Survival of mechanically ventilated patients admitted to intensive care units. Results from a tertiary care center between 2016-2018

Compact machine learning model for the accurate prediction of first 24-hour survival of mechanically ventilated patients

Background

The field of machine learning has been evolving and applied in medical applications. We utilised a public dataset, MIMIC-III, to develop compact models that can accurately predict the outcome of mechanically ventilated patients in the first 24 h of first-time hospital admission.

Methods

67 predictive features, grouped into 6 categories, were selected for the classification and prediction task. 4 tree-based algorithms (Decision Tree, Bagging, eXtreme Gradient Boosting and Random Forest), and 5 non-tree-based algorithms (Logistic Regression, K-Nearest Neighbour, Linear Discriminant Analysis, Support Vector Machine and Naïve Bayes), were employed to predict the outcome of 18,883 mechanically ventilated patients. 5 scenarios were crafted to mirror the target population as per existing literature. S1.1 reflected an imbalanced situation, with significantly fewer mortality cases than survival ones, and both the training and test sets played similar target class distributions. S1.2 and S2.2 featured balanced classes; however, instances from the majority class were removed from the test set and/or the training set. S1.3 and S 2.3 generated additional instances of the minority class via the Synthetic Minority Over-sampling Technique. Standard evaluation metrics were used to determine the best-performing models for each scenario. With the best performers, Autofeat, an automated feature engineering library, was used to eliminate less important features per scenario.

Results

Tree-based models generally outperformed the non-tree-based ones. Moreover, XGB consistently yielded the highest AUC score (between 0.91 and 0.97), while exhibiting relatively high Sensitivity (between 0.58 and 0.88) on 4 scenarios (1.2, 2.2, 1.3, and 2.3). After reducing a significant number of predictors, the selected calibrated ML models were still able to achieve similar AUC and MCC scores across those scenarios. The calibration curves of the XGB and BG models, both prior to and post dimension reduction in Scenario 2.2, showed better alignment to the perfect calibration line than curves produced from other algorithms.

Conclusion

This study demonstrated that dimension-reduced models can perform well and are able to retain the important features for the classification tasks. Deploying a compact machine learning model into production helps reduce costs in terms of computational resources and monitoring changes in input data over time.

Time to death and its predictors among adult patients on mechanical ventilation admitted to intensive care units in West Amhara comprehensive specialized hospitals, Ethiopia: a retrospective follow-up study

INTRODUCTION

Mechanical ventilation is the most common intervention for patients with respiratory failure in the intensive care unit. There is limited data from African countries, including Ethiopia on time to death and its predictors among patients on mechanical ventilators. Therefore, this study aimed to assess time to death and its predictors among adult patients on mechanical ventilation admitted in comprehensive specialized hospitals in West Amhara, Ethiopia.

METHODS

An institutional-based retrospective follow-up study was conducted from January 1, 2020, to December 31, 2022. A simple random sampling was used to select a total of 391 patients' charts. Data were collected using data the extraction tool, entered into Epi-data version 4.6.0, and exported to STATA version 14 for analysis. Kaplan-Meier failure curve and the log-rank test were fitted to explore the survival difference among groups. The Cox regression model was fitted, and variables with a p-value < 0.25 in the bivariable Cox regression were candidates for the multivariable analysis. In the multivariable Cox proportional hazard regression, an adjusted hazard ratio with 95% confidence intervals were reported to declare the strength of association between mortality and predictors when a p value is < 0.05.

RESULTS

A total of 391 mechanically ventilated patients were followed for 4098 days at risk. The overall mortality of patients on mechanical ventilation admitted to the intensive care units was 62.2%, with a median time to death of 16 days (95% CI: 11, 22). Those patients who underwent tracheostomy procedure (AHR = 0.40, 95% CI: 0.20, 0.80), received cardio-pulmonary resuscitation (AHR = 8.78, 95% CI: 5.38, 14.35), being hypotensive (AHR = 2.96, 95% CI: 1.11, 7.87), and had a respiratory rate less than 12 (AHR = 2.74, 95% CI: 1.48, 5.07) were statistically significant predictors of time to death among mechanically ventilated patients.

CONCLUSION

The mortality rate of patients on mechanical ventilation was found to be high and the time to death was short. Being cardiopulmonary resuscitated, hypotensive, and had lower respiratory rate were significant predictors of time to death, whereas patients who underwent tracheostomy was negatively associated with time to death. Tracheostomy is needed for patients who received longer mechanical ventilation, and healthcare providers should give a special attention for patients who are cardiopulmonary resuscitated, hypotensive, and have lower respiratory rate.

