New Research Shows EarlySign's AI Solution Can Identify Prediabetic Patients At High Risk Of Developing Diabetes

Machine Learning Model Utilizing Patient Data from Electronic Health Records Demonstrates Superiority Over Traditional Logic-Regression Model

Tel Aviv, Israel /PRNewswire/ - Medial EarlySign (earlysign.com), a leader in machine learning-based solutions to aid in the early detection and prevention of high-burden diseases, today announced the results of a new study assessing the effectiveness of its Pre2D-Flag AlgoMarker™ algorithmic model to identify individuals at high risk of progression from prediabetes to diabetes.

The peer-reviewed study, Prediction of Progression from Pre-Diabetes to Diabetes: Development and Validation of a Machine Learning Model, published in Diabetes/Metabolism Research and Reviews, evaluated the validity of EarlySign's machine-learning based model to identify individuals with prediabetes at increased risk of annual progression to diabetes. The retrospective data study was performed in conjunction with the Diabetes Unit, Department of Endocrinology and Metabolism at Hadassah Hebrew University of Jerusalem and the Medical Division of Maccabi Healthcare Services (MHS).

EarlySign's Pre2D-Flag AlgoMarker was found to be consistently superior to the alternative logistic-regression model for diabetes prediction, showing that machine learning can capture the subtle, multivariate relationships which linear models may be unable to detect.

"The growing prevalence of diabetes has become a universal health problem with approximately 12% ($727 billion) of global health expenditure spent on addressing diabetes and downstream complications," said Avivit Cahn, MD, Senior Endocrinologist at Hadassah University Hospital and lead author of the study. "Traditional risk stratification methods, relying on laboratory tests and the manual collection of clinical characteristics, tend to be limited. A machine learning model, trained to predict prediabetes to diabetes progression using electronic health records, can significantly lower the overall burden of diabetes through early identification and timely interventions delivered specifically to high risk populations, rather than generic programs targeting all prediabetic individuals."

EarlySign's Pred2D model was trained on data from The Health Improvement Network (THIN) database, the U.K. National Health Service's primary care database, and externally validated on the Canadian AppleTree and Israeli MHS data sets. The study was performed on a cohort of 852,454 individuals with prediabetes from the THIN database. The machine learning model was implemented using 69 variables, generated from 11 basic signals. It utilized information up to 10 years prior to the index date to predict the risk of developing diabetes in the following year. The algorithm showed reliable predictions when including less historical data i.e. 3-5 years and was also able to reasonably predict diabetes progression up to five years.

"This study demonstrates the potential efficacy of incorporating machine learning models into large clinical systems to identify high-risk individuals with prediabetes who stand to benefit from early intervention strategies," said Dr. Jeremy Orr, CEO of Medial EarlySign. "The continuity and versatility of our Pre2D-Flag AlgoMarker enables it to be employed in multiple clinical setups to facilitate interventions by clinicians within an actionable timeframe and help delay or postpone the onset of diabetes."

About Medial EarlySign
Medial EarlySign helps healthcare systems with early detection and prevention of high-burden diseases. Their suite of outcome-focused software solutions (AlgoMarkers™) find subtle, early signs of high-risk patient trajectories in existing lab results and ordinary EHR data already collected in the course of routine care. EarlySign's AlgoMarkers can help clients identify patients at high risk for conditions such as lower GI disorders, prediabetic progression to diabetes, and downstream diabetic complications such as chronic kidney disease (CKD). The algorithmic models developed using the company's machine learning approach are supported by peer-reviewed research published by internationally recognized health organizations and hospitals. Founded in 2013, Medial EarlySign is headquartered in Tel Aviv, Israel with US headquarters in Colorado. For more information, please visit: https://earlysign.com.

Follow Medial EarlySign on LinkedIn: Medial EarlySign and Twitter: @MedialEarlySign