Sustained Improvements After Intervention to Prevent Contrast‐Associated Acute Kidney Injury: A Randomized Controlled Trial

Abstract

Background

In the IMPROVE AKI (A Cluster‐Randomized Trial of Team‐Based Coaching Interventions to Improve Acute Kidney Injury) trial, a combination of team‐based coaching and data‐driven surveillance dashboards reduced the odds of AKI following cardiac catheterization by 46%. The objective of this study was to determine if improvements in AKI outcomes would be sustained after completion of the active intervention.

Methods and Results

A 2×2 factorial cluster‐randomized trial with an 18‐month active intervention phase (October 2019–March 2021) and an 18‐month sustainability phase (April 2021–September 2022) conducted among cardiac catheterization laboratories in 20 Veterans Affairs sites. Interventions included team‐based coaching in a virtual learning collaborative or technical assistance, with and without access to an automated surveillance reporting dashboard. Data were collected on procedures involving adult patients undergoing diagnostic coronary angiography or percutaneous coronary interventions and not receiving chronic dialysis. The main outcome was AKI within 7 days of cardiac catheterization among all participants and those with preexisting chronic kidney disease. In addition, survey and focused interview data were collected to understand barriers and facilitators to sustaining AKI improvements. In this phase, 440 of 4160 patients experienced AKI, including 216 of 1260 patients with chronic kidney disease. Compared with technical assistance alone, we observed a reduction in AKI among virtual learning collaborative + automated surveillance reporting sites (adjusted odds ratio, 0.60 [95% CI, 0.42–0.86]). Sites had implemented standardized orders (11), oral and intravenous hydration standing orders (13), and contrast limiting protocols (10).

Conclusions

Team‐based coaching coupled with data‐driven surveillance dashboards reduced AKI by 40% during the 18 months after active participation in the trial. Process improvement education, care process standardization, and automated outcome feedback may be effective and durable methods for reducing AKI.

Registration

URL: https://clinicaltrials.gov/; Unique Identifier: NCT03556293.

Nonstandard Abbreviations and Acronyms

AKI

acute kidney injury

ASR

automated surveillance reporting

EBI

evidence‐based interventions

VLC

virtual learning collaborative

TA

technical assistance

Clinical Perspective.

What Is New?

• Team‐based coaching and data dashboards are sustainable interventions for preventing acute kidney injury following cardiac catheterizations.

What Are the Clinical Implications?

• One‐time resource‐intensive implementation interventions can reduce acute kidney injury rates following cardiac catheterizations up to 18 months after the intervention

Although clinical research strongly supports the identification and testing of evidence‐based interventions (EBI), less focus is placed on evaluating the sustainability of the intervention outside the original and often contrived trial conditions. ¹ , ² , ³ Randomized controlled trials, even pragmatic ones, create a curated research environment dedicated to testing the validity of the intervention. ¹ , ⁴ , ⁵ In these circumstances, short‐term improvements in patient outcomes do not necessarily translate into sustained improvements over time. ¹ Moreover, the realities of routine clinical practice (eg, fewer resources, different workflow, etc.) or differences in a patient population (ie, groups underrepresented in research) that were carefully controlled during the trial may create significant challenges in the sustainment of EBIs. ¹ As such, the sustainability of EBIs following active interventions is considered a critical and understudied area of translational research. ²

Strategies to prevent contrast‐associated acute kidney injury (AKI) following cardiac catheterization are well known ⁶ , ⁷ and extensively researched, ⁸ , ⁹ , ¹⁰ , ¹¹ yet their effectiveness beyond the trial period remains unclear. ¹² , ¹³ , ¹⁴ In response to the critical need to evaluate the sustainability of previously validated AKI prevention tools, the IMPROVE AKI (A Cluster‐Randomized Trial of Team‐Based Coaching Interventions to Improve Acute Kidney Injury) trial ¹⁴ passively monitored fidelity to the AKI intervention(s) and AKI events for 18 months after cessation of the active intervention. In addition, the study conducted surveys and focused interviews with staff implementing the AKI prevention tools to better understand the barriers and facilitators to sustainment. The objective of this study was to evaluate the sustainability of the AKI outcomes after the intervention was withdrawn and to better understand factors affecting the sustainment of AKI prevention tools.

