Performance and comparison of assessment models to predict 30-day mortality in patients with hospital-acquired pneumonia

Interpretability-based machine learning for predicting the risk of death from pulmonary inflammation in Chinese intensive care unit patients

Objective

The objective of this research was to create a machine learning predictive model that could be easily interpreted in order to precisely determine the risk of premature death in patients receiving intensive care after pulmonary inflammation.

Methods

In this study, information from the China intensive care units (ICU) Open Source database was used to examine data from 2790 patients who had infections between January 2019 and December 2020. A 7:3 ratio was used to randomly assign the whole patient population to training and validation groups. This study used six machine learning techniques: logistic regression, random forest, gradient boosting tree, extreme gradient boosting tree (XGBoost), multilayer perceptron, and K-nearest neighbor. A cross-validation grid search method was used to search the parameters in each model. Eight metrics were used to assess the models' performance: accuracy, precision, recall, F1 score, area under the curve (AUC) value, Brier score, Jordon's index, and calibration slope. The machine methods were ranked based on how well they performed in each of these metrics. The best-performing models were selected for interpretation using both the Shapley Additive exPlanations (SHAP) and Local interpretable model-agnostic explanations (LIME) interpretable techniques.

Results

A subset of the study cohort's patients (120/1668, or 7.19%) died in the hospital following screening for inclusion and exclusion criteria. Using a cross-validated grid search to evaluate the six machine learning techniques, XGBoost showed good discriminative ability, achieving an accuracy score of 0.889 (0.874-0.904), precision score of 0.871 (0.849-0.893), recall score of 0.913 (0.890-0.936), F1 score of 0.891 (0.876-0.906), and AUC of 0.956 (0.939-0.973). Additionally, XGBoost exhibited excellent performance with a Brier score of 0.050, Jordon index of 0.947, and calibration slope of 1.074. It was also possible to create an interactive internet page using the XGBoost model.

Conclusion

By identifying patients at higher risk of early mortality, machine learning-based mortality risk prediction models have the potential to significantly improve patient care by directing clinical decision making and enabling early detection of survival and mortality issues in patients with pulmonary inflammation disease.

Accuracy of physician and nurse predictions for 28-day prognosis in ICU: a single center prospective study

The proportion of correctly predicted prognoses and factors associated with prediction accuracy are unknown. The objective of this study was to explore the accuracy of physician and nurse predictions of 28-day mortality in the ICU. This was a prospective observational single-center study. All medical staff in the ICU have access to patient data, can communicate with patients or clients, and can independently predict the prognosis of patients within 24 h of patient admission. The only question of the questionnaire survey was: What is the patient's outcome on day 28 (alive or death)? There were 2155 questionnaires completed by 18 physicians and 1916 submitted by 15 nurses. In the 312 patients included, the 28-day mortality rates were predicted by physicians and nurses. The overall proportion of correct prognosis prediction was 90.1% for physicians and 64.4% for nurses (P = 0.000). There was no significant difference in the overall correct proportion and average correct proportion among physicians with different seniority levels. The overall correct proportion and average correct proportion increased among nurses with seniority. Physicians in the ICU can moderately predict 28-day mortality in critically ill patients. Nurses with a seniority of less than 10 years in ICU cannot accurately predict 28-day mortality in critically ill patients. However, the accuracy of nurses' prediction of patients' 28-day prognosis increased with their seniority in the ICU.

The Battle of the Pneumonia Predictors: A Comprehensive Meta-Analysis Comparing the Pneumonia Severity Index (PSI) and the CURB-65 Score in Predicting Mortality and the Need for ICU Support

The CURB-65 (confusion, uremia, respiratory rate, blood pressure, age ≥ 65 years) score and the pneumonia severity index (PSI) are widely used and recommended in predicting 30-day mortality and the need for intensive care support in community-acquired pneumonia. This study aims to compare the performance of these two severity scores in both mortality prediction and the need for intensive care support. A systematic review and meta-analysis was carried out, following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) 2020 guidelines, and PubMed, Scopus, ScienceDirect, and Google Scholar were searched for articles published from 2012 to 2022. The reference lists of the included studies were also searched to retrieve possible additional studies. Twenty-five studies reporting prognostic information for CURB 65 and PSI were identified. ReviewManager (RevMan) 5.4.1 was used to produce risk ratios, and a random effects model was used to pool them. Both PSI and CURB-65 showed a high strength in identifying high-risk patients. However, CURB-65 was slightly better in early mortality prediction and had more sensitivity (96.7%) and specificity (89.3%) in predicting admission to intensive care support. Thus, CURB-65 seems to be the preferred tool in predicting mortality and the need for admission into intensive care support.

Risk factors for infection and mortality caused by carbapenem-resistant Klebsiella pneumoniae: a large multicentre case-control and cohort study

OBJECTIVES

To elucidate the predictors of carbapenem-resistant Klebsiella pneumoniae (CRKP) infection and help clinicians better identify CRKP infection at an early age.

METHODS

We conducted a multicentre case-control study of 422 patients with CRKP infection and 948 with carbapenem-susceptible K. pneumoniae (CSKP) infection from March to July 2017. Binary logistic regression was used to identify risk factors for CRKP infection. The subgroups of CRKP respiratory infection, intra-abdominal infection, and bloodstream infection were also evaluated. Patients were followed up for 28 days. Independent risk factors for 28-day crude mortality of CRKP infection were analysed using Cox proportional hazards regression models.