Time to recovery and its predictors among critically ill patients on mechanical ventilation from intensive care unit in Ethiopia: a retrospective follow up study

Introduction

For critically ill patients, mechanical ventilation is considered a pillar of respiratory life support. The mortality of victims in intensive care units is high in resource-constrained Sub-Saharan African countries. The recovery and prognosis of mechanically ventilated victims are unknown, according to evidence. The goal of the study was to see how long critically ill patients on mechanical ventilation survived.

Methods

A retrospective follow-up study was conducted. A total of 376 study medical charts were reviewed. Data was collected through reviewing medical charts. Data was entered into Epi-data manager version 4.6.0.4 and analyzed through Stata version 16. Descriptive analysis was performed. Kaplan- Meier survival estimates and log rank tests were performed. Cox proportional hazard model was undertaken.

Results

Median recovery time was 15 days (IQR: 6–30) with a total recovery rate of 4.49 per 100 person-days. In cox proportional hazard regression, diagnosis category {AHR: 1.690, 95% CI: (1.150- 2.485)}, oxygen saturation {AHR: 1.600, 95% CI: (1.157- 2.211)}, presence of comorbidities {AHR: 1.774, 95% CI: (1.250–2.519)}, Glasgow coma scale {AHR: 2.451, 95% CI: (1.483- 4.051)}, and use of tracheostomy {AHR: 0.276, 95% CI: (0.180–0.422)} were statistically significant predictors.

Discussion

Based on the outcomes of this study, discussions with suggested possible reasons and its implications were provided.

Conclusion and Recommendations

Duration and recovery rate of patients on mechanical ventilation is less than expected of world health organization standard. Diagnosis category, oxygen saturation, comorbidities, Glasgow coma scale and use of tracheostomy were statistically significant predictors. Mechanical ventilation durations should be adjusted for chronic comorbidities, trauma, and use of tracheostomy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12873-022-00689-3.

Characteristics of mechanically ventilated COVID-19 patients in the Al-Ahsa Region of Saudi Arabia: a retrospective study with survival analysis

BACKGROUND

About 5-10% of coronavirus disease 2019 (COVID-19) infected patients require critical care hospitalization and a variety of respiratory support, including invasive mechanical ventilation. Several nationwide studies from Saudi Arabia have identified common comorbidities but none were focused on mechanically ventilated patients in the Al-Ahsa region of Saudi Arabia.

OBJECTIVES

Identify characteristics and risk factors for mortality in mechanically ventilated COVID-19 patients.

DESIGN

Retrospective chart review SETTING: Two general hospitals in the Al-Ahsa region of Saudi Arabia PATIENTS AND METHODS: We included mechanically ventilated COVID-19 patients (>18 years old) admitted between 1 May and 30 November 2020, in two major general hospitals in the Al-Ahsa region, Saudi Arabia. Descriptive statistics were used to characterize patients. A multivariable Cox proportional hazards (CPH) model was used exploratively to identify hazard ratios (HR) of predictors of mortality.

MAIN OUTCOME MEASURES

Patient characteristics, mortality rate, extubation rate, the need for re-intubation and clinical complications during hospitalization.

SAMPLE SIZE AND CHARACTERISTICS

154 mechanically ventilated COVID-19 patients with median (interquartile range) age of 60 (22) years; 65.6% male.

RESULTS

Common comorbidities were diabetes (72.2%), hypertension (67%), cardiovascular disease (14.9%) and chronic kidney disease (CKD) (14.3%). In the multivariable CPH model, age >60 years old (HR=1.83, 95% CI 1.2-2.7, P=.002), CKD (1.61, 95% CI 0.9-2.6, P=.062), insulin use (HR=0.65, 95% CI 0.35-.08, P<.001), and use of loop diuretics (HR=0.51, 95% CI 0.4, P=.037) were major predictors of mortality.

CONCLUSION

Common diseases in mechanically ventilated COVID-19 patients from the Al-Ahsa region were diabetes, hypertension, other cardiovascular diseases, and CKD in this exploratory analysis.

LIMITATIONS

Retrospective, weak CPH model performance.

CONFLICTS OF INTEREST

None.

Rotating nasal masks with nasal prongs reduces the incidence of moderate to severe nasal injury in preterm infants supported by noninvasive ventilation

OBJECTIVE

To investigate the association between noninvasive ventilation delivery devices and the incidence of nasal septum injury in preterm infants.

METHODS

This retrospective singlecenter cohort study included preterm infants supported by noninvasive ventilation. The incidence of nasal injury was compared among three groups according to the noninvasive ventilation delivery device (G1 - nasal mask; G2 - binasal prongs; and G3, rotation of nasal mask with prongs). Nasal injury was classified according to the National Pressure Ulcer Advisory Panel as stages 1 - 4. Multivariate regression analyses were performed to estimate relative risks to identify possible predictors associated with medical device-related injuries.

