B44-04 Impact of an Artificial Intelligence (AI) Enabled Clinical Decision Support System for COPD Identification and Referral on Follow-up and Outcomes in Patients After an Acute Exacerbation of COPD Resulting in a Hospital Visit: A Multicenter Study Design

Rationale GOLD recommends timely follow-up to reduce readmissions and mortality in COPD patients following a severe exacerbation. However, 30-day follow-up rates remain suboptimal and range from ∼8% to 44% with a pulmonologist. Optimizing care coordination between the hospital visit and pulmonary clinics presents a key gap and opportunity to improve follow-up rates. We hypothesize that an AI-enabled clinical decision support system (CDSS) that is tailored to COPD and embedded within the emergency departments (EDs) and specialist workflows may constitute a novel and scalable approach for diverse hospital systems and patient populations. This study will investigate the ability of this first-of-its-kind system to improve specialist follow-up and patient outcomes after an acute exacerbation. Methods In a multicenter, longitudinal, pre-post implementation study (NCT07118306), we will deploy a CDSS, Viz COPD (Viz.ai, Inc.), across the EDs and pulmonary clinics in five geographically dispersed U.S. healthcare systems. A post-implementation cohort of 485 patients will be compared to a matched control retrospective cohort of patients who presented to the same study sites under the standard of care clinical workflow before CDSS implementation (Figure 1). Additionally, we will independently analyze an exploratory cohort of historical patients with no prior pulmonologist follow-up for potential follow-up after CDSS implementation. Inclusion criteria: ≥ 40 years, COPD diagnosis, dual or triple long-acting maintenance inhaler therapy (LABA/LAMA, LABA/ICS, or LABA/LAMA/ICS), and an acute exacerbation resulting in a hospital visit. Results Our primary endpoint will compare the proportion of patients with a pulmonologist follow-up visit within 30 days after discharge in the matched cohorts. Secondary endpoints will examine time from discharge to a pulmonologist follow-up; the proportion of pulmonologist follow-up within 90 days; exacerbation rate, healthcare resource utilization, cardiovascular events, and all-cause mortality at 30-, 90-, and 365-days post-exacerbation; and all-cause readmissions and the proportion of patients with pharmacological treatment consistent with the GOLD treatment recommendations at 30 and 90 days. Conclusion We aim to validate a novel approach to post-exacerbation COPD care by examining the ability of an AI-enabled clinical decision support system to improve pulmonologist follow up and clinical outcomes. The results of this study will inform clinical practice and may provide a new care pathway for reducing morbidity and mortality in this vulnerable patient population.