METHODS

To ensure scientific rigor and reproducibility, all data collection and analyses will be overseen by a steering committee discussed in the governance and organizational structure of the leadership team and research program. An analytic program file used to produce the analyses and a deidentified data set will be made available to third parties within the Veterans Affairs (VA) system to conduct confirmatory analyses for the proposed specific aims. All parties must be in compliance with VA regulations and access. All analytic code available upon request to the research team.

IMPROVE AKI is registered at ClinicalTrials.gov (identifier: NCT03556293). Information on its methodology, including the Consolidated Standards of Reporting Trials list, was published elsewhere ¹⁴ , and the protocol is available in Supplement Methods 1. Briefly, this was a 2×2 factorial design hybrid type 1 effectiveness implementation trial conducted within cardiac catheterization laboratories in VA medical centers. Sites were randomized to 1 of 4 interventions (see Interventions section for more details), and randomization via a random number generator occurred at the hospital level in a 2×2 factorial design where block randomization was stratified by VA medical center region (1–4) and patient volume (<800, 800+). ¹⁵ Sites that elected to participate were enrolled by research staff and were blinded to their assigned intervention until approved by the lead statistician. The VA central institutional review board approved the trial with waiver of patient informed consent because the intervention was provider focused and recommending routine care delivery.

Trial Methods

Eligible patients included those aged 18 years+ who underwent a diagnostic coronary angiogram or percutaneous coronary intervention at participating sites between October 1, 2019, and March 31, 2021. Patients were excluded if they had a history of hemodialysis or peritoneal dialysis. Enrolled patients were evaluated for chronic kidney disease (CKD), defined as an estimated glomerular filtration rate <60 mL for the most current record within the year before the procedure. ¹⁶ If preprocedure estimated glomerular filtration rates were unavailable, the presence of a CKD diagnosis code in the year before the procedure was used to define preexisting CKD. All 76 VA cardiac catheterization laboratories were invited to participate, and 20 were accepted and randomized.

AKI Prevention Toolkit

All 20 participating sites were provided the AKI Prevention Toolkit, which included 3 core interventions: (1) standardized order sets, (2) increased intravenous and oral fluids, and (3) reduced contrast volume. ¹⁴ , ¹⁷ All interventions were supported by the Kidney Disease Improving Global Outcomes as effective strategies for preventing AKI. ¹⁸ Instructions for implementing these interventions in clinical practice were developed and tested in a pilot study. ¹⁷

Interventions

A 2×2 factorial design allows for a simultaneous comparison of multiple interventions. ¹⁹ These trials typically have 2 factors, each of which has 2 levels. ¹⁹ For this trial, the 2 factors were the implementation strategies (Table 1), and the 2 levels included the presence or absence of a surveillance reporting tool. Technical assistance (TA) and team‐based coaching with a virtual learning collaborative (VLC) were the 3 implementation strategies. The automated surveillance reporting (ASR) dashboard provided site‐level risk‐adjusted AKI performance over time and performance comparisons against other VA sites regionally and nationally. Sites were randomized into 1 of 4 groups as follows: (1) TA, (2) TA+ASR, (3) VLC, or (4) VLC+ASR.

Table 1.

Intervention Details

Intervention	Description
Implementation strategy
Technical assistance	60‐min monthly meeting led by AKI improvement specialist for each site Reviewed AKI Prevention Toolkit and discussed site implementation questions
Virtual learning collaborative	Sites built multi‐disciplinary teams with interventional cardiologists, laboratory managers and technicians, nurses, cardiology administration, nephrology, and quality improvement departments Sites were assigned quality improvement and collaborative improvement coaches 60‐min group monthly training calls with structured agenda covering education on one of the clinical interventions in the toolkit, team updates, quality improvement knowledge
Information support
Automated surveillance reporting	Sites had access to dashboards displaying AKI performance over time Estimates of risk‐adjusted performance the site level and comparisons to regional and national Veterans Affairs benchmarks Control charts to highlight temporal trends in site‐level performance In‐depth patient‐level data with key features related to AKI risk and procedure information
No automated surveillance reporting	Access to routinely collected care patient data

AKI indicates acute kidney injury.

