An Organizational-Level Program of Intervention for AKI: A Pragmatic Stepped Wedge Cluster Randomized Trial

Abstract

Background Variable standards of care may contribute to poor outcomes associated with AKI. We evaluated whether a multifaceted intervention (AKI e-alerts, an AKI care bundle, and an education program) would improve delivery of care and patient outcomes at an organizational level. Methods A multicenter, pragmatic, stepped-wedge cluster randomized trial was performed in five UK hospitals, involving patients with AKI aged $18 years. The intervention was introduced sequentially across fixed three-month periods according to a randomly determined schedule until all hospitals were exposed. The primary outcome was 30-day mortality, with pre-specified secondary endpoints and a nested evaluation of care process delivery. The nature of the intervention precluded blinding, but data collection and analysis were independent of project delivery teams. Results We studied 24,059 AKI episodes, finding an overall 30-day mortality of 24.5%, with no difference between control and intervention periods. Hospital length of stay was reduced with the intervention (decreases of 0.7, 1.1, and 1.3 days at the 0.5, 0.6, and 0.7 quantiles, respectively). AKI incidence increased and was mirrored by an increase in the proportion of patients with a coded diagnosis of AKI. Our assessment of process measures in 1048 patients showed improvements in several metrics including AKI recognition, medication optimization, and fluid assessment. Conclusions A complex, hospital-wide intervention to reduce harm associated with AKI did not reduce 30-day AKI mortality but did result in reductions in hospital length of stay, accompanied by improvements in in quality of care. An increase in AKI incidence likely reflected improved recognition.

A Pragmatic Step Forward: AKI and Beyond

“…One of the greatest challenges for embedded pragmatic trials is achieving adherence to the intervention. Lower than desired uptake of the care bundle components likely reduced the effect of the intervention in the trial by Selby et al 2 This trial may be the first in nephrology to use stepped wedge cluster randomization. Cluster randomization is often used in pragmatic trials to simplify implementation of the intervention by clinical personnel and reduce the likelihood of contamination across treatment arms.…”

“…Specific strategies for supportive care have been recommended in consensus guidelines, but they have not been tested in rigorous randomized clinical trials. 1 In this issue of the Journal of the American Society of Nephrology, Selby et al 2 report the results of a multicenter, pragmatic, stepped wedge cluster randomized trial of a multicomponent supportive care intervention for AKI. The trial was conducted in five hospitals in the United Kingdom over a 27-month period, during which there were 24,059 episodes of AKI.…”

“…6 By leveraging the clinical infrastructure, embedded pragmatic trials markedly reduce the cost of research, and because interventions are tested under the real world conditions in which they will ultimately be applied, interventions found to be effective should be readily adopted and sustained after the trial has ended. The trial by Selby et al 2 was highly pragmatic: the eligibility criteria were nonrestrictive; the settings were diverse; the investigators used clinically acquired rather than trial-generated data; the intervention was implemented by clinicians rather than researchers; and by design, the intervention was tailored to fit local needs. One of the greatest challenges for embedded pragmatic trials is achieving adherence to the intervention.…”

“…For example, because the timing for moving from the control period to the intervention period is established for each cluster before the start of the trial and typically cannot be modified, lower than anticipated enrollment during a control period cannot necessarily be addressed by extending the duration of enrollment, a strategy often employed in trials that use parallel group randomization. Stepped wedge cluster randomization seemed to work well for Selby et al, 2 and it is likely that their experience will be helpful to other investigators weighing the advantages and disadvantages of this approach.…”

Beyond Biomarkers: Machine Learning in Diagnosing Acute Kidney Injury

“…Such a study using e-alerts for AKI defined by serum creatinine level, and not machine learning surveillance as in the present study, found a reported increase in AKI, a reduced length of stay, and improvements in the quality of care but no change in the 30-day AKI mortality rate. 17 Perhaps machine learningemediated surveillance, with earlier recognition of AKI, will offer additional improvements in the care of the patient with AKI. The exciting prospect that machine learningemediated surveillance may confer salutary outcomes in AKI and reduce attendant morbidity and mortality should be examined in further studies.…”