discrimination and acceptable calibration. Since the postwiring MI-SYNTAX score does not improve mortality prediction, the prewiring MI-SYNTAX score may be preferred for the 2-year mortality prediction using the Updated LCSS.

Mortality Prediction from Hospital-Acquired Infections in Trauma Patients Using an Unbalanced Dataset.

ObjectivesMachine learning has been widely used to predict diseases, and it is used to derive impressive knowledge in the healthcare domain. Our objective was to predict in-hospital mortality from hospital-acquired infections in trauma patients on an unbalanced dataset.MethodsOur study was a cross-sectional analysis on trauma patients with hospital-acquired infections who were admitted to Shiraz Trauma Hospital from March 20, 2017, to March 21, 2018. The study data was obtained from the surveillance hospital infection database. The data included sex, age, mechanism of injury, body region injured, severity score, type of intervention, infection day after admission, and microorganism causes of infections. We developed our mortality prediction model by random under-sampling, random over-sampling, clustering (k-mean)-C5.0, SMOTE-C5.0, ADASYN-C5.5, SMOTE-SVM, ADASYN-SVM, SMOTE-ANN, and ADASYN-ANN among hospital-acquired infections in trauma patients. All mortality predictions were conducted by IBM SPSS Modeler 18.ResultsWe studied 549 individuals with hospital-acquired infections in a trauma hospital in Shiraz during 2017 and 2018. Prediction accuracy before balancing of the dataset was 86.16%. In contrast, the prediction accuracy for the balanced dataset achieved by random under-sampling, random over-sampling, clustering (k-mean)-C5.0, SMOTE-C5.0, ADASYN-C5.5, and SMOTE-SVM was 70.69%, 94.74%, 93.02%, 93.66%, 90.93%, and 100%, respectively.ConclusionsOur findings demonstrate that cleaning an unbalanced dataset increases the accuracy of the classification model. Also, predicting mortality by a clustered under-sampling approach was more precise in comparison to random under-sampling and random over-sampling methods.

Blood-based epigenetic estimators of chronological age in human adults using DNA methylation data from the Illumina MethylationEPIC array

BackgroundEpigenetic clocks have been recognized for their precise prediction of chronological age, age-related diseases, and all-cause mortality. Existing epigenetic clocks are based on CpGs from the Illumina HumanMethylation450 BeadChip (450 K) which has now been replaced by the latest platform, Illumina MethylationEPIC BeadChip (EPIC). Thus, it remains unclear to what extent EPIC contributes to increased precision and accuracy in the prediction of chronological age.ResultsWe developed three blood-based epigenetic clocks for human adults using EPIC-based DNA methylation (DNAm) data from the Norwegian Mother, Father and Child Cohort Study (MoBa) and the Gene Expression Omnibus (GEO) public repository: 1) an Adult Blood-based EPIC Clock (ABEC) trained on DNAm data from MoBa (n = 1592, age-span: 19 to 59 years), 2) an extended ABEC (eABEC) trained on DNAm data from MoBa and GEO (n = 2227, age-span: 18 to 88 years), and 3) a common ABEC (cABEC) trained on the same training set as eABEC but restricted to CpGs common to 450 K and EPIC. Our clocks showed high precision (Pearson correlation between chronological and epigenetic age (r) > 0.94) in independent cohorts, including GSE111165 (n = 15), GSE115278 (n = 108), GSE132203 (n = 795), and the Epigenetics in Pregnancy (EPIPREG) study of the STORK Groruddalen Cohort (n = 470). This high precision is unlikely due to the use of EPIC, but rather due to the large sample size of the training set.ConclusionsOur ABECs predicted adults’ chronological age precisely in independent cohorts. As EPIC is now the dominant platform for measuring DNAm, these clocks will be useful in further predictions of chronological age, age-related diseases, and mortality.

