Abstract

Introduction There is uncertainty about the effectiveness of oral fluid restriction in patients with acute heart failure treated with intravenous loop diuretics, due to a lack of high-quality evidence from randomised controlled trials (RCTs). RCTs of non-pharmacological therapies are challenging and costly to undertake in the acute setting using conventional approaches. We sought to investigate the feasibility of conducting a pragmatic RCT that was integrated into the electronic health record (EHR), in the setting of acute unplanned care. Our primary aim was to determine the feasibility of using an automated, interruptive alert to invite clinicians to enrol patients into an RCT of fluid restriction in patients presenting with fluid overload. Methods THIRST Alert was a single-centre parallel group, open-label, randomised controlled trial conducted to pilot a novel and efficient approach to trial conduct in a hospital setting. Patient screening, recruitment, randomisation, and outcome ascertainment were all conducted through the EHR and routine care processes. A proportionate model of verbal opt-out consent was used. Over a 6-month period, clinicians who prescribed more than one dose of intravenous furosemide within 48 hours of an unplanned patient admission were exposed to an alert which invited them to assess whether their patient was suitable for inclusion into the trial. Patients <18 and those admitted to the care of surgical or maternity teams were excluded. Enrolled patients underwent simple 1:1 randomisation by the EHR, and were allocated to either oral fluid restriction of 1L/ day or no oral fluid restriction. The co-primary outcome measures were the number of patients enrolled and the documented difference in oral fluid intake between the intervention and control group in the 48 hours after randomised allocation. The trial did not involve any additional investigator input or patient follow-up. Results Between 3 May 2023 and 1 November 2023, a total of 1,191 alerts were triggered. 23 of 141 eligible patients (16%) were enrolled on the trial by routine care clinicians (table 1). In 21 of 23 patients (91%), there was evidence of adherence to the randomised treatment allocation: a clinical order concordant with the randomised allocation was recorded in the health record in 12/12 patients allocated to fluid restriction arm and 9/11 patients allocated to no restriction (figure 1). For 19/23 (83%) of the patients enrolled in the study, heart failure was included in the hospital episode statistics for the admission. In the intention to treat (ITT) analysis, the documented oral intake of the restricted group was 1170 ml (930–1620 ml IQR) and 650 ml (75–1100 ml IQR) in the unrestricted group. The mean number of documented entries for the primary outcome measure of fluid intake was 6.6 for the restricted group and 4.9 for the unrestricted group. Conclusions To our knowledge, the THIRST Alert trial is one of the first pragmatic RCTs delivered entirely through the electronic health record system and without the direct intervention of a research team. Our study demonstrates the feasibility of conducting low-cost and efficient trials during the routine care process to generate evidence that can inform practice and improve patient outcomes. Paradoxically, we observed a higher documented fluid intake in the fluid restriction group which may indicate differences in measurement and documentation in patients with active treatments. Further studies are required to determine whether oral fluid restriction is an effective adjunct to diuretic treatment in this setting or if it is a low-value intervention that contributes to care complexity and patient thirst without clinical benefit. Trial registration: NCT05869656. Funded by NIHR UCLH Biomedical Research Centre. Conflict of Interest None