Sites receiving TA were involved in monthly calls where site team members could ask questions and discuss challenges. Sites participating in VLC created multidisciplinary teams, were assigned coaches from the research team, and met monthly on structured calls that included education on the clinical interventions in the toolkit, quality improvement planning, and team progress updates. Sites randomized to ASR were provided access to the surveillance dashboard and discussed the tool during their respective monthly calls. Sites not randomized to receive ASR as part of their intervention had access to standard clinical information systems and existing local data displays.

The active intervention phase of the trial lasted 18 months (October 2019 through March 2021). In the subsequent 18 months (sustainability phase: April 2021 through September 2022), the monthly TA and VLC calls stopped, and research staff did not contact site teams. The ASR dashboard was maintained during this period and remained available if site teams chose to access it independently. Sites were not reminded of this tool's availability during this period.

Outcome

The primary clinical outcome was AKI occurring within 7 days from the cardiac catheterization procedure. AKI was defined using Kidney Disease Improving Global Outcomes: ≥0.30 mg/dL or ≥50% increase in serum creatinine over the last baseline or onset of dialysis. ²⁰ Preprocedural serum creatinine was used to establish the baseline and calculated as the most recent creatinine from 365 days before up to and including the day of the procedure. The 30‐day AKI was evaluated as a secondary outcome. If multiple postprocedure serum creatinine were measured, the maximum value within days 1 to 7 or 1 to 30 was used to define the 7‐day and 30‐day outcomes, respectively.

Data Collection

Data were obtained from the Veterans Health Information Systems and Technology Architecture/Computerized Patient Record System electronic health record. Most data were collected from the Corporate Data Warehouse, which receives a nightly copy of the electronic health record. A portion of the data was collected from the Clinical Assessment, Reporting, and Tracking System for Cardiac Catheterization Laboratories application within the electronic health record, which specifically collects detailed preprocedural and periprocedural information for patients undergoing cardiac catheterization. ²¹ The AKI risk model supporting the ASR dashboard leveraged these 2 data sources. ²² The original AKI risk model was adjusted before the action phase and then maintained through the pandemic via a novel algorithmic vigilance and updating system. ²³ , ²⁴

Power and Sample Size

The target enrollment in each intervention arm was 1983 participants. Using data from national VA catheterization data, the AKI proportion among patients with CKD was assumed to be 0.2700 and 0.2025 under the null and alternative hypotheses, respectively. The intracluster correlation coefficient was informed from the intervention arm of a pilot study and was assumed to be 0.0009. ²⁵ Using an F‐test based on a normal approximation of the distribution of the proportion of AKI cases under each of the 4 intervention strategies, the power to detect any difference across these strategies against the null was >99%.

Statistical Analysis

Hierarchical logistic regression models were used to estimate the effect of the implementation strategy and information support interventions, including their interactions for 7‐day AKI. Using data on procedures occurring in the 18 months following the end of the active intervention phase of the trial, we fit models to compare AKI outcomes with team‐based coaching (VLC), surveillance dashboard with technical assistance (TA+ASR), and the combination of both (VLC+ASR) against technical assistance alone (TA). ¹⁴ To further understand whether any impact of the interventions varied by study phase (active versus sustainability) and improve the estimation of effects with larger available sample sizes, we also fit a model using all procedures during the 36 months of active intervention and the sustainability phase. This model included an interaction between phase and intervention. All models were adjusted for the following patient‐level characteristics: age, race, tobacco use, anemia, heart failure, CKD, diabetes, hypertension, prior percutaneous coronary intervention, prior coronary artery bypass grafting, prior myocardial infarction, peripheral vascular disease, ejection fraction, shock, kidney function, and procedural urgency. ²² Race was categorized as Asian American, Black, American Indian, Pacific Islander, and Unknown versus White due to sample size restrictions. Models included site‐level random effects and baseline risk‐adjusted performance 12 months before the trial. Satterthwaite small‐sample correction was used due to the small number of clusters. Pairwise comparisons were adjusted for multiple comparisons using the Bonferroni adjustment. All analyses were conducted on the entire population and patients with CKD. All described analyses were repeated for 30‐day postprocedural AKI. Analyses were performed using R 4.1.2. ²⁶

Qualitative Assessments

A survey was sent to all participating sites to assess what process changes occurred during the action phase of the trial and whether any of those process changes stopped or were changed during the sustainment phase of the trial (Supplement Methods 2). The survey results informed semistructured interviews conducted with participants at 36 months by providing information regarding which AKI toolkit strategies had been implemented and sustained. Thirteen sites participated in the interviews (1 VLC+ASR site completed a survey but did not participate in an interview). The semistructured interviews were conducted through open‐ended inquiry into barriers and facilitators to implementation and sustainability at the practice/clinic level (eg, staffing, organizational culture) and outer context (eg, patient characteristics, policies). The interview guide is available in Supplement Methods 3.

