Background: Heart failure is common and confers high mortality and morbidity. Our system sought to improve heart-failure-patient care, working to reduce mortality index (MORTi) and 30-day readmission. The effort included a combination of education and incentives followed by EHR-based clinical decisions support systems (CDSS) within order sets, notes, and more. At our recently-merged hospitals none of the interventions applied (control). The CDSS was deployed in Epic, but a subset of hospitals use Cerner, allowing a natural experiment to explore incremental effects.
Methods: Patients were analyzed by seasonal controls – “Pre1” (7/1/23-10/31/23) and “Pre2” (12/1/23-6/30/24)” for two intervention periods – “Post1” (7/1/24-10/31/24 – education and incentives) and “Post2” (12/1/24-6/30/25 – EHR CDSS addition). Data was obtained from Vizient and analyzed in Excel. Changes of interest were Post2 versus Pre2 and Post1 versus Pre1. Data is reported as “index” (observed/Vizient-model expected). Given interacting variables, results are reported for descriptive purposes only.
Results: A total of 17,280 (30-day readmission) and 17,692 (MORTi) patients were analyzed across 20 hospitals. Control saw MORTi change across Pre1->Pre2->Post1->Post2 1.04->0.91->0.69->1.12 and 30-day readmission index (30Di) 1.10->1.21->1.28->1.24. Cerner hospitals (education and incentives in Post1/Post2) saw MORTi change 0.63->0.77->0.70->1.20 and 30Di change 1.61->1.40->1.19->1.14. Finally, Epic (Post1/Post2: education and incentives; Post2: EHR CDSS added) saw MORTi change 1.24->1.21->1.21->0.82 and 30Di change 1.30->1.19->1.13->1.06.Exploring seasonally-compared ratios, control saw a Post1vPre1 MORTi ratio of 0.66 (0.78 observed ratio/1.18 expected ratio) and Post1vPre1 30Di ratio of 1.17 (1.17/1.01). By comparison, the Post1vPre1 Cerner MORTi ratio was 1.11 (1.22/1.10) and 30Di ratio was 0.74 (0.75/1.01) while Epic MORTi ratio was 0.97 (0.92/0.95) and 30Di ratio was 0.87 (0.88/1.00). With the addition of EHR CDSS affecting Epic hospitals in Post2, control Post2vPre2 MORTi ratio was 1.23 (1.18/0.96) and 30Di ratio was 1.03 (1.04/1.01). Cerner MORTi ratio was 1.55 (1.44/0.93) and 30Di ratio was 0.82 (0.81/1.00). Finally, Epic MORTi ratio was 0.68 (1.00/1.49) and 30Di ratio was 0.89 (0.94/1.05).
Conclusions: The results suggest education and incentives have limited impact on MORTi but are associated with improved 30Di as Post1vPre1 relative MORTi was modestly reduced (0.97) for Epic hospitals and higher for Cerner hospitals (1.11) while the control was lower (0.66) as compared to the 30Di ratio was 0.97, 0.74, and 1.17 for Epic, Cerner, and control, respectively. In Post2, though there were not additive EHR CDSS 30Di effects – ratios were 0.89, 0.82, and 1.03 for Epic, Cerner, and control, respectively, there was an outsized MORTi effect. The MORTi ratio was 0.68 for Epic, whereas Cerner (1.55) and control (1.23) were higher. This 0.68 was composed of an unchanged observed ratio (1.00) and increased expected (1.49). Given controls had higher observed ratios (1.44 and 1.18) this suggests there were improvements in both unadjusted outcomes and expected outcomes. We suspect this was driven by CDSS diuretic and GDMT emphasis and complexity-capture support, respectively. If these CDSS improvements continue, it will be worth sharing as a 32% relative reduction in heart failure MORTi would have meaningful benefit to patients nationally.