RESULTS

Among the 300 infants included in the study, the incidence of medical device-related injuries in the rotating group was significantly lower than that in the continuous mask or prong groups (n = 68; 40.48%; p value < 0.01).The basal prong group presented more stage 2 injuries (n = 15; 55.56%; p < 0.01). Staying ≥ 7 days in noninvasive ventilation was associated with a higher frequency of medical device-related injuries, regardless of device (63.81%; p < 0.01). Daily increments in noninvasive ventilation increased the risk for nasal injury by 4% (95%CI 1.02 - 1.06; p < 0.01). Higher birth weight indicated protection against medical device-related injuries. Each gained gram represented a decrease of 1% in the risk of developing nasal septum injury (RR: 0.99; 95%CI 0.99 - 0.99; p < 0.04).

CONCLUSION

Rotating nasal masks with nasal prongs reduces the incidence of moderate to severe nasal injury in comparison with single devices. The addition of days using noninvasive ventilation seems to contribute to medical device-related injuries, and higher birth weight is a protective factor.

Machine Learning Prediction Models for Mechanically Ventilated Patients: Analyses of the MIMIC-III Database

Background: Mechanically ventilated patients in the intensive care unit (ICU) have high mortality rates. There are multiple prediction scores, such as the Simplified Acute Physiology Score II (SAPS II), Oxford Acute Severity of Illness Score (OASIS), and Sequential Organ Failure Assessment (SOFA), widely used in the general ICU population. We aimed to establish prediction scores on mechanically ventilated patients with the combination of these disease severity scores and other features available on the first day of admission.

Methods: A retrospective administrative database study from the Medical Information Mart for Intensive Care (MIMIC-III) database was conducted. The exposures of interest consisted of the demographics, pre-ICU comorbidity, ICU diagnosis, disease severity scores, vital signs, and laboratory test results on the first day of ICU admission. Hospital mortality was used as the outcome. We used the machine learning methods of k-nearest neighbors (KNN), logistic regression, bagging, decision tree, random forest, Extreme Gradient Boosting (XGBoost), and neural network for model establishment. A sample of 70% of the cohort was used for the training set; the remaining 30% was applied for testing. Areas under the receiver operating characteristic curves (AUCs) and calibration plots would be constructed for the evaluation and comparison of the models' performance. The significance of the risk factors was identified through models and the top factors were reported.

Results: A total of 28,530 subjects were enrolled through the screening of the MIMIC-III database. After data preprocessing, 25,659 adult patients with 66 predictors were included in the model analyses. With the training set, the models of KNN, logistic regression, decision tree, random forest, neural network, bagging, and XGBoost were established and the testing set obtained AUCs of 0.806, 0.818, 0.743, 0.819, 0.780, 0.803, and 0.821, respectively. The calibration curves of all the models, except for the neural network, performed well. The XGBoost model performed best among the seven models. The top five predictors were age, respiratory dysfunction, SAPS II score, maximum hemoglobin, and minimum lactate.

Conclusion: The current study indicates that models with the risk of factors on the first day could be successfully established for predicting mortality in ventilated patients. The XGBoost model performs best among the seven machine learning models.

Novel predictor markers for early differentiation between transient tachypnea of newborn and respiratory distress syndrome in neonates

Neonatal Respiratory Distress Syndrome (RDS) and Transient Tachypnea of newborn (TTN) are common similar neonatal respiratory diseases. Study the early predictor markers in differentiation between TTN and RDS in neonates. A prospective case control study which was done in Neonatal Intensive Care Unit (NICU) of Tanta University Hospital (TUH) from September 2016 to March 2018. Three groups of neonates were included in the study: RDS group (45 neonates), TTN group (45 neonates), and control group (45 healthy neonates). There were statistically significant difference (SSD) between our studied three groups as regard serum Malondialdehyde (MDA), Superoxide dismutase SOD, Lactate dehydrogenase (LDH), and blood PH and P-values were 0.001* for these comparative parameters. The ROC curve of RDS cases revealed that the serum MDA Cut off, sensitivity and specificity were 1.87 mmol/L, 98%, 96%, respectively which had the highest sensitivity and specificity followed by the serum SOD then the serum LDH and lastly the blood PH while in TTN cases, the serum MDA Cut off, sensitivity and specificity were 0.74 mmol/L, 96%, 93%, respectively then the serum SOD then the serum LDH and lastly the blood PH. Serum MDA, SOD, LDH, and PH had a beneficial role as early predictors in differentiation between TTN and RDS in neonates.