ORLANDO – Researchers have been able to sustain a dramatic reduction in hypoglycemia incidents at nine St. Louis–area hospitals, thanks to a computer algorithm that warns medical staff when patients appear to be on the road to dangerously low blood sugar levels.
The 6-year retrospective system-wide study, which was released at the annual scientific sessions of the American Diabetes Association, found that the use of the alert system lowered the annual occurrence of severe hypoglycemia events by 41% at the hospitals.
In at-risk patients – those with blood glucose levels under 90 mg/dL – the system considers several variables, such as their weight, creatinine clearance, insulin therapy, and basal insulin doses. If the algorithm considers that a patient is at high risk of a sub–40-mg/dL glucose level – dangerously low – it sends a single alert to medical staff during the patient’s stay.
The idea is that the real-time alerts will go to nurses or pharmacists who will review patient charts and then contact physicians. The doctors are expected to “make clinically appropriate changes,” Dr. Tobin said.
Earlier, Dr. Tobin and colleagues prospectively analyzed the alert system’s effectiveness at a single hospital for 5 months. The trial, a cohort intervention study, tracked 655 patients with a blood glucose level under 90 mg/dL.
In 2014, the researchers reported the results of that trial: The alert identified 390 of the patients as being at high risk for severe hypoglycemia (blood glucose under 40 mg/dL). The frequency of severe hypoglycemia events was just 3.1% in this population vs. 9.7% in unalerted patients who were also deemed to be at high risk (J Hosp Med. 2014: 621-6).
For the new study, researchers extended the alert system to nine hospitals and tracked its use from 2011 to 2017.
During all visits, the number of severe hypoglycemic events fell from 2.9 to 1.7 per 1,000 at-risk patient days. (P less than .001)
At one hospital, Dr. Tobin said, the average monthly number of severe hypoglycemia incidents fell from 40 to 12.
Researchers found that the average blood glucose level post alert was 93 mg/dL vs. 74 mg/dL before alert. They also reported that the system-wide total of alerts per year ranged from 4,142 to 5,649.
“The current data reflected in our poster show that the alert process is sustainable over a wide range of clinical settings, including community hospitals of various size and complexity, as well as academic medical centers,” Dr. Tobin said.
The alert system had no effect on hyperglycemia, Dr. Tobin said.
In regard to expense, Dr. Tobin said it’s small because the alert system uses existing current staff and computer systems. Setup costs, he said, included programming, creating the alert infrastructure, and staff training
No study funding is reported. Dr. Tobin reports relationships with Novo Nordisk (advisory board, speaker’s bureau) and MannKind (speaker’s bureau). The other authors report no relevant disclosures.
SOURCE: Tobin G et al. ADA 2018. Abstract 397-P.