The Effectiveness of an Electronic Decision Support Algorithm to Optimize Recommendations of SGLT2i and GLP-1RA in Patients with Type 2 Diabetes upon Discharge from Internal Medicine Wards

Abstract

Background/Objectives: Despite the established cardiovascular benefit of sodium-glucose cotransporter-2 inhibitors (SGLT2is) and glucagon-like peptide-1 receptor agonists (GLP-1RAs), these medications are under-prescribed in patients with type 2 diabetes. Our study aims to examine the effectiveness of a clinical decision support system (CDSS) in improving the recommendation rate of SGLT2i and GLP-1RA upon discharge. Methods: We developed an algorithm to automatically recommend SGLT2is and GLP-1RAs for eligible patients with type 2 diabetes upon discharge, based on current guidelines. Data were collected from electronic medical records of all eligible patients ≥18 years old hospitalized in one of five internal medicine wards at Beilinson Hospital. The primary outcome was to evaluate the rate of physician recommendation of SGLT2is and GLP-1RAs at discharge, before and after algorithm implementation. Results: Our study included 1318 patients in the pre-algorithm group and 970 in the post-algorithm group. The recommendation rate of SGLT2is and GLP-1RAs was 8.5% in the pre-algorithm group and 22.7% in the post-algorithm. The odds ratio (OR) of recommendation in the post- vs. pre-algorithm group was 3.151 (95% CI: 2.467-4.025, p < 0.0001). Recommendation rates increased in all subgroups analyzed, notably in patients hospitalized due to heart failure (recommendation rate pre-algorithm: 14.6% vs. post-algorithm: 49.02%). Conclusions: This study demonstrates the benefit of a CDSS in improving the recommendation rate of SGLT2is and GLP-1RAs in patients with type 2 diabetes upon discharge from hospitalization. Future studies should assess the impact of the algorithm on recommendation rates in other wards, medication utilization, and long-term outcomes.

Keywords: GLP-1RA; SGLT2i; clinical decision support system; electronic decision support algorithm; type 2 diabetes.

Figure 1

The rates of recommendation before and after algorithm implementation and the odds ratio of recommendation in the post- versus pre-algorithm groups for the overall cohort and by sex and age. The rate of recommendation improved in all groups, with an odds ratio and lower limit of the confidence interval greater than 1 in all groups.

Figure 2

The rates of recommendation before and after algorithm implementation and the odds ratio of recommendation in the post- versus pre-algorithm groups by comorbidity, medication status, and discharge diagnosis. The rate of recommendation improved in all groups, with an odds ratio and lower limit of the confidence interval greater than 1 in all comorbidity and medication groups. The highest rate of recommendation was seen in patients with a comorbidity of heart failure after algorithm implementation, with a recommendation rate of 49.0%. The odds ratio and lower limit of the confidence interval was greater than 1 in patients with a discharge diagnosis of heart failure, ischemic stroke, and “other”.