B-type natriuretic peptide in predicting the severity of community-acquired pneumonia

An explainable machine learning-based model to predict intensive care unit admission among patients with community-acquired pneumonia and connective tissue disease

BACKGROUND

There is no individualized prediction model for intensive care unit (ICU) admission on patients with community-acquired pneumonia (CAP) and connective tissue disease (CTD) so far. In this study, we aimed to establish a machine learning-based model for predicting the need for ICU admission among those patients.

METHODS

This was a retrospective study on patients admitted into a University Hospital in China between November 2008 and November 2021. Patients were included if they were diagnosed with CAP and CTD during admission and hospitalization. Data related to demographics, CTD types, comorbidities, vital signs and laboratory results during the first 24 h of hospitalization were collected. The baseline variables were screened to identify potential predictors via three methods, including univariate analysis, least absolute shrinkage and selection operator (Lasso) regression and Boruta algorithm. Nine supervised machine learning algorithms were used to build prediction models. We evaluated the performances of differentiation, calibration, and clinical utility of all models to determine the optimal model. The Shapley Additive Explanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) techniques were performed to interpret the optimal model.

RESULTS

The included patients were randomly divided into the training set (1070 patients) and the testing set (459 patients) at a ratio of 70:30. The intersection results of three feature selection approaches yielded 16 predictors. The eXtreme gradient boosting (XGBoost) model achieved the highest area under the receiver operating characteristic curve (AUC) (0.941) and accuracy (0.913) among various models. The calibration curve and decision curve analysis (DCA) both suggested that the XGBoost model outperformed other models. The SHAP summary plots illustrated the top 6 features with the greatest importance, including higher N-terminal pro-B-type natriuretic peptide (NT-proBNP) and C-reactive protein (CRP), lower level of CD4 + T cell, lymphocyte and serum sodium, and positive serum (1,3)-β-D-glucan test (G test).

CONCLUSION

We successfully developed, evaluated and explained a machine learning-based model for predicting ICU admission in patients with CAP and CTD. The XGBoost model could be clinical referenced after external validation and improvement.

Usefulness of a predictive model to hospitalize patients with low-risk community-acquired pneumonia

INTRODUCTION

A high proportion of patients with low-risk community-acquired pneumonia (CAP) (classes I-III of the Pneumonia Severity Index) are hospitalized. The purpose of this study was to determine whether validated severity scales are used in clinical practice to make admission decisions, identify the variables that influence this decision, and evaluate the potential predictive value of these variables.

MATERIALS AND METHODS

A prospective, observational study of patients ≥ 18 years of age with a diagnosis of low-risk CAP hospitalized or referred from the Emergency Department to outpatient consultations. A multivariate logistic regression predictive model was built to predict the decision to hospitalize a patient.

RESULTS

The study population was composed of 1,208 patients (806 inpatients and 402 outpatients). The severity of CAP was estimated in 250 patients (20.7%). The factors that determined hospitalization were "abnormal findings in complementary studies" (643/806: 79.8%; due to respiratory failure in 443 patients) and "signs of clinical deterioration" [64/806 (7.9%): hypotension (16/64, 25%); hemoptoic expectoration (12/64, 18.8%); tachypnea (10/64, 15.6%)]. In total, ambulatory management was not contraindicated in 24.7% of hospitalized patients (199). The predictive model built to decide about hospitalization had a good power of discrimination (AUC 0.876; 95%CI: 0.855-0.897).

CONCLUSIONS

Scales are rarely used to estimate the severity of CAP at the emergency department. The decision to hospitalize or not a patient largely depends on the clinical experience of the physician. Our predictive model showed a good power to discriminate the patients who required hospitalization. Further studies are warranted to validate these results.

Validation of risk prediction for outcomes of severe community-acquired pneumonia among under-five children in Amhara region, Northwest Ethiopia

BACKGROUND

Globally there are over 1,400 cases of pneumonia per 100,000 children every year, where children in South Asia and Sub-Saharan Africa are disproportionately affected. Some of the cases develop poor treatment outcome (treatment failure or antibiotic change or staying longer in the hospital or death), while others develop good outcome during interventions. Although clinical decision-making is a key aspect of the interventions, there are limited tools such as risk scores to assist the clinical judgment in low-income settings. This study aimed to validate a prediction model and develop risk scores for poor outcomes of severe community-acquired pneumonia (SCAP).

METHODS

A cohort study was conducted among 539 under-five children hospitalized with SCAP. Data analysis was done using R version 4.0.5 software. A multivariable analysis was done. We developed a simplified risk score to facilitate clinical utility. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and calibration plot. Bootstrapping was used to validate all accuracy measures. A decision curve analysis was used to evaluate the clinical and public health utility of our model.

RESULTS

The incidence of poor outcomes of pneumonia was 151(28%) (95%CI: 24.2%-31.8%). Vaccination status, fever, pallor, unable to breastfeed, impaired consciousness, CBC abnormal, entered ICU, and vomiting remained in the reduced model. The AUC of the original model was 0.927, 95% (CI (0.90, 0.96), whereas the risk score model produced prediction accuracy of an AUC of 0.89 (95%CI: 0.853-0.922. Our decision curve analysis for the model provides a higher net benefit across ranges of threshold probabilities.