Interviews were completed with participants from 13 sites, were 20 to 30 minutes long, and audio‐recorded for transcription and analysis. Transcripts were managed and qualitatively coded using Atlas.ti desktop (version 9.1.3). Guided by the Consolidated Framework for Implementation Research, We conducted multiple rounds of coding, including an initial round to index text excerpts into large categories (barriers to implementation or sustainment by prevention strategy; facilitators of implementation or sustainment by prevention strategy) followed by focused coding of these text excerpts into inductively derived descriptive categories of specific barriers and facilitators at the practice/clinic level and outer context. We aggregated data in code reports, and through analytic memoing, we developed key thematic findings regarding barriers and facilitators to implementation and sustainment.

RESULTS

Quantitative Results

During the 18 months sustainability phase, sites conducted 4160 eligible procedures involving 3939 unique patients (Table 2). In the subpopulation with CKD, there were 1260 procedures from 1201 unique patients. Across study arms, patients varied by race, ethnicity, tobacco use, anemia, history of percutaneous coronary intervention, myocardial infarction type, urgency status, and inpatient status of the procedure.

Table 2.

Study Population by Intervention for 18‐Month Sustainability Phase

Patient and procedural characteristics	All study sites	TA	TA+ASR	VLC	VLC+ASR
	N (%)	N (%)	N (%)	N (%)	N (%)
Total procedures	4160	518	1192	1385	1065
Unique patients	3939	502	1092	1321	1024
Age, y (median [interquartile range])	71 [64–75]	71 [64–75]	70 [64–75]	70 [63–75]	71 [64–75]
Female sex	149 (4)	19 (4)	42 (4)	49 (4)	39 (4)
Race
Black	892 (21)	60 (12)	335 (28)	383 (28)	114 (11)
White	2894 (70)	418 (81)	784 (66)	885 (64)	807 (76)
Other*	374 (9)	40 (8)	73 (6)	117 (8)	144 (14)
Hispanic ethnicity	435 (11)	35 (7)	57 (5)	161 (12)	182 (17)
Tobacco use	1588 (38)	202 (39)	523 (44)	514 (37)	349 (33)
Anemia	2008 (48)	260 (50)	642 (54)	645 (47)	461 (43)
Congestive heart failure	1274 (31)	147 (28)	403 (34)	396 (29)	328 (31)
Chronic kidney disease	1260 (30)	150 (29)	373 (31)	396 (29)	341 (32)
Diabetes	2131 (51)	265 (51)	624 (52)	711 (51)	531 (50)
Hypertension	1162 (28)	134 (26)	359 (30)	411 (30)	258 (24)
Prior PCI	1388 (33)	151 (29)	501 (42)	429 (31)	307 (29)
Procedure type
Catheterization only	2421 (58)	274 (53)	621 (52)	859 (62)	667 (63)
PCI only	592 (14)	98 (19)	162 (14)	225 (16)	107 (10)
Catheterization and PCI	1147 (28)	146 (28)	409 (34)	301 (22)	291 (27)
Urgency
Elective	2339 (56)	272 (53)	615 (52)	675 (49)	777 (73)
Emergent	169 (4)	10 (2)	68 (6)	73 (5)	18 (2)
Urgent	1649 (40)	236 (46)	508 (43)	635 (46)	270 (25)
Salvage	3 (0)	0 (0)	1 (0)	2 (0)	0 (0)
Inpatient status on day of procedure	2943 (71)	361 (70)	910 (76)	1120 (81)	552 (52)
Myocardial infarction type
Non‐STEMI	915 (22)	142 (27)	255 (21)	306 (22)	212 (20)
STEMI	110 (3)	12 (2)	42 (4)	47 (3)	9 (1)
Neither	3135 (75)	364 (70)	895 (75)	1032 (75)	844 (79)
Number of diseased vessels†
0	279 (16)	20 (8)	101 (18)	86 (16)	72 (18)
1	1313 (76)	199 (82)	436 (76)	388 (74)	290 (73)
2	139 (8)	24 (10)	32 (6)	50 (10)	33 (8)
3	8 (1)	1 (0)	2 (0)	2 (0)	3 (1)
Number of stents†
0	405 (23)	33 (14)	221 (39)	75 (14)	76 (19)
1	796 (46)	129 (53)	222 (39)	273 (52)	172 (43)
2	376 (22)	58 (24)	89 (16)	124 (24)	105 (26)
3	99 (6)	13 (5)	24 (4)	37 (7)	25 (6)
4	49 (3)	9 (4)	12 (2)	14 (3)	14 (4)
5	7 (0)	1 (0)	2 (0)	2 (0)	2 (1)
6	4 (0)	1 (0)	0 (0)	1 (0)	2 (1)
7	1 (0)	0 (0)	1 (0)	0 (0)	0 (0)
8	2 (0)	0 (0)	0 (0)	0 (0)	2 (1)

