Prediction of hospital mortality among critically ill patients in a single centre in Asia: comparison of artificial neural networks and logistic regression-based model.

INTRODUCTION: This study compared the performance of the artificial neural network (ANN) model with the Acute Physiologic and Chronic Health Evaluation (APACHE) II and IV models for predicting hospital mortality among critically ill patients in Hong Kong. METHODS: This retrospective analysis included all patients admitted to the intensive care unit of Pamela Youde Nethersole Eastern Hospital from January 2010 to December 2019. The ANN model was constructed using parameters identical to the APACHE IV model. Discrimination performance was assessed using area under the receiver operating characteristic curve (AUROC); calibration performance was evaluated using the Brier score and Hosmer-Lemeshow statistic. RESULTS: In total, 14 503 patients were included, with 10% in the validation set and 90% in the ANN model development set. The ANN model (AUROC=0.88, 95% confidence interval [CI]=0.86-0.90, Brier score=0.10; P in Hosmer-Lemeshow test=0.37) outperformed the APACHE II model (AUROC=0.85, 95% CI=0.80-0.85, Brier score=0.14; P<0.001 for both comparisons of AUROCs and Brier scores) but showed performance similar to the APACHE IV model (AUROC=0.87, 95% CI=0.85-0.89, Brier score=0.11; P=0.34 for comparison of AUROCs, and P=0.05 for comparison of Brier scores). The ANN model demonstrated better calibration than the APACHE II and APACHE IV models. CONCLUSION: Our ANN model outperformed the APACHE II model but was similar to the APACHE IV model in terms of predicting hospital mortality in Hong Kong. Artificial neural networks are valuable tools that can enhance real-time prognostic prediction.

Impact of appropriate empirical antibiotics on clinical outcomes in Klebsiella pneumoniae bacteraemia.

INTRODUCTION: Klebsiella pneumoniae infections can cause septic shock, multiorgan failure, and death. METHODS: This retrospective cohort study included adults with K pneumoniae bacteraemia treated from 1 January 2009 to 30 June 2017. Demographics, microbiology, and outcomes were analysed. The primary outcome was 90-day all-cause mortality; secondary outcomes were intensive care unit (ICU) and hospital mortalities, ICU and hospital lengths of stay, and ICU ventilator duration. RESULTS: In total, 984 patients had K pneumoniae bacteraemia; of them, 686 received appropriate empirical antibiotics. Overall, 205 patients required intensive care. Older age (odds ratio [OR]=1.60; 95% confidence interval [CI]=1.120-2.295; P=0.010), chronic kidney disease (OR=1.81; 95% CI=1.181- 2.785; P=0.007), mechanical ventilation (OR=1.79; 95% CI=1.188-2.681; P=0.005), pneumonia (OR=1.50; 95% CI=1.030-2.187; P=0.034), and carbapenem-resistant or extended-spectrum betalactamase (ESBL)-producing isolates (OR=12.51; 95% CI=7.886-19.487; P<0.001) were associated with greater risk of inappropriate empirical treatment. Ninety-day mortality was significantly higher among patients with inappropriate empirical treatment; independent predictors included pneumonia (hazard ratio [HR]=2.94; 95% CI=2.271-3.808; P<0.001), gastrointestinal infection (HR=2.77; 95% CI=2.055-3.744; P<0.001), failed empirical antibiotics (HR=2.45; 95% CI=1.928-3.124; P<0.001), older age (HR=1.79; 95% CI=1.356-2.371; P<0.001), solid tumour (HR=1.77; 95% CI=1.401-2.231; P<0.001), carbapenem-resistant or ESBL-producing isolates (HR=1.64; 95% CI=1.170-2.297; P=0.004), patients admitted through the Department of Medicine (HR=1.39; 95% CI=1.076-1.800; P=0.012), and higher total Sequential Organ Failure Assessment score (HR=1.09; 95% CI=1.058-1.112; P=0.023). Among ICU patients, inappropriate empirical antibiotic treatment was significantly associated with increased ventilator duration and 90-day mortality. CONCLUSIONS: Klebsiella pneumoniae bacteraemia was associated with high 90-day and ICU mortalities; 90-day mortality increased with inappropriate empirical antibiotic treatment.

Burden of pneumococcal disease: 8-year retrospective analysis from a single centre in Hong Kong.