CONCLUSIONS

Our model has excellent discrimination and calibration performance. Similarly, the risk score model has excellent discrimination and calibration ability with an insignificant loss of accuracy from the original. The models can have the potential to improve care and treatment outcomes in the clinical settings.

Cardiac function in bronchiolitis: Not only a right ventricle matter

OBJECTIVES

Extrapulmonary manifestations of bronchiolitis have been previously studied, with some identifying right ventricle (RV) diastolic/systolic dysfunction. We hypothesized that severe cases of bronchiolitis would have cardiac dysfunction resulting an increase in N-terminal pro-hormone B-type natriuretic peptide (NT-proBNP) values and worse outcomes. Therefore, the objective was to evaluate the existence of cardiac dysfunction and to determine its association with severe bronchiolitis.

METHODS

This prospective cohort study included children hospitalized for bronchiolitis under 1-year old between January 2019 and March 2020. At admission, an echocardiography was performed and plasma levels of NT-proBNP were measured. To analyze outcomes, the cohort was divided into two groups based on the need for positive pressure respiratory support (PPRS), and both were compared to healthy infants.

STATISTICS

bivariant analysis, significant differences p < 0.05.

RESULTS

One hundred eighty-one patients were included; median age was 2 months. Seventy-three patients required PPRS. Compared to controls, patients requiring PPRS showed worse RV systolic function, with lower tricuspid annular-plane systolic excursion (p = 0.002) and parameters of worse right and left diastolic function (trans-tricuspid E and A wave [p = 0.004 and p = 0.04, respectively] and tricuspid tissue doppler imaging [TDI] e' [p = 0.003], trans-mitral E and mitral TDI a' [p = 0.02 and p = 0.005, respectively]). An NT-ProBNP greater than 3582 pg/dl predicts the need for longer necessity of PPRS in patients younger than 2 months.

CONCLUSIONS

In addition to the expected RV systolic dysfunction, patients with severe bronchiolitis have parameters of global diastolic worse function possibly secondary to intrinsic myocardial involvement. NT-ProBNP values at admission had strong discriminatory power to predict worse outcomes.

Do N-terminal pro-brain natriuretic peptide levels determine the prognosis of community acquired pneumonia?

OBJECTIVE

Pneumonia is a leading cause of mortality worldwide, especially in the elderly. The use of clinical risk scores to determine prognosis is complex and therefore leads to errors in clinical practice. Pneumonia can cause increases in the levels of cardiac biomarkers such as N-terminal pro-brain natriuretic peptide (NT-proBNP). The prognostic role of the NT-proBNP level in community acquired pneumonia (CAP) remains unclear. The aim of this study was to evaluate the prognostic role of the NT-proBNP level in patients with CAP, as well as its correlation with clinical risk scores.

METHODS

Consecutive inpatients with CAP were enrolled in the study. At hospital admission, venous blood samples were collected for the evaluation of NT-proBNP levels. The Pneumonia Severity Index (PSI) and the Confusion, Urea, Respiratory rate, Blood pressure, and age ≥ 65 years (CURB-65) score were calculated. The primary outcome of interest was all-cause mortality within the first 30 days after hospital admission, and a secondary outcome was ICU admission.

RESULTS

The NT-proBNP level was one of the best predictors of 30-day mortality, with an area under the curve (AUC) of 0.735 (95% CI: 0.642-0.828; p < 0.001), as was the PSI, which had an AUC of 0.739 (95% CI: 0.634-0.843; p < 0.001), whereas the CURB-65 had an AUC of only 0.659 (95% CI: 0.556-0.763; p = 0.006). The NT-proBNP cut-off level found to be the best predictor of ICU admission and 30-day mortality was 1,434.5 pg/mL.

CONCLUSIONS

The NT-proBNP level appears to be a good predictor of ICU admission and 30-day mortality among inpatients with CAP, with a predictive value for mortality comparable to that of the PSI and better than that of the CURB-65 score.

Expanded A-DROP Score: A New Scoring System for the Prediction of Mortality in Hospitalized Patients with Community-acquired Pneumonia