Performance of Intensive Care Unit Severity Scoring Systems Across Different Ethnicities

Background: Despite wide utilisation of severity scoring systems for case-mix determination and benchmarking in the intensive care unit, the possibility of scoring bias across ethnicities has not been examined. Recent guidelines on the use of illness severity scores to inform triage decisions for allocation of scarce resources such as mechanical ventilation during the current COVID-19 pandemic warrant examination for possible bias in these models. We investigated the performance of three severity scoring systems (APACHE IVa, OASIS, SOFA) across ethnic groups in two large ICU databases in order to identify possible ethnicity-based bias.Method: Data from the eICU Collaborative Research Database and the Medical Information Mart for Intensive Care were analysed for score performance in Asians, African Americans, Hispanics and Whites after appropriate exclusions. Discrimination and calibration were determined for all three scoring systems in all four groups.Finding: While measurements of discrimination - area under the receiver operating characteristic curve (AUROC) - were significantly different among the groups, they did not display any discernible systematic patterns of bias. In contrast, measurements of calibration - standardised mortality ratio (SMR) - indicated persistent, and in some cases significant, patterns of difference between Hispanics and African Americans versus Asians and Whites. The differences between African Americans and Whites were consistently statistically significant. While calibrations were imperfect for all groups, the scores consistently demonstrated a pattern of over-predicting mortality for African Americans and Hispanics.Interpretation:The systematic differences in calibration across ethnic groups suggest that illness severity scores reflect bias in their predictions of mortality.Funding Statement: Funding: LAC is funded by the National Institute of Health through NIBIB R01 EB017205. There was no specific funding for this study.Declaration of Interests: RS received writing fees for healthcare reports from Crystallise UK Ltd. None of the other authors declare any conflict of interest.Ethics Approval Statement: Research using the eICU-CRD is exempt from institutional review board (IRB) approval due to the retrospective design, lack of direct patient intervention, and the security schema, for which the re-identification risk was certified as meeting safe harbour standards by an independent privacy expert (Privacert, Cambridge, MA) (Health Insurance Portability and Accountability Act Certification no. 1031219-2). The data in MIMIC-III has been previously de-identified, and the IRBs of the Massachusetts Institute of Technology (No. 0403000206) and BIDMC (2001-P-001699/14) both approved the use of the database for research.

A Racially Unbiased, Machine Learning Approach to Prediction of Mortality: Algorithm Development Study.

BackgroundRacial disparities in health care are well documented in the United States. As machine learning methods become more common in health care settings, it is important to ensure that these methods do not contribute to racial disparities through biased predictions or differential accuracy across racial groups.ObjectiveThe goal of the research was to assess a machine learning algorithm intentionally developed to minimize bias in in-hospital mortality predictions between white and nonwhite patient groups.MethodsBias was minimized through preprocessing of algorithm training data. We performed a retrospective analysis of electronic health record data from patients admitted to the intensive care unit (ICU) at a large academic health center between 2001 and 2012, drawing data from the Medical Information Mart for Intensive Care–III database. Patients were included if they had at least 10 hours of available measurements after ICU admission, had at least one of every measurement used for model prediction, and had recorded race/ethnicity data. Bias was assessed through the equal opportunity difference. Model performance in terms of bias and accuracy was compared with the Modified Early Warning Score (MEWS), the Simplified Acute Physiology Score II (SAPS II), and the Acute Physiologic Assessment and Chronic Health Evaluation (APACHE).ResultsThe machine learning algorithm was found to be more accurate than all comparators, with a higher sensitivity, specificity, and area under the receiver operating characteristic. The machine learning algorithm was found to be unbiased (equal opportunity difference 0.016, P=.20). APACHE was also found to be unbiased (equal opportunity difference 0.019, P=.11), while SAPS II and MEWS were found to have significant bias (equal opportunity difference 0.038, P=.006 and equal opportunity difference 0.074, P<.001, respectively).ConclusionsThis study indicates there may be significant racial bias in commonly used severity scoring systems and that machine learning algorithms may reduce bias while improving on the accuracy of these methods.

