Machine learning models to evaluate mortality in pediatric patients with pneumonia in the intensive care unit.

OBJECTIVES: This study aimed to predict mortality in children with pneumonia who were admitted to the intensive care unit (ICU) to aid decision-making. STUDY DESIGN: Retrospective cohort study conducted at a single tertiary hospital. PATIENTS: This study included children who were admitted to the pediatric ICU at the National Taiwan University Hospital between 2010 and 2019 due to pneumonia. METHODOLOGY: Two prediction models were developed using tree-structured machine learning algorithms. The primary outcomes were ICU mortality and 24-h ICU mortality. A total of 33 features, including demographics, underlying diseases, vital signs, and laboratory data, were collected from the electronic health records. The machine learning models were constructed using the development data set, and performance matrices were computed using the holdout test data set. RESULTS: A total of 1231 ICU admissions of children with pneumonia were included in the final cohort. The area under the receiver operating characteristic curves (AUROCs) of the ICU mortality model and 24-h ICU mortality models was 0.80 (95% confidence interval [CI], 0.69-0.91) and 0.92 (95% CI, 0.86-0.92), respectively. Based on feature importance, the model developed in this study tended to predict increased mortality for the subsequent 24 h if a reduction in the blood pressure, peripheral capillary oxygen saturation (SpO2), or higher partial pressure of carbon dioxide (PCO2) were observed. CONCLUSIONS: This study demonstrated that the machine learning models for predicting ICU mortality and 24-h ICU mortality in children with pneumonia have the potential to support decision-making, especially in resource-limited settings.

Clinical characteristics of hospitalized children with community-acquired pneumonia and respiratory infections: Using machine learning approaches to support pathogen prediction at admission.

BACKGROUND: Acute respiratory infections (ARIs) are common in children. We developed machine learning models to predict pediatric ARI pathogens at admission. METHODS: We included hospitalized children with respiratory infections between 2010 and 2018. Clinical features were collected within 24 h of admission to construct models. The outcome of interest was the prediction of 6 common respiratory pathogens, including adenovirus, influenza virus types A and B, parainfluenza virus (PIV), respiratory syncytial virus (RSV), and Mycoplasma pneumoniae (MP). Model performance was estimated using area under the receiver operating characteristic curve (AUROC). Feature importance was measured using Shapley Additive exPlanation (SHAP) values. RESULTS: A total of 12,694 admissions were included. Models trained with 9 features (age, event pattern, fever, C-reactive protein, white blood cell count, platelet count, lymphocyte ratio, peak temperature, peak heart rate) achieved the best performance (AUROC: MP 0.87, 95% CI 0.83-0.90; RSV 0.84, 95% CI 0.82-0.86; adenovirus 0.81, 95% CI 0.77-0.84; influenza A 0.77, 95% CI 0.73-0.80; influenza B 0.70, 95% CI 0.65-0.75; PIV 0.73, 95% CI 0.69-0.77). Age was the most important feature to predict MP, RSV and PIV infections. Event patterns were useful for influenza virus prediction, and C-reactive protein had the highest SHAP value for adenovirus infections. CONCLUSION: We demonstrate how artificial intelligence can assist clinicians identify potential pathogens associated with pediatric ARIs upon admission. Our models provide explainable results that could help optimize the use of diagnostic testing. Integrating our models into clinical workflows may lead to improved patient outcomes and reduce unnecessary medical costs.

Using Text Content From Coronary Catheterization Reports to Predict 5-Year Mortality Among Patients Undergoing Coronary Angiography: A Deep Learning Approach.

