SUMMARY: UCLA researchers in the Department of Psychiatry and Biobehavioral Sciences have invented a novel algorithm that uses electronic health records to determine a patient’s risk of having undiagnosed two diabetes mellitus.
BACKGROUND: Type two diabetes mellitus, often simply referred to as diabetes, affects 1 in 10 patients in the U.S. or about 34 million people. Patients are diagnosed with diabetes after completing a laboratory test that measures the body’s ability to regulate blood sugar. Due to insufficient testing, it is estimated that 25% of diabetes cases are undiagnosed. This predicament poses a severe risk to patients as type two diabetes mellitus is dangerous, and if left untreated, can be fatal. Often diabetes is only diagnosed after the patient has already experienced pain and suffering. Many patients can live healthy lives if treatment for diabetes is administered early. Early detection of diabetes can reduce the physical burden patients live with as well as the financial burden that untreated diabetes has on our society. Population-wide screening however is not cost-effective. Therefore, a scalable method for diagnosing diabetes is needed to ensure people with diabetes get the treatment they need.
While blood sugar tests are still needed to confirm diabetes diagnosis, other methods exist for identifying individuals at risk for developing diabetes. These methods, which rely on patterns in electronic health records, are theoretically ideal, as the cost of running an algorithm is far less than performing blood tests on every individual. Unfortunately, many of these algorithms rely on specific laboratory tests or have low accuracy. There is a clear and pressing need for better algorithms that use readily available electronic health record data to accurately predict individuals at risk for diabetes.
INNOVATION: Researchers at UCLA led by Dr. Ariana Anderson have developed a novel algorithm that uses electronic health records to accurately determine if a patient is likely to develop type two diabetes mellitus. This algorithm is robust to the issues of missing data that limit the accuracy of competing algorithms and outperforms conventional diabetes screening guidelines by including novel risk factors. Researchers also trained their model to utilize diverse formats of health records, thereby increasing the capacity of the algorithm. The proposed innovation offers an invaluable method for clinicians to accurately predict patients’ susceptibility to developing diabetes, reducing economic strain and medical harm caused by late detection of disease onset.
POTENTIAL APPLICATIONS:
ADVANTAGES:
DEVELOPMENT-TO-DATE: An algorithm has been created and tested on a variety of electronic health records from U.S. patients.
Related Papers:
A. E. Anderson, W. T. Kerr, A. Thames, T. Li, J. Xiao, and M. S. Cohen. Electronic Health Record Phenotyping Improves Detection and Screening of Type 2 Diabetes in the General United States Population: A Cross-Sectional, Unselected, Retrospective Study. Journal of Biomedical Informatics. 2016.