Machine learning approach for predicting inhalation injury in patients with burns.

BACKGROUND: The coronavirus disease pandemic has had a tangible impact on bronchoscopy for burn inpatients due to isolation and triage measures. We utilised the machine-learning approach to identify risk factors for predicting mild and severe inhalation injury and whether patients with burns experienced inhalation injury. We also examined the ability of two dichotomous models to predict clinical outcomes including mortality, pneumonia, and duration of hospitalisation. METHODS: A retrospective 14-year single-centre dataset of 341 intubated patients with burns with suspected inhalation injury was established. The medical data on day one of admission and bronchoscopy-diagnosed inhalation injury grade were compiled using a gradient boosting-based machine-learning algorithm to create two prediction models: model 1, mild vs. severe inhalation injury; and model 2, no inhalation injury vs. inhalation injury. RESULTS: The area under the curve (AUC) for model 1 was 0·883, indicating excellent discrimination. The AUC for model 2 was 0·862, indicating acceptable discrimination. In model 1, the incidence of pneumonia (P < 0·001) and mortality rate (P < 0·001), but not duration of hospitalisation (P = 0·1052), were significantly higher in patients with severe inhalation injury. In model 2, the incidence of pneumonia (P < 0·001), mortality (P < 0·001), and duration of hospitalisation (P = 0·021) were significantly higher in patients with inhalation injury. CONCLUSIONS: We developed the first machine-learning tool for differentiating between mild and severe inhalation injury, and the absence/presence of inhalation injury in patients with burns, which is helpful when bronchoscopy is not available immediately. The dichotomous classification predicted by both models was associated with the clinical outcomes.

Machine Learning Approach for Predicting Inhalation Injury in Patients with Burns

Background The coronavirus disease pandemic has had a tangible impact on bronchoscopy for burn inpatients due to isolation and triage measures. We utilised the machine-learning approach to identify risk factors for predicting mild and severe inhalation injury (INH) and whether patients with burns experienced INH. We also examined the ability of two dichotomous models to predict clinical outcomes including mortality, pneumonia, and duration of hospitalisation. Methods A retrospective 14-year single-centre dataset of 341 intubated patients with burns with suspected INH was established. The medical data on day one of admission and bronchoscopy-diagnosed INH grade were compiled using a gradient boosting-based machine-learning algorithm to create two prediction models: model 1, mild vs. severe INH;and model 2, no INH vs. INH. Results The area under the curve (AUC) for model 1 was 0·883, indicating excellent discrimination. The AUC for model 2 was 0·862, indicating acceptable discrimination. In model 1, the incidence of pneumonia (P<0·001) and mortality rate (P<0·001), but not duration of hospitalisation (P=0·1052), were significantly higher in patients with severe INH. In model 2, the incidence of pneumonia (P<0·001), mortality (P<0·001), and duration of hospitalisation (P=0·021) were significantly higher in patients with INH. Conclusions We developed the first machine-learning tool for differentiating between mild and severe INH, and the absence/presence of INH in patients with burns, which is helpful when bronchoscopy is not available immediately. The dichotomous classification predicted by both models was associated with the clinical outcomes.

Development of a semi-structured, multifaceted, computer-aided questionnaire for outbreak investigation: e-Outbreak Platform.

Aggressive tracing of contacts of confirmed cases is crucial to Taiwan's successful control of the early spread of COVID-19. As the pandemic lingers, an epidemiological investigation that can be conducted efficiently in a timely manner can help decrease the burden on the health personnel and increase the usefulness of such information in decision making. To develop a new tool that can improve the current practice of epidemiological investigation by incorporating new technologies in digital platform and knowledge graphs. To meet the various needs of the epidemiological investigation, we decided to develop an e-Outbreak Platform that provides a semi-structured, multifaceted, computer-aided questionnaire for outbreak investigation. There are three major parts of the platform: (1) a graphic portal that allows users to have an at-glance grasp of the functions provided by the platform and then choose the one they need; (2) disease-specific questionnaires that can accommodate different formats of the information, including text typing, button selection, and pull-down menu; and (3) functions to utilize the stored information, including report generation, statistical analyses, and knowledge graphs displaying contact tracing. When the number of outbreak investigation increases, the knowledge graphs can be extended to encompass other persons appearing in the same location at the same time, i.e., constituting a potential contact cluster. The information extracted can also be used to display the tracing on a map in animation. Overall, this system can provide a basis for further refinement that can be generalized to a variety of outbreak investigations.