ASR indicates automated surveillance reporting; PCI, percutaneous coronary intervention; STEMI, ST‐segment–elevation myocardial infarction; TA, technical assistance; and VLC, virtual learning collaborative.

*Other race is defined as patient race reporting other than White or Black.

^†Excluding cath‐only procedures.

Compared with the 18‐month active intervention phase, patient demographics and comorbidities did not vary significantly across phases, although sustainability phase procedures were more likely to be elective (56% versus 52%) and patients undergoing catheterization only (58% versus 55%; Table S1). Some secular trends occurred in all arms of the study, particularly during the COVID‐19 pandemic when elective procedures were initially not performed and then gradually reintroduced over time (Figures S1 and S2).

The proportion of procedures in which patients developed AKI within 7 days was 11% (n=440) during the sustainability phase and 11% (n=510) during the active intervention phase. ¹⁴ In the subpopulation with CKD, AKI occurred after 17% (n=216) and 18% (n=235) of procedures in the sustainability and active intervention phases, respectively. ¹⁴ AKI was least common in the VLC+ASR arm during both study phases (7% and 8% among all procedures during the sustainability and active intervention phases, respectively; among procedures in the subpopulation with CKD, the corresponding values were 13% and 14%). Overall, approximately 50% of postprocedural creatinine values were missing, which could not be assessed, and Table S2 provides a summary of creatinine missingness across both phases of the trial by arm.

Table 3 provides absolute and case‐mix adjusted outcomes by intervention for the sustainability phase. Adjusting for multiple comparisons, the combination of VLC+ASR was associated with a significant reduction in the odds of 7‐day AKI compared with TA alone in both study phases (sustainability phase odds ratio [OR], 0.60 [95% CI, 0.42–0.86], P=0.006; active intervention phase OR, 0.54 [95% CI, 0.40–0.74], P=0.0001). ¹⁴ The Figure displays ORs in phase‐specific and combined models. Evaluating both phases together, there was a 43% reduction in the odds of 7‐day AKI among sites with VLC+ASR compared with TA alone (OR, 0.57 [95% CI, 0.45–0.72]; P<0.008). This translates to an absolute 5% decrease in case‐mix adjusted AKI rate in the VLC+ASR sites compared with TA alone during the sustainability phase (adjusted difference, −5.0 [95% CI, −5.2 to −4.8]; Table 3). Across all study arms, case‐mix adjusted AKI rates were approximately 1% lower in the sustainability phase than in the active intervention phase (Table 3). Similar patterns were observed within the subpopulation with CKD, although comparisons did not achieve statistical significance.

Table 3.

Seven‐Day Postprocedural AKI Outcomes in the Sustainability Phase by Intervention