Background: Current predictive models for patients undergoing coronary angiography have complex parameters which limit their clinical application. Coronary catheterization reports that describe coronary lesions and the corresponding interventions provide information of the severity of the coronary artery disease and the completeness of the revascularization. This information is relevant for predicting patient prognosis. However, no predictive model has been constructed using the text content from coronary catheterization reports before. Objective: To develop a deep learning model using text content from coronary catheterization reports to predict 5-year all-cause mortality and 5-year cardiovascular mortality for patients undergoing coronary angiography and to compare the performance of the model to the established clinical scores. Method: This retrospective cohort study was conducted between January 1, 2006, and December 31, 2015. Patients admitted for coronary angiography were enrolled and followed up until August 2019. The main outcomes were 5-year all-cause mortality and 5-year cardiovascular mortality. In total, 11,576 coronary catheterization reports were collected. BioBERT (bidirectional encoder representations from transformers for biomedical text mining), which is a BERT-based model in the biomedical domain, was utilized to construct the model. The area under the receiver operating characteristic curve (AUC) was used to assess model performance. We also compared our results to the residual SYNTAX (SYNergy between PCI with TAXUS and Cardiac Surgery) score. Results: The dataset was divided into the training (60%), validation (20%), and test (20%) sets. The mean age of the patients in each dataset was 65.5 ± 12.1, 65.4 ± 11.2, and 65.6 ± 11.2 years, respectively. A total of 1,411 (12.2%) patients died, and 664 (5.8%) patients died of cardiovascular causes within 5 years after coronary angiography. The best of our models had an AUC of 0.822 (95% CI, 0.790-0.855) for 5-year all-cause mortality, and an AUC of 0.858 (95% CI, 0.816-0.900) for 5-year cardiovascular mortality. We randomly selected 300 patients who underwent percutaneous coronary intervention (PCI), and our model outperformed the residual SYNTAX score in predicting 5-year all-cause mortality (AUC, 0.867 [95% CI, 0.813-0.921] vs. 0.590 [95% CI, 0.503-0.684]) and 5-year cardiovascular mortality (AUC, 0.880 [95% CI, 0.873-0.925] vs. 0.649 [95% CI, 0.535-0.764]), respectively, after PCI among these patients. Conclusions: We developed a predictive model using text content from coronary catheterization reports to predict the 5-year mortality in patients undergoing coronary angiography. Since interventional cardiologists routinely write reports after procedures, our model can be easily implemented into the clinical setting.

Clinical characteristics of influenza with or without Streptococcus pneumoniae co-infection in children.

BACKGROUND/PURPOSE: Influenza is frequently complicated with bacterial co-infection. This study aimed to disclose the significance of Streptococcus pneumoniae co-infection in children with influenza. METHODS: We retrospectively reviewed medical records of pediatric patients hospitalized for influenza with or without pneumococcal co-infection at the National Taiwan University Hospital from 2007 to 2019. Clinical characteristics and outcomes were compared between patients with and without S. pneumoniae co-infection. RESULTS: There were 558 children hospitalized for influenza: 494 had influenza alone whereas 64 had S. pneumoniae co-infection. Patients with S. pneumoniae co-infection had older ages, lower SpO2, higher C-Reactive Protein (CRP), lower serum sodium, lower platelet counts, more chest radiograph findings of patch and consolidation on admission, longer hospitalization, more intensive care, longer intensive care unit (ICU) stay, more mechanical ventilation, more inotropes/vasopressors use, more surgical interventions including video-assisted thoracoscopic surgery (VATS) and extracorporeal membrane oxygenation (ECMO), and higher case-fatality rate. CONCLUSION: Compared to influenza alone, patients with S. pneumoniae co-infection had more morbidities and mortalities. Pneumococcal co-infection is considered when influenza patients have lower SpO2, lower platelet counts, higher CRP, lower serum sodium, and more radiographic patches and consolidations on admission.

Clinical characteristics of recurrent pneumonia in children with or without underlying diseases.