RESULTS

Longer stay of hospitalization, stay in the ICU, previous exposure to antibacterial agents (especially carbapenems, quinolones, aminoglycosides, and tigecycline), invasive procedures, intravascular catheter use, tracheotomy, or admission to ICU in the preceding 90 days were risk factors for CRKP infection. Carbapenem exposure was the only common predictor of different types of CRKP infection. The 28-day mortality of CRKP infection was 13.1% and was independently associated with sex, admitted unit, and types of infection.

CONCLUSIONS

Strict policies for antibiotic use, cautious decisions regarding the implementation of invasive procedures, and careful management of patients with catheters, especially intravascular catheters, are necessary to handle CRKP infection.

MiR-940 Serves as a Diagnostic Biomarker in Patients with Sepsis and Regulates Sepsis-Induced Inflammation and Myocardial Dysfunction

Introduction

Sepsis is a heterogeneous syndrome with a life-long threat caused by infection. This study aimed to investigate the clinical function of miR-940 and its influence on cardiomyocyte models.

Methods

The relative expression of miR-940 was assessed by qRT-PCR and the roles in the clinical diagnosis of miR-940 were revealed by the ROC curve. The relationship between miR-940 and clinical parameters was validated by Pearson analysis. The sepsis rat models were established by treatment with cecal ligation and perforation (CLP) and clinical items including left ventricular systolic pressure (LVSP), left ventricular and end-diastolic pressure (LVEDP), maximum rate of increase/decrease in left ventricular blood pressure (± dp/dt_max) as well as troponin (cTnl), creatine kinase isoenzyme (CK-MB), TNF-α, IL-1β, and IL-6 were detected.

Results

The finding of qRT-PCR accentuated that the relative expression of miR-940 was significantly decreased in sepsis patients and CLP-stimulated models. The ROC curve proposed that miR-940 could be a satisfactory diagnostic biomarker for sepsis patients. Pearson analysis reinforced the expression of miR-940 was negatively associated with the PCT, WBC, CRP, Scr, SOFA score, and APACHE II score. The outcome of CLP-steered rat verified that overexpression of miR-940 inhibited the detrimental effects of CLP on myocardial dysfunction and inflammation reactions.

Conclusion

The downregulation of miR-940 was reported and it might be an underlying diagnostic marker in sepsis patients. Overexpression of miR-940 protected myocardial function from damage and inflammation induced by CLP.

Association between IL-18 gene polymorphisms and Hashimoto thyroiditis

BACKGROUND

Hashimoto's thyroiditis (HT), which is also called lymphocytic thyroiditis, is the most frequent autoimmune thyroid disease (AITD), in which T helper-1 lymphocytes mediate the disease. IL-18 is expressed in thyroid follicular cells (TFCs) during HT. The findings of studies aimed at investigating the relationship between IL-18 and HT are highly contradictory. In this study, we aimed to investigate the association between IL-18 gene polymorphisms and HT.

METHODS AND RESULTS

The study included 97 patients diagnosed with HT and 86 volunteers in the healthy control group. The IL18-607C/A (rs1946518) and -137G/C (rs187238) genotypes were determined using the polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) method. No significant difference in the mean age or sex was observed between the groups (p = 0.763 and p = 0.658, respectively). The -137 IL18 CG genotype was more frequent in HT patients than in controls. In HT patients, the risk of the IL-18 CG genotype was more than 2.237 times higher (OR 2.237%95 Cl 1.195-4.187, p = 0.039) than that of the G/G genotype. Additionally, the -607 AC genotype was more frequent in the control group than in the HT group (in individuals with the IL18 CG genotype).

CONCLUSIONS

According to our results, the CG genotype might be a risk factor for HT. Conversely, there is a possibility that the AC genotype plays a protective role against the condition. However, further studies will contribute to new solutions by revealing the molecular and cellular mechanisms of HT.

Prediction of Mortality in Patients After Oncologic Gastrointestinal Surgery: Comparison of the ASA, APACHE II, and POSSUM Scoring Systems

Scoring systems have been developed to predict the expected mortality and morbidity in surgical procedures. In this study, our aim was to compare the ASA (American Society of Anesthesiologists), APACHE (Acute Physiology and Chronic Health Evaluation) II, POSSUM (Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity) scoring systems as predictors of mortality in patients who underwent gastrointestinal oncologic surgery, followed, and were admitted to the intensive care unit during the postoperative period. We examined the files of 82 patients who underwent oncologic gastrointestinal surgery and followed up in the intensive care units (ICUs). The patients’ APACHE II scores and predicted mortality rates (PMR) according to the APACHE II, POSSUM, and ASA scores were ​​calculated. The receiver operator characteristic (ROC) curve analysis was used when evaluating the performances of the ASA, APACHE, and POSSUM scoring systems in terms of accurate assessment of mortality. Accordingly, the area under the curve (AUC) = 0.5 no distinction, 0.5 <AUC <0.7 discriminative power of the test is statistically not significant, 0.7 <AUC <0.8 acceptable, 0.8 <AUC <0.9 very good and 0.9 <AUC <1 perfect. The evaluations showed that APACHE II had the best performance with 0.81, followed by POSSUM, which had an acceptable level at 0.78. On the other hand, the ASA score was 0.63 and its discriminative power was identified as statistically insignificant. Our results show that the POSSUM and APACHE II scoring systems were better at predicting mortality than the ASA scoring system for the prediction of mortality in the postoperative period. Both the POSSUM and APACHE II scoring systems can be confidently used for the prediction of mortality in patients undergoing operations due to oncologic gastrointestinal diseases.