Effect of Integrating Machine Learning Mortality Estimates With Behavioral Nudges to Clinicians on Serious Illness Conversations Among Patients With Cancer

Serious illness conversations (SICs) are structured conversations between clinicians and patients about prognosis, treatment goals, and end-of-life preferences. Interventions that increase the rate of SICs between oncology clinicians and patients may improve goal-concordant care and patient outcomes. To determine the effect of a clinician-directed intervention integrating machine learning mortality predictions with behavioral nudges on motivating clinician-patient SICs. This stepped-wedge cluster randomized clinical trial was conducted across 20 weeks (from June 17 to November 1, 2019) at 9 medical oncology clinics (8 subspecialty oncology and 1 general oncology clinics) within a large academic health system in Pennsylvania. Clinicians at the 2 smallest subspecialty clinics were grouped together, resulting in 8 clinic groups randomly assigned to the 4 intervention wedge periods. Included participants in the intention-to-treat analyses were 78 oncology clinicians who received SIC training and their patients (N = 14 607) who had an outpatient oncology encounter during the study period. (1) Weekly emails to oncology clinicians with SIC performance feedback and peer comparisons; (2) a list of up to 6 high-risk patients (≥10% predicted risk of 180-day mortality) scheduled for the next week, estimated using a validated machine learning algorithm; and (3) opt-out text message prompts to clinicians on the patient's appointment day to consider an SIC. Clinicians in the control group received usual care consisting of weekly emails with cumulative SIC performance. Percentage of patient encounters with an SIC in the intervention group vs the usual care (control) group. The sample consisted of 78 clinicians and 14 607 patients. The mean (SD) age of patients was 61.9 (14.2) years, 53.7% were female, and 70.4% were White. For all encounters, SICs were conducted among 1.3% in the control group and 4.6% in the intervention group, a significant difference (adjusted difference in percentage points, 3.3; 95% CI, 2.3-4.5; P < .001). Among 4124 high-risk patient encounters, SICs were conducted among 3.6% in the control group and 15.2% in the intervention group, a significant difference (adjusted difference in percentage points, 11.6; 95% CI, 8.2-12.5; P < .001). In this stepped-wedge cluster randomized clinical trial, an intervention that delivered machine learning mortality predictions with behavioral nudges to oncology clinicians significantly increased the rate of SICs among all patients and among patients with high mortality risk who were targeted by the intervention. Behavioral nudges combined with machine learning mortality predictions can positively influence clinician behavior and may be applied more broadly to improve care near the end of life. ClinicalTrials.gov Identifier: NCT03984773.

Validation of a Machine Learning Algorithm to Predict 180-Day Mortality for Outpatients With Cancer

