Researchers have developed an algorithm that can evaluate electronic health records to determine whether the patient has diabetes, FierceHealthIT reports.
Details of Algorithm
The algorithm, published in BioMedCentral, focuses on determining the earliest possible data of diagnosis in close to real time.
It evaluates information that is regularly documented and can be extracted from structured data fields, including:
- Past medical history;
- Problem list;
- Medications; and
- Laboratory results.
Each element is given a point value. After reaching a certain threshold, the algorithm identifies the presence of diabetes and calculates the earliest date that the disease could have been diagnosed.
Algorithm’s Accuracy
The researchers compared the algorithm’s findings with the opinion of a physician and found that the algorithm agreed on the date of diagnosis in 78.4% of cases.
It established a date of diagnosis that was within three months of the physician’s date in 94% of cases.
Experts say the algorithm could be effective in reaching patients who do not visit a physician regularly (Hall, FierceHealthIT, 8/2). Source