Editorial 

September 6, 2022

Clinical Decision Support to Prevent Acute Kidney Injury After Cardiac Catheterization: Moving Beyond Process to Improving Clinical Outcomes

Eric R. Gottlieb, Mallika Mendu

JAMA. 2022;328(9):831-832. doi:10.1001/jama.2022.14070

Over the past decade, with the advent of meaningful use as part of the 2009 Health Information Technology for Economic and Clinical Health (HITECH) Act, the US health care system has experienced a digital transformation, ensuring that the electronic health record is foundational to the practice of medicine.¹ Electronic health records offer the possibility of aggregating myriad discrete clinical data to inform real-time practice in the form of clinical decision support. However, for numerous reasons (including a focus on process as opposed to outcome metrics, disruption of established clinical workflows, and inadequate resourcing), clinical decision support has largely fallen short.²

Case studies of successful clinical decision support interventions can provide valuable information on best practices, including institutional and policy environments that facilitate this type of innovation. In this issue of JAMA, James et al³ report a pragmatic, cluster randomized clinical trial combining education, clinical decision support–based algorithms, and scheduled audit and feedback to study the effect of this multifaceted intervention on the development of acute kidney injury (AKI) after intravenous contrast administration during cardiac catheterization.

The authors used a step-wedged cluster design, initially including 34 interventional cardiologists (participants) in Alberta, Canada. Clusters of participants were randomized to undergo crossover from the preintervention period (provided usual care and served as the control group) to the intervention period (8 start dates; included only 31 participants due to 3 retirements). The trial included 3 elements. First, prior to the intervention period, participants received a 1-hour training session related to AKI and prevention strategies. Second, they were provided with clinical decision support during catheterization procedures, including automated predictions of AKI risk, calculated optimal intravenous contrast volume, a hemodynamic-guided intravenous fluid volume target that used left ventricular end-diastolic pressure, and real-time guidance on a safe patient-specific contrast volume target coupled with alerts when the target was reached. Third, the participating cardiologists received quarterly audits with data on adherence to intravenous contrast volume guidelines, intravenous fluid administration, and AKI incidence among their patients who were at high risk.

The authors included 4032 patients in the intervention (4327 catheterization procedures) and demonstrated a reduction in AKI incidence to 7.2% during the intervention period from 8.6% during the preintervention (control) period (time-adjusted odds ratio [OR] accounting for clustering, 0.72 [95% CI, 0.56-0.93], P = .01). This was accompanied by reductions in mean intravenous contrast volume to 93.1 mL (SD, 61.2 mL) during the intervention period from 112.7 mL (SD, 67.7 mL) during the control period, and the proportion of cases exceeding the patient-specific contrast volume target was reduced to 38.1% from 51.7%, respectively (time-adjusted OR, 0.77 [95% CI, 0.65-0.91]). There was an increase in mean volume of intravenous crystalloid administered to 851.4 mL (SD, 596.4 mL) during the intervention period from 650.0 mL (467.4 mL) during the control period and a decrease in the proportion of eligible patients who received less than the hemodynamic-guided intravenous fluid volume target to 60.8% from 75.1%, respectively (time-adjusted OR, 0.68 [95% CI, 0.53-0.87]). There were no significant differences in the rates of major adverse cardiovascular events or major adverse kidney events, although the study was not powered to detect these differences.

In addition to the notable clinical outcomes observed in this pragmatic trial by James et al,³ the study design is a model for effective development, implementation, and evaluation of clinical decision support focused on meaningful clinical outcomes. The authors addressed many common pitfalls related to clinical decision support efforts by (1) selecting a clinically relevant outcome, (2) establishing clinician trust and buy-in, (3) incorporating algorithms based on consistent, evidence-based best practices with the aim of reducing variation, (4) ensuring seamless integration into clinical workflow with clear, actionable guidance at the point of care, and (5) establishing accountability with ongoing feedback.