Machine learning (ML) algorithms can identify patients with cancer at risk of short-term mortality to inform treatment and advance care planning. However, no ML mortality risk prediction algorithm has been prospectively validated in oncology or compared with routinely used prognostic indices. To validate an electronic health record-embedded ML algorithm that generated real-time predictions of 180-day mortality risk in a general oncology cohort. This prognostic study comprised a prospective cohort of patients with outpatient oncology encounters between March 1, 2019, and April 30, 2019. An ML algorithm, trained on retrospective data from a subset of practices, predicted 180-day mortality risk between 4 and 8 days before a patient's encounter. Patient encounters took place in 18 medical or gynecologic oncology practices, including 1 tertiary practice and 17 general oncology practices, within a large US academic health care system. Patients aged 18 years or older with outpatient oncology or hematology and oncology encounters were included in the analysis. Patients were excluded if their appointment was scheduled after weekly predictions were generated and if they were only evaluated in benign hematology, palliative care, or rehabilitation practices. Gradient-boosting ML binary classifier. The primary outcome was the patients' 180-day mortality from the index encounter. The primary performance metric was the area under the receiver operating characteristic curve (AUC). Among 24 582 patients, 1022 (4.2%) died within 180 days of their index encounter. Their median (interquartile range) age was 64.6 (53.6-73.2) years, 15 319 (62.3%) were women, 18 015 (76.0%) were White, and 10 658 (43.4%) were seen in the tertiary practice. The AUC was 0.89 (95% CI, 0.88-0.90) for the full cohort. The AUC varied across disease-specific groups within the tertiary practice (AUC ranging from 0.74 to 0.96) but was similar between the tertiary and general oncology practices. At a prespecified 40% mortality risk threshold used to differentiate high- vs low-risk patients, observed 180-day mortality was 45.2% (95% CI, 41.3%-49.1%) in the high-risk group vs 3.1% (95% CI, 2.9%-3.3%) in the low-risk group. Integrating the algorithm into the Eastern Cooperative Oncology Group and Elixhauser comorbidity index-based classifiers resulted in favorable reclassification (net reclassification index, 0.09 [95% CI, 0.04-0.14] and 0.23 [95% CI, 0.20-0.27], respectively). In this prognostic study, an ML algorithm was feasibly integrated into the electronic health record to generate real-time, accurate predictions of short-term mortality for patients with cancer and outperformed routinely used prognostic indices. This algorithm may be used to inform behavioral interventions and prompt earlier conversations about goals of care and end-of-life preferences among patients with cancer.

Development and Validation of 2 Composite Aging Measures Using Routine Clinical Biomarkers in the Chinese Population: Analyses From 2 Prospective Cohort Studies.

This study aimed to: (i) develop 2 composite aging measures in the Chinese population using 2 recent advanced algorithms (the Klemera and Doubal method and Mahalanobis distance); and (ii) validate the 2 measures by examining their associations with mortality and disease counts. Based on data from the China Nutrition and Health Survey (CHNS) 2009 wave (N = 8119, aged 20-79 years, 53.5% women), a nationwide prospective cohort study of the Chinese population, we developed Klemera and Doubal method-biological age (KDM-BA) and physiological dysregulation (PD, derived from Mahalanobis distance) using 12 biomarkers. For the validation analysis, we used Cox proportional hazard regression models (for mortality) and linear, Poisson, and logistic regression models (for disease counts) to examine the associations. We replicated the validation analysis in the China Health and Retirement Longitudinal Study (CHARLS, N = 9304, aged 45-99 years, 53.4% women). Both aging measures were predictive of mortality after accounting for age and gender (KDM-BA, per 1-year, hazard ratio [HR] = 1.14, 95% confidence interval [CI] = 1.08, 1.19; PD, per 1-SD, HR = 1.50, 95% CI = 1.33, 1.69). With few exceptions, these mortality predictions were robust across stratifications by age, gender, education, and health behaviors. The 2 aging measures were associated with disease counts both cross-sectionally and longitudinally. These results were generally replicable in CHARLS although 4 biomarkers were not available. We successfully developed and validated 2 composite aging measures-KDM-BA and PD, which have great potentials for applications in early identifications and preventions of aging and aging-related diseases in China.

Predictions of diabetes complications and mortality using hba1c variability: a 10-year observational cohort study.

