Randomized trial of AKI alerts in hospitalized patients
Wilson, Francis (2020), Randomized trial of AKI alerts in hospitalized patients, Dryad, Dataset, https://doi.org/10.5061/dryad.4f4qrfj95
Objective: To determine whether electronic health record (EHR) alerts for Acute Kidney Injury (AKI) would improve patient outcomes of mortality, dialysis and progression of AKI.
Design: Double-blinded, multicenter, parallel, randomized, controlled trial of an electronic AKI alert versus usual care (no alert). Participants were electronically identified and randomized via a best practice alert build using simple randomization with allocation concealment.
Setting: Six diverse hospitals (four teaching and two non-teaching) ranging from small community hospitals to large tertiary care centers.
Participants: 6,030 adult inpatients with AKI, as defined by the Kidney Disease: Improving Global Outcomes (KDIGO) creatinine criteria.
Interventions: An EHR-based “pop-up” alert for AKI with an associated AKI order set upon provider opening of the patient’s medical record.
Main Outcome Measures: A composite of AKI progression, receipt of dialysis, or death within 14 days of randomization. Pre-specified secondary outcomes included per-hospital outcome rates and rates of various AKI care practices.
Results: 6,030 patients were randomized over 22 months. The primary outcome occurred in 653 (21.4%) patients in the alert group and 622 (20.9%) in the usual care group (relative risk 1.02, 95% confidence interval [CI] 0.93 to 1.13, p=0.67). Per-hospital analysis revealed worse outcomes in the two non-teaching hospitals (N=765, 13%), where alerts were associated with a relative risk of the primary outcome of 1.49 (95% CI, 1.12 to 1.98, p=0.006). More deaths (15.6% in the alert group vs. 8.6% in the usual care group) occurred at these centers (p=0.003). Certain AKI care practices were increased in the alert group but did not appear to mediate these outcomes.
Conclusions: Alerts did not reduce rates of our primary outcome among hospitalized patients with AKI. The overall lack of clinical benefit and signals of harm in non-teaching hospitals should lead to a re-evaluation of existing AKI alerting systems.
Trial Registration: ClinicalTrials.gov NCT02753751.
Time to Event Analysis Note: The "time_to" data reflects the time from randomization to the event of interest, discharge, or 14 days (whichever comes first). Be careful when analyzing this! In the paper, we change this value to 14 for people who were discharged alive prior to day 14 without the event of interest. This assumes, for example, that if a patient was not dialyzed prior to their (alive) discharge on day 10 after randomization, they were not dialyzed at day 14 (a relatively safe assumption). But we leave the data in more raw form in this dataset to facilitate other analyses.
National Institutes of Health, Award: R01DK113191
National Institutes of Health, Award: P30DK079310