There are several established prognostic scoring systems for community-acquired pneumonia (CAP). The Pneumonia Severity Index (PSI) is a prediction rule consisting of 20 variables to identify low-risk patients with CAP. Although PSI had high discrimination ability, it is complex to calculate and difficult to use in busy hospital settings. The CURB-65 score is much simpler to use than is PSI, but it has lower sensitivity for predicting mortality compared with PSI. The A-DROP score is a modified version of the CURB-65 score and provides similar predictive power to that of CURB-65. This study was performed to determine whether a simpler score (CURB-65, A-DROP), expanded with a small number of additional variables, can predict mortality more accurately than PSI. We conducted a retrospective observational study of 1,031 patients with CAP who were hospitalized at a tertiary teaching hospital. We used age, sex, comorbidities, vital signs, and laboratory findings as prognostic variables. We compared the PSI, CURB-65, and A-DROP scores using receiver operating characteristic curve analysis. The areas under the curves (AUCs) of PSI, CURB-65, and A-DROP were 0.735, 0.701, and 0.730, respectively.Multivariable analysis identified malignancy [odds ratio (OR): 2.17, 95% confidence interval (CI): 1.13–4.17], respiration rate ≥ 24/min [OR: 2.18, 95% CI: 1.24–3.82], heart rate ≥ 100/min [OR: 2.92, 95% CI: 1.68–5.08], albumin ≤ 3.09 g/dL [OR: 3.85, 95% CI: 2.09–7.07], lactate > 1.7 mmol/L [OR: 2.59, 95% CI: 1.53–4.38], and N-terminal prohormone brain natriuretic peptide > 500 pg/mL [OR: 2.23, 95% CI: 1.26–3.95] as prognostic factors. Using the prognostic variables identified in the multivariable analysis, we assembled a new scoring system, the expanded A-DROP score. The AUC of this score for the prediction of 28-day mortality was 0.834 (95% CI: 0.794–0.874). Bootstrap validation yielded an estimated AUC of 0.833, indicating negligible overfitting of the model.The expanded A-DROP score is a relatively simple and effective scoring system, and its predictive value was superior to those of other scoring systems.

Prognostic value of transthoracic echocardiography and biomarkers of cardiac dysfunction in community-acquired pneumonia

The aim of this study was to determine the prognostic role of echocardiography and compare with admission N-terminal proB-type natriuretic peptide (NT-proBNP) levels in adult patients with community-acquired pneumonia (CAP). Consecutive adult patients hospitalized with CAP were prospectively enrolled and followed-up until hospital discharge or death. Echocardiography was performed within the first 48 hours. Complicated hospitalization (CH) was defined as intensive care unit admission, need for mechanical ventilation or in-hospital mortality. This study was registered with ClinicalTrials.gov, number NCT02441855. A total of 15 CH (13.5%) occurred among 111 patients with CAP. CAP patients with a CH compared with those without CH had significantly higher NT-proBNP values (1267.4±1146.1 vs. 305.6±545.7 pg/mL, p <0.001) and troponin I (23.8±24.3 vs. 10.3±6.3 ng/mL, p 0.02) but had lower left ventricle ejection fraction (52.7±8.7 vs. 60.5±6.7%, p <0.001) and tricuspid annular plane systolic excursion (TAPSE), which is a measure of right ventricular systolic function (17.1±4.4 vs. 21.8±4 mm; p 0.001). Patients with elevation of NT-proBNP and decreased TAPSE at presentation had a significantly higher probability of CH (60%) than patients with either elevated NT-proBNP or decreased TAPSE (40%). Patients with neither elevated NT-proBNP nor decreased TAPSE had a 0% probability of CH. This is the first study to demonstrate that decreased right ventricular systolic function is associated with increased rates of adverse events in patients with CAP.

Validation of a B-type natriuretic peptide as a prognostic marker in pneumonia patients: a prospective cohort study

OBJECTIVES

To validate a B-type natriuretic peptide (BNP) as a prognostic marker in pneumonia patients.

DESIGN

A prospective cohort study.

SETTING

Kanazawa Medical University Himi Municipal (a 250-bed community hospital in Himi-shi, Toyama-ken, Japan).

PARTICIPANTS

All patients diagnosed with pneumonia by the physician and admitted to our hospital between 1 January 2012 and 31 March 2015 whose BNP levels had been determined in the first 24 h of admission. A total of 673 patients were enrolled. Of these, BNP levels were measured for a total of 369 patients on admission.

INTERVENTION

After enrolment, baseline, demographic, clinical and laboratory characteristics including levels of suspected prognostic markers for pneumonia proposed in previous papers, were collected. All patients were followed up until discharge. During analysis, they were divided into categories as follows: community-acquired pneumonia (CAP), aspiration pneumonia (AP), healthcare-associated pneumonia (HCAP) and pneumonia with acute heart failure (PAHF). A univariate and multivariable Cox-regression analysis were applied to each parameter to identify predictors of death. Three cut-off points, namely 40, 100 and 200 pg/mL, as well as the mean, were applied when comparing BNP levels.

MAIN OUTCOME MEASURES

30-day mortality.

RESULTS

Of the 369 patients finally included, 137 were diagnosed with CAP, 122 with AP, 74 with HCAP, and 36 with PAHF. In the univariate analysis, BNP levels (mean, cut-off points 100 pg/mL and 200 pg/mL, p<0.01, respectively) were associated with death in CAP, and similar situation was found for BNP (cut-off points 200 pg/mL, p<0.05) in AP, but not for HCAP, or PAHF. In multivariable Cox-regression analysis, BNP remained an independent mortality predictor (HR 10.01, 95% CI 1.32 to 75.7, p=0.03) in CAP.

CONCLUSIONS

BNP levels may be a useful single prognostic marker for CAP. Further research for validation is warranted.