BACKGROUND: Recurrent pneumonia is uncommon in children and few studies investigate the clinical impact of underlying diseases on this issue. This study aimed to explore the difference in clinical manifestations, pathogens, and prognosis of recurrent pneumonia in children with or without underlying diseases. METHODS: We conducted a retrospective study of pediatric recurrent pneumonia from 2007 to 2019 in National Taiwan University Hospital. Patients under the age of 18 who had two or more episodes of pneumonia in a year were included, and the minimum interval of two pneumonia episodes was more than one month. Aspiration pneumonia was excluded. Demographic and clinical characteristics of patients were collected and compared. RESULTS: Among 8508 children with pneumonia, 802 (9.4%) of them had recurrent pneumonia. Among these 802 patients, 655 (81.7%) had underlying diseases including neurological disorders (N = 252, 38.5%), allergy (N = 211, 32.2%), and cardiovascular diseases (N = 193, 29.5%). Children without underlying diseases had more viral bronchopneumonia (p < 0.001). Children with underlying diseases were more likely to acquire Staphylococcus aureus (p = 0.001), and gram-negative bacteriae, more pneumonia episodes (3 vs 2, p < 0.001), a longer hospital stay (median: 7 vs. 4 days, p < 0.001), a higher ICU rate (28.8% vs 3.59%, p < 0.001), and a higher case-fatality rate (5.19% vs 0%, p < 0.001) than those without underlying diseases. CONCLUSION: Children with underlying diseases, prone to have recurrent pneumonia and more susceptible to resistant microorganisms, had more severe diseases and poorer clinical outcomes. Therefore, more attention may be paid on clinical severity and the therapeutic plan.

RSV pneumonia with or without bacterial co-infection among healthy children.

BACKGROUND: Respiratory syncytial virus (RSV) is a common cause of childhood pneumonia, but there is limited understanding of whether bacterial co-infections affect clinical severity. METHODS: We conducted a retrospective cohort study at National Taiwan University Hospital from 2010 to 2019 to compare clinical characteristics and outcomes between RSV with and without bacterial co-infection in children without underlying diseases, including length of hospital stay, intensive care unit (ICU) admission, ventilator use, and death. RESULTS: Among 620 inpatients with RSV pneumonia, the median age was 1.33 months (interquartile range, 0.67-2 years); 239 (38.6%) under 1 year old; 366 (59.0%) males; 201 (32.4%) co-infected with bacteria. The three most common bacteria are Streptococcus pneumoniae, Staphylococcus aureus and Haemophilus influenzae. The annually seasonal analysis showed that spring and autumn were peak seasons, and September was the peak month. Compared with single RSV infection, children with bacterial co-infection were younger (p = 0.021), had longer hospital stay (p < 0.001), needed more ICU care (p = 0.02), had higher levels of C-reactive protein (p = 0.009) and more frequent hyponatremia (p = 0.013). Overall, younger age, bacterial co-infection (especially S. aureus), thrombocytosis, and lower hemoglobin level were associated with the risk of requiring ICU care. CONCLUSION: RSV related bacterial co-infections were not uncommon and assoicated with ICU admission, especially for young children, and more attention should be given. For empirical antibacterial treatment, high-dose amoxicillin-clavulanic acid or ampicillin-sulbactam was recommended for non-severe cases; vancomycin and third-generation cephalosporins were suggested for critically ill patients requiring ICU care.

Preoperative biliary drainage associated with biliary stricture after pancreaticoduodenectomy: a population-based study.

BACKGROUND: The rate of preoperative biliary drainage for pancreaticoduodenectomy has been increasing despite most recent evidence that favors avoiding it. Only a few studies have focused on late surgical complications - biliary stricture after pancreaticoduodenectomy and have produced only inconclusive results. We evaluate the role of preoperative biliary drainage in the formation of biliary stricture after pancreaticoduodenectomy. METHODS: The Taiwan National Health Insurance Program is a mandatory health care plan that covers nearly the entire population of 23 million in this country. A retrospective study was conducted to analyze the database compiled by the Taiwan National Health Insurance between January 2000 and December 2011. We included only patients with at least 2 years of follow-up. A cohort of 2,087 patients with preoperative diagnosis of biliary obstruction that underwent pancreaticoduodenectomy was evaluated. RESULTS: A total of 212 (10.1%) of the 2,087 studied patients needed intervention for biliary stricture after pancreaticoduodenectomy. The median time to biliary stricture formation was 15.2 months (range: 1.2-89.7 months). The cumulative biliary stricture rate was 6.9% (1 year), 15.8% (5 years), and 18.5% (10 year). Multivariate analysis showed preoperative biliary drainage (hazard ratio 1.78, 95% CI 1.27-2.50, P = 0.001) associated with biliary stricture after pancreaticoduodenectomy. CONCLUSIONS: Preoperative biliary drainage increases biliary stricture rate after pancreaticoduodenectomy.