Predicting Sepsis Mortality in a Population-Based National Database: Machine Learning Approach.

BACKGROUND: Although machine learning (ML) algorithms have been applied to point-of-care sepsis prognostication, ML has not been used to predict sepsis mortality in an administrative database. Therefore, we examined the performance of common ML algorithms in predicting sepsis mortality in adult patients with sepsis and compared it with that of the conventional context knowledge-based logistic regression approach. OBJECTIVE: The aim of this study is to examine the performance of common ML algorithms in predicting sepsis mortality in adult patients with sepsis and compare it with that of the conventional context knowledge-based logistic regression approach. METHODS: We examined inpatient admissions for sepsis in the US National Inpatient Sample using hospitalizations in 2010-2013 as the training data set. We developed four ML models to predict in-hospital mortality: logistic regression with least absolute shrinkage and selection operator regularization, random forest, gradient-boosted decision tree, and deep neural network. To estimate their performance, we compared our models with the Super Learner model. Using hospitalizations in 2014 as the testing data set, we examined the models' area under the receiver operating characteristic curve (AUC), confusion matrix results, and net reclassification improvement. RESULTS: Hospitalizations of 923,759 adults were included in the analysis. Compared with the reference logistic regression (AUC: 0.786, 95% CI 0.783-0.788), all ML models showed superior discriminative ability (P<.001), including logistic regression with least absolute shrinkage and selection operator regularization (AUC: 0.878, 95% CI 0.876-0.879), random forest (AUC: 0.878, 95% CI 0.877-0.880), xgboost (AUC: 0.888, 95% CI 0.886-0.889), and neural network (AUC: 0.893, 95% CI 0.891-0.895). All 4 ML models showed higher sensitivity, specificity, positive predictive value, and negative predictive value compared with the reference logistic regression model (P<.001). We obtained similar results from the Super Learner model (AUC: 0.883, 95% CI 0.881-0.885). CONCLUSIONS: ML approaches can improve sensitivity, specificity, positive predictive value, negative predictive value, discrimination, and calibration in predicting in-hospital mortality in patients hospitalized with sepsis in the United States. These models need further validation and could be applied to develop more accurate models to compare risk-standardized mortality rates across hospitals and geographic regions, paving the way for research and policy initiatives studying disparities in sepsis care.

Use of Calcium Channel Blockers and Risk of Active Tuberculosis Disease: A Population-Based Analysis.

Calcium channel blockers (CCBs) are known to reduce the availability of iron-an important mineral for intracellular pathogens. Nonetheless, whether the use of CCBs modifies the risk of active tuberculosis in the clinical setting remains unclear. To determine whether CCBs may modify the risk of active tuberculosis disease, we conducted a nested case-control study using the National Health Insurance Research Database of Taiwan between January 1999 and December 2011. Conditional logistic regression and disease risk score adjustment were used to calculate the risk of active tuberculosis disease associated with CCB use. Subgroup analyses investigated the effect of different types of CCBs and potential effect modification in different subpopulations. A total of 8164 new active tuberculosis cases and 816 400 controls were examined. Use of CCBs was associated with a 32% decrease in the risk of active tuberculosis (relative risk [RR], 0.68 [95% CI, 0.58-0.78]) after adjustment with disease risk score. Compared with nonuse of CCBs, the use of dihydropyridine CCBs was associated with a lower risk of tuberculosis (RR, 0.63 [95% CI, 0.53-0.79]) than nondihydropyridine CCBs (RR, 0.73 [95% CI, 0.57-0.94]). In contrast, use of ß-blockers (RR, 0.99 [95% CI, 0.83-1.12]) or loop diuretics (RR, 0.88 [95% CI, 0.62-1.26]) was not associated with lower risk of tuberculosis. In subgroup analyses, the risk of tuberculosis associated with the use of CCBs was similar among patients with heart failure or cerebrovascular diseases. Our study confirms that use of dihydropyridine CCBs decreases the risk of active tuberculosis.

High-sensitivity-cardiac troponin for accelerated diagnosis of acute myocardial infarction: A systematic review and meta-analysis.

OBJECTIVES: Cardiovascular disease is the leading cause of mortality and morbidity. Serial troponin tests have been endorsed as essential diagnostic steps to rule out/-in acute myocardial infarction (AMI), and hs-cTn assays have shown promise in enhancing the accuracy and efficiency of AMI diagnosis in the emergency department (ED). METHODS: A systematic review and meta-analysis of diagnostic test accuracy studies were conducted to compare the diagnostic performance of various accelerated diagnostic algorithms of hs-cTn assays for patients with symptoms of AMI. Random-effects bivariate meta-analysis was conducted to estimate the summary sensitivity, specificity, likelihood ratios, and area under receiver operating characteristic curve. RESULTS: In the systematic review consisting of 56 studies and 67,945 patients, both hs-cTnT and hs-cTnI-based 0-, 1-, 2- and 0-1 h algorithms showed a pooled sensitivity >90%. The hs-cTnI-based algorithm showed a pooled specificity >80%. The hs-cTnT-based algorithms had a specificity of 68% for the 0-h algorithm and of around 80% for the 1-, 2-, and 0-1 h algorithms. The heterogeneities of all diagnostic algorithms were mild (I2 < 50%). CONCLUSION: Both hs-cTnI- and hs-cTnT-based accelerated diagnostic algorithms have high sensitivities but moderate specificities for early diagnosis of AMI. Overall, hs-cTnI-based algorithms have slightly higher specificities in early diagnosis of AMI. For patients presenting ED with typical symptoms, the use of hs-cTnT or hs-cTnI assays at the 99th percentile may help identify patients with low risk for AMI and promote early discharge from the ED.

Effect of Renin-Angiotensin-Aldosterone System Inhibitors on Short-Term Mortality After Sepsis: A Population-Based Cohort Study.

Antagonists of the renin-angiotensin-aldosterone system (RAAS), including ACEIs (angiotensin-converting enzyme inhibitors) and ARBs (angiotensin II receptor blockers), may prevent organ failure. We, therefore, investigated whether specific RAAS inhibitors are associated with reduced mortality in patients with sepsis.We conducted a population-based retrospective cohort study using multivariable propensity score-based regression to control for differences among patients using different RAAS inhibitors. A multivariable-adjusted Cox proportional-hazards regression model was used to determine the association between RAAS inhibitors and sepsis outcomes. To directly compare ACEI users, ARB users, and nonusers, a 3-way propensity score matching approach was performed. Results were pooled with previous evidence via a random-effects meta-analysis. A total of 52 727 patients were hospitalized with sepsis, of whom 7642 were prescribed an ACEI and 4237 were prescribed an ARB. Using propensity score-matched analyses, prior ACEI use was associated with decreased 30-day mortality (hazard ratio, 0.84 [95% CI, 0.75-0.94]) and 90-day mortality (hazard ratio, 0.83 [95% CI, 0.75-0.92]) compared with nonuse. Prior ARB use was associated with an improved 90-day survival (hazard ratio, 0.88 [95% CI, 0.83-0.94]). These results persisted in sensitivity analyses focusing on patients without cancer and patients with hypertension. By contrast, no beneficial effect was found for antecedent ß-blockers exposure (hazard ratio, 0.99 [95% CI, 0.94-1.05]). The pooled estimates obtained from the meta-analysis was 0.71 (95% CI, 0.58-0.87) for prior use of ACEI/ARB.The short-term mortality after sepsis was substantially lower among those who were already established on RAAS inhibitor treatment when sepsis occurred.