	Total procedures	No. 7‐d AKI events	% 7‐d AKI				7‐d AKI by stage
			Absolute	Case‐mix* adjusted	Adjusted difference from TA group	Adjusted change from active phase	Stage1	Stage2	Stage3
							N (%)	N (%)	N (%)
All procedures
TA	518	62	12.0	13.3 [13.1 to 13.5]	…	−1.1 [−1.4 to −0.8]	57 (11.0)	4 (0.8)	1 (0.2)
TA+ASR	1192	127	10.7	10.2 [10.1 to 10.4]	−3.0 [−3.3 to −2.8]	−1.2 [−1.4 to −1.0]	109 (9.1)	10 (0.8)	8 (0.7)
VLC§	1385	178	12.9	10.9 [10.7 to 11.0]	−2.4 [−2.7 to −2.2]	−0.7 [−0.9 to −0.5]	152 (11.0)	12 (0.9)	14 (1.0)
VLC+ASR	1065	73	6.9	8.3 [8.2 to 8.4]	−5.0 [−5.2 to −4.8]	−0.8 [−1.0 to −0.6]	66 (6.2)	2 (0.2)	5 (0.5)
Procedures in population with chronic kidney disease
TA	150	26	17.3	18.6 [18.2 to 19.0]	…	−0.7 [−1.3 to −0.1]	23 (15.3)	2 (1.3)	1 (0.7)
TA+ASR	373	76	20.4	18.8 [18.4 to 19.2]	0.2 [−0.3 to 0.8]	−1.9 [−2.5 to −1.3]	62 (16.6)	6 (1.6)	8 (2.1)
VLC	396	71	17.9	15.7 [15.3 to 16.0]	−2.9 [−3.5 to −2.4]	−0.3 [−0.8 to 0.2]	60 (15.2)	2 (0.5)	9 (2.4)
VLC+ASR	341	43	12.6	15.9 [15.5 to 16.2]	−2.7 [−3.3 to −2.2]	−1.4 [−1.9 to −0.8]	37 (10.9)	2 (0.6)	4 (1.2)

AKI indicates acute kidney injury; ASR, automated surveillance reporting; TA, technical assistance; and VLC, virtual learning collaborative.

^*Case‐mix adjusted rates assume patient feature distributions in each intervention group were the same as the distribution among all eligible procedures in the combined active intervention and sustainability phases.

Figure 1. Hierarchically risk‐adjusted odds ratios for 7‐day AKI comparing interventions to technical assistance in active intervention phase, sustainability phase, and both phases combined for all patients (A) and only among patients with CKD (B).

Patient characteristics were included in adjustment, such as age, race, tobacco use, anemia, heart failure, CKD, diabetes, hypertension, prior percutaneous coronary intervention, prior coronary artery bypass surgery, prior myocardial infarction, peripheral vascular disease, left ventricle ejection fraction, shock status, laboratory‐based creatinine assessment, procedural urgency, and site baseline‐level performance. AKI indicates acute kidney injury; and CKD, chronic kidney disease.

There were no significant interactions between implementation strategy and information support interventions in the phase‐specific or combined models. In the model combining active intervention and sustainability phase data, no significant interaction existed between the study phase and intervention. Both the implementation strategy and information support interventions had significant main effects in the combined model (P<0.001).

Findings were similar in sensitivity analyses using 30‐day AKI outcomes (Table S3; Figure S3). Evaluating both phases together, there was a 39% reduction in the odds of 30‐day AKI among sites with VLR+ASR compared with TA alone (OR, 0.61 [95% CI, 0.48–0.77]).

Qualitative Results

Results related to implementing the AKI toolkit during the action phase of the trial are reported elsewhere. ⁸ Among the 20 participating sites, 14 submitted survey responses, and 13 participated in semistructured interviews. Among the 14 sites that responded to the survey, 11 reported using standardized orders, 13 reported using standing orders for oral and intravenous hydration, and 10 reported contrast‐limiting protocols. A summary of the survey responses regarding whether the sites implemented standardized orders, oral and intravenous hydration, and contrast limiting protocols is shown in Table 4. The full list of barriers and facilitators to implementing and sustaining the AKI toolkit during the sustainability phase is available in Tables S4 through S7. Broadly, there were no discernable patterns to barriers and facilitators by prevention strategy (of the AKI toolkit). Participants reported that the key barriers to implementing the AKI toolkit centered on staff turnover and limited staffing; challenges related to organizational culture, including lack of buy‐in and resistance to change; building and using standardized I order sets; fitting new practices into existing workflows; coordinating with other departments to implement new practices; and patient‐level challenges, including limited follow‐up. In terms of challenges sustaining the AKI toolkit, participants reported staff turnover, limited staffing, and limited patient follow‐up as key barriers.

Table 4.