Emerging evidence suggests that HbA1c variability, in addition to HbA1c itself, can be used as a predictor for mortality. The present study aims to examine the predictive power of mean HbA1c and HbA1c variability measures for diabetic complications as well as mortality. The retrospective observational study analyzed diabetic patients who were prescribed insulin at outpatient clinics of the Prince of Wales Hospital and Shatin Hospital, Hong Kong, from 1 January to 31 December, 2009. Standard deviation (SD), root mean square (RMS), and coefficient of variation were used as measures of HbA1c variability. The primary outcomes were all-cause and cardiovascular mortality. Secondary outcomes were diabetes-related complications. The study cohort consists of 3424 patients, including 3137 patients with at least three HbA1c measurements. The low mean HbA1c subgroup had significantly shorter time-to-death for all-cause mortality (P < 0.001) but not cardiovascular mortality (P = 0.920). The high Hba1c subgroup showed shorter time-to-death for all-cause (P < 0.001) and cardiovascular mortality (P < 0.001). Mean Hba1c and Hba1c variability predicted all-cause as well as cardiovascular-specific mortality. In terms of secondary outcomes, mean HbA1c and HbA1c variability significantly predicted diabetic ketoacidosis/hyperosmolar hyperglycemic state/diabetic coma, neurological, ophthalmological, and renal complications. A significant association between dichotomized HbA1c variability and hypoglycemia frequency was found (P < 0.0001). High HbA1c variability is associated with increased risk of all-cause and cardiovascular mortality, as well as diabetic complications. The association between hypoglycemic frequency, HbA1c variability, and mortality suggests that intermittent hypoglycemia resulting in poorer outcomes in diabetic patients.

Peripheral arterial tonometry as a method of measuring reactive hyperaemia correlates with organ dysfunction and prognosis in the critically ill patient: a prospective observational study.

Predictions of mortality may help in the selection of patients who benefit from intensive care. Endothelial dysfunction is partially responsible for many of the organic dysfunctions in critical illness. Reactive hyperaemia is a vascular response of the endothelium that can be measured by peripheral arterial tonometry (RH-PAT). We aimed to assess if reactive hyperaemia is affected by critical illness and if it correlates with outcomes. Prospective study with a cohort of consecutive patients admitted to an Intensive Care Unit. RH-PAT was accessed on admission and on the 7th day after admission. Early and late survivors were compared to non-survivors. The effect of RH-PAT variation on late mortality was studied by a logistic regression model. The association between RH-PAT and severity scores and biomarkers of organic dysfunction was investigated by multivariate analysis. 86 patients were enrolled. Mean ln(RHI) on admission was 0.580 and was significantly lower in patients with higher severity scores (p < 0.01) and early non-survivors (0.388; p = 0.027). The model for prediction of early-mortality estimated that each 0.1 decrease in ln(RHI) increased the odds for mortality by 13%. In 39 patients, a 2nd RH-PAT measurement was performed on the 7th day. The variation of ln(RHI) was significantly different between non-survivors and survivors (− 24.2% vs. 63.9%, p = 0.026). Ln(RHI) was significantly lower in patients with renal and cardiovascular dysfunction (p < 0.01). RH-PAT is correlated with disease severity and seems to be an independent marker of early mortality, cardiovascular and renal dysfunctions. RH-PAT variation predicts late mortality. There appears to be an RH-PAT impairment in the acute phase of severe diseases that may be reversible and associated with better outcomes.

Optimal Harvesting Decision Paths When Timber and Water Have an Economic Value in Uneven Forests

We developed an uneven-aged forest economic decision-making framework that combines: (i) a size-structured matrix model, based on growth and mortality predictions of a dynamic process-based forest landscape model, (ii) an optimal control model that determines the dynamics of control and state variables, which in turn are defined by tree harvesting and forest stock, respectively, and (iii) a water yield function that depends on changes in the leaf area index (LAI), the latter being affected by forest management. This framework was used to simulate the effects of economic-driven harvesting decisions on water yields on a catchment of South-Western Swiss Alps when both timber and water benefits are considered. Water benefits are estimated as environmental prices considering current water demands for drinking, irrigation and hydropower production. We simulated optimal harvesting decisions given the initial forest structure at each 200 m × 200 m grid cells, a set of restrictions to harvesting, and specific species survival, recruitment and growth probabilities, all of which are affected by the stand’s LAI. We applied this model using different harvesting restriction levels over a period of 20 to 40-years, and accounting for single and joint timber and water benefits. The results suggested that at the environmental prices estimated at the catchment area, water benefits have a slight influence on harvesting decisions, but when water is accounted for, harvesting decisions would include more tree species and different diameter classes, which, in principle, is expected to favor more diverse forest structures.