Presentations Highlight Dascena’s Machine Learning Algorithms’ Ability to Provide Earlier Detection of Ischemic Stroke, Pulmonary Embolism and Deep Vein Thrombosis

OAKLAND, Calif.--Dascena, Inc., a machine learning company, today announced that the company will present data from three different studies of its machine learning algorithms at the Society for Critical Care Medicine’s (SCCM) Annual Congress, taking place virtually from January 31 to February 12, 2021. The presentations highlight data from studies of three Dascena-developed algorithms for the prediction of ischemic stroke, pulmonary embolism (PE) or deep vein thrombosis (DVT). The research on PE was honored with a STAR Research Award, which is presented by SCCM to recognize excellence in critical care research and high-quality original investigative research.

“These three disease areas all represent areas of high unmet need, because if left untreated or not treated soon enough, they can result in devastating impacts on patients,” said Ritankar Das, chief executive officer of Dascena. “While prediction of these diseases can enable early intervention and possibly disease prevention, no systems previously existed that have been reliably shown to predict any of these outcomes. We were motivated to address these unmet needs, and to demonstrate the diverse potential of machine learning to address important clinical problems.”

All three algorithms demonstrate strong predictive performance, with results supporting that all algorithms are capable of identifying patients before the outcome of interest occurs. Details of the presentations are as follows:

Poster Name: Enriching the Study Population for Ischemic Stroke Therapeutic Trials Using Machine Learning
Details: A machine learning algorithm was developed to predict ischemic stroke within the next 12 months and trained and tested on data from 1,207 patient encounters. Findings demonstrated that it is possible to use machine learning to identify patients at high risk of stroke within a 12-month period, and that the algorithm may have applications to clinical trial enrollment by enabling identification of patients at high risk of ischemic stroke.

Poster Name: A Machine Learning Algorithm for Predicting Pulmonary Embolism Among Hospitalized Patients
Details: Dascena developed a machine learning algorithm to predict PE before clinical onset in hospitalized patients with improvements over the standard of care Geneva score, which was trained and tested on data from 63,841 patient encounters. Findings showed that the algorithm led to earlier and more accurate predictions of PE compared to standard systems, and that the algorithm can be easily integrated into existing inpatient workflows.

Poster Name: A Machine Learning Approach to Predict Deep Venous Thrombosis Among Hospitalized Patients
Details: Dascena developed a machine learning algorithm designed to predict DVT with greater accuracy and lead time than current systems allow, which was evaluated using electronic health record data from nearly 100,000 patients. Results showed that it is possible to accurately predict DVT up to 24 hours in advance of clinical diagnosis, and machine learning methods may offer an advantage over traditional risk scores, such as the IMPROVE score.

“The potential of these algorithms is immense,” said Anna Siefkas, SM, scientific writer, Dascena. “In addition to improving patient outcomes by enabling early interventions, these algorithms may improve our ability to study these conditions by identifying patients suitable for enrollment in clinical trials. This was one of the motivations behind the stroke prediction algorithm, in particular. We’re excited to share these findings with the scientific community and hope that our findings will lead to better care and treatment of patients in the future.”

About Dascena

Dascena develops machine learning predictive algorithms for the next generation of precision medicine. Dascena’s infrastructure platform brings patient data to life to enable early disease prevention and improved care outcomes for patients. For more information, visit Follow us on LinkedIn.

Read on Business Wire