Summary of the Survey Responses From the Sites Indicating Whether They Had or Implemented Electronic Health Record Standardized Cardiac Catheterization Orders, Whether They Implemented a Protocol for Oral and Intravenous Hydration for Patients, and Whether They Implemented a Contrast Limitation Protocol

Category	Implemented standardized cardiac catheterization orders	Implemented oral and intravenous orders	Implemented contrast limiting protocol
TA	2/3 (66%)	3/3 (100%)	1/3 (33%)
TA+ASR	1/3 (33%)	2/3 (66%)	1/3 (33%)
VLC	3/3 (100%)	3/3 (100%)	3/3 (100%)
VLC+ASR	5/5 (100%)	5/5 (100%)	5/5 (100%)

ASR indicates automated surveillance reporting; TA, technical assistance; and VLC, virtual learning collaborative.

The key facilitators of implementing the AKI toolkit prevention strategies identified by participants highlight the importance of organizational culture in shaping implementation. Participants noted strong buy‐in and the presence of local champions as central to sustaining prevention strategies. Other key facilitators included high‐functioning teams, with nurse involvement and communication/coordination between team members, as well as robust coordination with other departments. Additional facilitators of sustainment included having permanent I order sets in place, maintaining awareness of the prevention strategies, and access to data (Table 4).

DISCUSSION

In this cluster‐randomized trial, patients in sites in the VA randomized to team‐based coaching with ASR dashboards were less likely to develop AKI compared with sites receiving TA alone (control) even when the active intervention was complete. Sites that received the VLC+ASR interventions had a 40% reduction in the odds of AKI among all patients undergoing cardiac catheterization, and the overall reduction in the combined active intervention and sustainability phases was 43%. To the authors' knowledge, this study is the first to evaluate the sustainability of EBI for preventing AKI following cardiac catheterization, and the findings support the notion that a time‐limited, intensive (and costly) health care system investment and engagement could result in sustained improvements in the incidence of AKI following cardiac catheterization. The sustained reduction in AKI outcomes suggests the AKI toolkit may offer long‐term effectiveness in preventing AKI, up to 18 months following the initial implementation. Moreover, the results offer early support for disseminating the intervention to other clinical communities. ²⁷ Key barriers to AKI toolkit implementation and sustainment were also identified via qualitative interviews and included limited staffing, organizational culture, and standardized I order sets. In contrast, key facilitators included strong local champions, high‐functioning teams, and nurse involvement. Together, these results suggest one‐time intensive multi‐modal educational interventions with dedicated staffing and organizational resources can have lasting impacts on AKI rates following cardiac catheterizations.

Care bundles, like the AKI toolkit, are a structured set of evidence‐based practices, treatments, or EBIs, that can improve care and outcomes when implemented. ²⁸ Existing randomized controlled trials evaluating care bundle implementation for cardiac catheterization identified important reductions in the rate of AKI, including a 21% reduction in the Northern New England study, ¹⁷ an 18% reduction in the Canadian Contrast RISK (Contrast Reducing Injury Sustained by Kidneys) trial, ¹² an 68% reduction in POSEIDON (Prevention of Contrast Renal Injury With Different Hydration Strategies), ²⁹ and a 46% reduction in the active phase of the IMPROVE AKI study. ¹⁴ Although these results are important, evidence of their sustainability outside clinical trials is crucial to evaluate the effectiveness of the intervention in the real world. ² , ³⁰ Despite the importance of sustainability evaluation following cessation of clinical trial interventions, this research largely remains understudied, ¹ , ² , ³ with only 23% of EBIs being sustained after 2 years following the intervention. ³¹ , ³²

Central to understanding the sustainability of AKI outcomes following the AKI toolkit intervention is identifying the processes and factors that may have facilitated or hindered its continuation. ² Qualitative results from our study speak directly to such factors. Barriers and facilitators affecting compliance with AKI toolkit implementation sustainment represent contextual factors known to influence the successful implementation and sustainment of quality improvement initiatives. ³³ , ³⁴ Like IMPROVE AKI, previous studies identified organizational resources, culture, and leadership, along with staff buy‐in and workflow constraints, as barriers to the successful implementation of pediatric cardiopulmonary resuscitation procedures, ³³ geriatric medical home models, ³⁵ and veterans support programs. ³⁶ As identified in the IMPROVE AKI trial, prompting strong team communication and coordination, ³⁷ identifying a local champion, ³⁸ and ensuring a clear prevention strategy ² are consistently cited as important facilitators to sustaining EBI in practice.