The choice of target condition is important because resources should be invested in addressing problems that are common, costly, subject to variation in practice, and have modifiable outcomes such as AKI and, in this study specifically, contrast-induced AKI. Although the true population risk of contrast-induced AKI is debated, the risk is likely significant in this cohort of patients undergoing catheterization,⁴ of whom more than half had diabetes and one-third had heart failure. A recent meta-analysis⁵ showed a 9% risk of contrast-induced AKI after angiography (120 studies; 974 898 patients) and a 0.5% risk of requiring dialysis (111 studies; 858 305 patients). A 2012-2013 French study⁶ that involved 1 047 329 hospital stays estimated that contrast-induced AKI increased hospital length of stay by 15.8 days and increased the cost of hospitalization by more than €12 000 (approximately $15 000 adjusted for inflation).

To optimize practice, James et al³ leveraged published studies that had demonstrated improved clinical outcomes, including the POSEIDON (Prevention of Contrast Renal Injury With Different Hydration Strategies) trial,⁷which showed that in patients with chronic kidney disease and 1 or more comorbidities, fluid administration guided by left ventricular end-diastolic pressure was associated with a reduced risk of AKI. Another controlled study aimed at initiation of kidney replacement therapy for patients with AKI,⁸ which was based on a similar rationale of variation in clinical practice influencing outcomes, showed that implementation of a point-of-care algorithm (122 patients) compared with a sham control (102 patients) was associated with reduced intensive care unit length of stay, hospital length of stay, and use of kidney replacement therapy in cases of physician-perceived treatment futility.

The intervention outlined in the study by James et al³ was consistent with study design principles from the burgeoning field of implementation science.⁹ The initial educational session likely promoted clinician awareness and garnered buy-in for the intervention.¹⁰ The graphical display of key data points, which was paired with in-person reminders from procedure staff, provided clear, actionable guidance at the point of care and increased attention to intravenous contrast volume.¹¹ In addition, the quarterly audit aligned with principles of effective practice feedback and accountability, such as including data specific to the practitioner and stratified by AKI risk, preempting pushback based on variation in severity of illness and baseline comorbidities.¹²

Clinical innovators such as physicians, nurses, pharmacists, and other allied health professionals depend on adequate resourcing to provide technical, administrative, and financial infrastructure. It is notable that the authors received support to implement this innovative clinical decision support approach not only from their institution, but also from the provincial Canadian government, including a dedicated innovation grant. Although the HITECH Act facilitated the use of digital technology in health care in the US, implementation has been variable and has not led to robust infrastructure across all health systems.¹³ In addition, despite seminal grants focused on funding pragmatic care delivery innovation studies, funding for dissemination and implementation research is limited in contrast to traditional bench and epidemiological research.¹⁴ There is also an unmet opportunity for the Centers for Medicare & Medicaid Services to implement quality measures that reflect clinical outcomes, such as contrast-induced AKI, which could incentivize investment in improvement efforts such as clinical decision support tools.¹⁵

The study by James et al³ has several limitations. As the authors note, the trial included a relatively small number of cardiologists in a single region in Canada, and further work is needed to determine whether the findings are generalizable and durable. Notably, patients (68% in the intervention group and 67% in the control [preintervention] group) and participating cardiologists (87%) in this trial were predominantly male. A major limitation is that data on race and ethnicity were not reported. Although the authors took steps to minimize the risk of contamination from secular trends in event rates due to the pre-post design, the possibility of regression to the mean cannot be ignored.

The report by James et al³ in this issue of JAMA demonstrates that a novel and comprehensive clinical decision support intervention that goes beyond process metrics improves outcomes for high-risk patients undergoing cardiac catheterization. It serves as a model for innovation that could be applied across health care settings and specialties. Future iterations of clinical decision support that are focused on clinical outcomes and are fully integrated into clinical workflows at the point of care could facilitate more efficient, reliable care and could begin to be used to address disparities in care delivery. The full potential of clinical decision support, when effectively designed, has yet to be realized.

Back to top

Article Information

Corresponding Author: Eric R. Gottlieb, MD, MS, Hospitalist Physician Services, Mount Auburn Hospital, 300 Mount Auburn St, Cambridge, MA 02138 (eric.gottlieb@mah.org).

Conflict of Interest Disclosures: Dr Gottlieb reported having an ownership interest in Reveal Pharmaceuticals Inc. No other disclosures were reported.