Utilisation of a datathon to explore dysphagia post invasive mechanical ventilation.

Introduction: A datathon is an event designed to bring researchers, clinicians, and data scientist together. Machine learning enables exploratory data analysis and hypothesis generation for clinically relevant questions in a collaborative environment. At present, it is unclear which patients are at greatest risk of dysphagia post invasive mechanical ventilation in an Australian population. A recent international meta-analysis demonstrated that 49% of all intubated patients present with dysphagia. Objectives/Aims: To utilise a datathon as a platform for trend analysis with access to large health databases to determine the incidence, and possible risk factors for dysphagia post invasive mechanical ventilation across the state of Victoria. Method(s): This pilot study occurred at a regional critical care datathon. The data have been extracted from the Victorian Admitted Episodes Database, using minimal data entry fields and ICD 10-AM codes. All results presented are preliminary findings based on trend analysis. Result(s): In 2017 across Victoria, 15,033 patients underwent invasive mechanical ventilation. Of these 1,221 (8%) were identified as dysphagic. Incidence of dysphagia was noted to increase with length of ventilation, however plateaued at 24% after 4 days. 18% of patients with dysphagia were ventilated for less than 48 hours. Patients with delirium were more likely to present with dysphagia (23.5% vs. 5.5%). Based on data visualisation age did not appear to increase the risk of dysphagia until 70+ years. Conclusion(s): Prolonged ventilation is a frequently reported risk for dysphagia. Despite this, 1 in 5 patients with dysphagia in this study were for ventilated for 48 hours or less. It is unclear whether the lower incidence (8%) compared to previous reports reflects underdiagnosis and recognition of dysphagia post mechanical ventilation, or differences in international practice.