Limitations

There are potential limitations in the IMPROVE‐AKI sustainability assessment. First, site‐level randomization may have generated a patient‐level imbalance between sites. To address this limitation, the analysis adjusted for patient‐level characteristics. In addition, models were generated to account for baseline features to address concerns about between‐cluster imbalance. However, any patient‐level imbalance in the observed data did not confirm or deny unmeasured confounding, which is why cluster randomization was implemented. Second, approximately 50% of postprocedural creatinine values were missing and could not be assessed.

Because AKI occurs in this population, it is a crucial patient safety requirement to ensure patients have serum creatinine measurements shortly after catheterization. Missingness within each arm was consistent across pre, action, and sustainability phases. Third, it is possible sites in the trial accessed operational data for surveillance and quality improvement, including control sites, which potentially reduces the strength of the findings. Fourth, research staff interacted with sites after the sustainability phase to prevent attention bias, possibly introducing recall bias in qualitative assessments. Fifth, the annual catheterization volume of participating sites was <500 procedures (per site, per year), limiting the generalizability of findings, especially to facilities outside the VA or larger medical centers. Future work will scale this study's interventions outside the VA and into larger academic medical centers. Finally, implementation outside the VA may be challenged by data access, but a comparable approach could be used within other data networks to support care improvement.

Conclusions

In conclusion, this sustainability study in the IMPROVE AKI trial showed improvements in the postprocedural AKI rates sustained in the 18 months following the active intervention phase. Using an initial intensive combined coaching and ongoing informational data and analytic synthesis support may provide lasting reductions in the rates of AKI among patients undergoing cardiac catheterization. Furthermore, the successful sustainment of the AKI toolkit may be influenced by contextual attributes, including staff resources, strong leadership, and workflow dynamics.

Sources of Funding

The IMPROVE AKI cluster randomized trial was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (Grant/Award Number: R01DK113201) (multiple principal investigators Brown, Matheny, and Solomon).

Disclosures

Agarwal reports employment with VA Medical Center, Wright State University, and an advisory or leadership role for the American College of Cardiology Board of Trustees—Ohio Chapter (an unpaid position). S.A. Athar reports employment with Loma Linda VA Medical Center. E. Carpenter‐Song served as a consultant for Westat during 2022 to support the use of qualitative methods in 2 studies: (1) a study examining an employment program for veterans transitioning to civilian life and (2) a study examining evidence‐based supported employment for young adults with serious mental illnesses. E. Carpenter‐Song was part of a research team that received funding in 2020 from the Bristol Meyers Squibb Foundation to conduct research examining the impact of COVID‐19 on rural health systems and communities. S. Girotra reports research funding from National Heart, Lung, and Blood Institute (R01HL160734, R56HL158803, and R01HL166305). T.J. Helton reports employment with James H. Quillen VA Medical Center and ownership interest in Apple, AirBNB, ASML holding, Confluent, Crowdstrike, Microsoft, Shopify, and Upstart. C. Leung reports employment with Orlando Veterans Affairs Medical Center. M.E. Matheny reports employment with Department of Veterans Affairs and consultancy agreements with NIH‐VA‐DoD Pain Management Grant Consortium (PMC3), and membership in Informatics & Methods Section, SMRB Study Section, VA HSR&D; Steering Committee—Indianapolis VA HSR&D COIN Center; and Steering Committee—VA HSR&D VIREC. A.J. O'Malley reports consultancy agreements with JB Associates and an advisory or leadership role for Statistics in Medicine and Observational Studies. M.E. Plomondon reports employment with Veterans Health Administration, Washington, DC. R. Solomon reports consultancy agreements with MediBeacon, Inc., PLC Inc., and Sonogenix, Inc.; research funding from REATA and Vera Pharmaceuticals; and advisory or leadership roles for MediBeacon, PLC Med, Inc., and Sonogenix. M.I. Vidovich reports consultancy agreements with Boston Scientific, research funding from Boston Scientific, patents or royalties from Merit Medical, and an advisory or leadership role for Intersocietal Accreditation Commission. S.W. Waldo reports employment with Rocky Mountain Regional VA Medical Center and research funding from Cardiovascular Systems Incorporated and Janssen Pharmaceuticals. L. Zubkoff reports employment with Birmingham VA Healthcare System. All remaining authors have nothing to disclose.

Supporting information

Data S1

Tables S1–S7

Figures S1–S3