In an attempt to identify and treat patients with diabetes earlier, researchers have developed an algorithm that can spot the disease in EHRs almost as effectively as a physician in close to real time.
While other algorithms have been developed to identify diabetes from EHR data, this effort, published by BioMedCentral, focused on the earliest possible date of diagnosis.
Such efforts to identify and reach out to those affected are at the heart of population health management, Bonnie S. Cassidy, senior director of HIM innovation at Nuance Communications, wrote in an article about the emerging role of population health information management professionals. It could be effective in reaching patients who might not visit their doctors regularly.
The algorithm focused on data elements routinely documented and extractable from structured data fields, such as past medical history, problem list, medications, and laboratory results within the EHR. Each was given a points value. Once a threshold was reached, it identified the presence of diabetes and determined the earliest date that a diagnosis could have been made.
The researchers checked the algorithm's accuracy against the opinion of a doctor and found perfect agreement on the date of diagnosis in 78.4 percent of cases. It also set the date of diagnosis within three months of a physician's chart review date in 94 percent of cases.
New York University, NYU Langone Medical Center and Independence Blue Cross (IBC) recently announced a collaboration to apply artificial intelligence to prediction of diabetes and other diseases. Researchers will create algorithms that can be applied to medical and pharmacy claims data to identify patients with either undiagnosed or pre-diabetes.
Tina Buop, CIO of La Clinica de la Raza, a community health center in Oakland, Calif., earlier this year explained how her organization is using predictive analytics to understand whether diabetic patients are progressing--and to get them back in to see the physician as necessary.
To learn more:
- read the research