Generating evidence using electronic alerts during routine care: a fully automated randomised controlled trial of oral fluid restriction in acute heart failure (THIRST Alert Trial)
Yang Chen, Daniel Higgins, Yogini Jani, Nausheen Saleem, Kris Chafer, Ewan McFarlane, Timothy Roberts, Steve Harris, Bryan Williams, Matthew R Sydes, Folkert W Asselbergs, Anoop D Shah, R Thomas LumbersAbstract
Background
Many medical treatments lack robust evidence of safety and effectiveness from randomised controlled trials (RCTs). This is partly due to the cost and complexity of performing traditional RCTs and because randomisation is not routinely embedded in clinical care. We aimed to evaluate the feasibility of conducting a pragmatic RCT fully integrated into the electronic health record (EHR) system to streamline patient identification, randomisation, treatment allocation, and outcome assessment in patients admitted with acute heart failure.
Methods and Findings
THIRST Alert was a single-centre parallel-group, open-label, feasibility RCT embedded in a hospital EHR system from May 3 to November 1, 2023. Adult patients who received more than one dose of intravenous furosemide within 48 hours of admission were eligible. An interruptive alert was triggered when physicians accessed the medication order chart of eligible patients, inviting them to enrol the patient in the study. Enrolled patients were randomised to either ‘oral fluid restriction of 1L per day’ or ‘no fluid restriction’. The co-primary feasibility outcomes were the total number of patients recruited and the documented difference in oral fluid intake within 48 hours of randomisation. 23 patients (16%) were enrolled from 145 eligible patient admissions; there were no repeat admissions among enrolled patients. A total of 1,191 enrolment alerts were triggered, reaching 216 individual clinicians. 22/23 trial participants (96%) had a diagnosis of heart failure. No significant difference in oral fluid intake was observed between the treatment groups (median difference 518 mL; 95% CI –235 to 1270; P = 0.18). Documented oral intake was numerically higher in the oral fluid restriction group (1168 mL; IQR 932–1620 mL) compared to the group without restriction (650 mL; IQR 75–1102 mL), which may reflect documentation bias rather than a true increase in oral fluid intake or a lack of intervention efficacy.
Conclusion
This study demonstrates the feasibility of a pragmatic RCT fully integrated within a hospital EHR system in acute care. Although alert-to-enrolment rate was modest, the overall recruitment rate was comparable to conventional acute care RCTs, highlighting the potential of EHR-embedded trials to efficiently address evidence gaps in the management of conditions such as acute heart failure. ClinicalTrials.gov NCT05869656.