AI Case Study
Cleveland Clinic and Microsoft identify at-risk patients in ICU to prevent the occurrence of cardiac failure with machine learning ensembles
Cleveland Clinic worked with Microsoft to develop predictive models to identify at-risk patients of cardiac failure in ICU. Analysing vital data collected from ICU along with lab test and patient information, they developed Boosted Decision Tree models. They believe the predictive model will allow for timely intervention by medics.
Healthcare Providers And Services
"Using the data collected from monitoring units in ICUs over a period of time, we want to predict if a patient will need to be administered vasopressors in the near future, to prevent the occurrence of cardiac failure. The data included the vitals collected in ICU, the lab tests and patient information from a set of patients over a period of time.
To make predictions, a binary classification model was developed, to help determine if a patient will need vasopressor therapy in the next eight-hour window. The time window was decided based on the time requirement for intervention to be effective."
R And D
Core Research And Development
"Cleveland Clinic is a non-profit academic medical centre providing clinical and hospital care. It’s a leader in research, education and health information and at the forefront of developing and applying new technology. Cleveland Clinic launched eHospital in April 2014 and monitors a total of 100 beds in six ICUs. Cleveland Clinic recently teamed up with Microsoft to utilize predictive and advanced analytics to identify potential at-risk patients under ICU care."
"Sudden cardiac arrest is a major public health problem with more than 450,000 individuals affected annually. Vasopressors are a part of pulseless sudden cardiac arrest management protocol. These are drugs that cause the blood vessels to constrict, which, in turn, causes blood pressure to rise in an attempt to restore spontaneous circulation."
Unquantified improvements but claim improvements.
"Given the industry-wide shortage of primary care physicians and nurses, ML models like these can provide a helping hand and an extra set of eyes, especially during critical situations. Using Microsoft’s big data and advanced analytics capabilities, healthcare providers can use their data to assist physicians and nurses with critical medical decisions, and improve patient care and experience."
"To make predictions, a binary classification model was developed, to help determine if a patient will need vasopressor therapy in the next eight-hour window. The time window was decided based on the time requirement for intervention to be effective.
Boosted decision trees gave the most accurate results. The trained model was then used for making predictions and deployed as a web service."
"The data included the vitals collected in ICU, the lab tests and patient information from a set of patients over a period of time."