Predictive Modeling and Machine Learning

We strive to significantly improve lives through cutting-edge applications of machine learning. Predictive models provide insights from the complex patterns and correlations found in our rich and massive data, and these insights are translated into actions.

Clalit’s large heterogeneous population allows us to retrospectively identify early signs for various diseases and indicators of health outcomes. These early markers are combined in predictive models to produce individualized risk scores, acting as a highly sensitive and specific tool for flagging individuals at highest risk.

Our models include both supervised and unsupervised techinques, on tabular data, time series and natural language processing. We have unique know how in developing these models under real life limitations, integrating medical domain knowledge into the full machine learning pipeline.

Clalit’s frontline clinical teams employ these models in a proactive manner, utilizing the targeted risk assessments for preventive measures. Healthcare managers also use our models to identify clinical subgroups of patients at-risk. Finally models are also used in resource and program planning, such as for vaccine outreach targeting or innovative patient selection.

We are continuously expanding our tools set, with active research in various areas, all dedicated to ensuring that only the highest quality models reach the field.