PURPOSE: Streptococcus pneumoniae is a common pathogen involved in community-acquired pneumonia. Invasive pneumococcal disease is often associated with higher co-morbidity rates, but mortality-related findings have been inconclusive. This study investigated predictors of 30-day mortality and invasive pneumococcal disease. METHODS: This retrospective analysis included adults with pneumococcal disease who were admitted to Pamela Youde Nethersole Eastern Hospital from 1 January 2011 to 31 December 2018. Demographics, microbiological characteristics, and outcomes were compared between 30-day survivors and non-survivors, and between patients with invasive disease and those with non-invasive disease. Intensive care unit (ICU) subgroup analysis was performed. The primary outcome was 30-day all-cause mortality; secondary outcomes were ICU and hospital mortalities, and ICU and hospital lengths of stay. RESULTS: In total, 792 patients had pneumococcal disease; 701 survived and 91 (11.5%) died within 30 days. Notably, 106 (13.4%) patients had invasive pneumococcal disease and 170 (21.5%) patients received intensive care. Vasopressor use (odds ratio [OR]=4.96, P<0.001), chronic kidney disease (OR=3.62, P<0.001), positive urinary antigen test results (OR=2.57, P=0.001), and advanced age (OR=2.19, P=0.010) were independent predictors for 30-day mortality by logistic regression analysis. Among critically ill patients, chronic kidney disease (OR=4.64, P<0.001), higher APACHE IV score (OR=3.73, P=0.016), and positive urinary antigen test results (OR=2.94, P=0.008) were predictors for 30-day mortality. Logistic regression analysis revealed that chronic kidney disease (OR=3.10, P<0.001) was a risk factor for invasive pneumococcal disease. CONCLUSION: Advanced age, vasopressor use, chronic kidney disease, and positive urinary antigen test results were independent predictors for 30-day mortality in patients with pneumococcal disease.

A case of severe empyema with acute respiratory distress syndrome caused by Slackia exigua requiring veno-venous extracorporeal membrane oxygenation.

Slackia exigua (S. exigua) is an obligatory anaerobic coccobacillus under the family of Coriobacteriaceae. It is a rare cause of pyogenic extraoral infections. We report a 58-year-old lady with good past health presented with fulminant community-acquired pneumonia causing acute respiratory distress syndrome caused by S. exigua requiring veno-venous extra-corporeal membrane oxygenation (VV-ECMO). Bacterial identification can be challenging and often require 16 S rRNA and MALDI-TOF MS. The patient was treated with amoxicillin-clavulanic acid according to sensitivity and made significant recovery.

Clinical predictors and outcomes of Klebsiella pneumoniae bacteraemia in a regional hospital in Hong Kong.

BACKGROUND: Klebsiella pneumoniae (KP) infection is associated with high morbidity and mortality. Multidrug resistance, especially extended-spectrum ß-lactamase (ESBL) production, in KP is endemic worldwide. AIM: To evaluate the clinical characteristics and outcomes of patients with KP bacteraemia in critical care and general ward settings. METHODS: Adult patients admitted to a regional hospital in Hong Kong from January 1st, 2009 to June 30th, 2016 (7.5 years) with KP bacteraemia were included. Demographics, clinical features, microbiological characteristics, and outcomes were analysed. FINDINGS: Among 853 patients, 178 (20.9%) required critical care and 176 (20.6%) died within 30 days of hospital admission. Thirty-day survivors were younger (P<0.001), had milder disease (defined by Sequential Organ Failure Assessment score) (P<0.001), presented with hepatobiliary sepsis (P<0.001) or urosepsis (P<0.001), less septic shock (P=0.013), fewer invasive organ supports (P<0.001), and had appropriate empirical antibiotics (P<0.001). Cox regression analysis showed that respiratory tract infection (hazard ratio: 2.99; 95% confidence interval: 2.061-4.337; P≤0.001), gastrointestinal tract infection (excluding hepatobiliary system) (2.763; 1.761-4.337; P≤0.001), mechanical ventilation (2.202; 1.506-3.221; P≤0.001), medical case (1.830; 1.253-2.672; P=0.002), inappropriate empirical antibiotics (1.716; 1.267-2.324; P≤0.001), female (1.699; 1.251-2.307; P<0.001), age >65 years (1.692; 1.160-2.467; P=0.006), and presence of solid tumour (1.457; 1.056-2.009; P=0.022) were independent risk factors for 30-day mortality. Unexpectedly, diabetes mellitus was associated with better 30-day survival (P=0.002). A total of 102 patients (12.0%) had infections with ESBL-producing strains, which were not associated with higher 30-day mortality. CONCLUSION: KP bacteraemia is associated with high 30-day mortality. Site of infection, patients' comorbidities and appropriate use of empirical antibiotic are important predictors of patients' outcomes.