Electronic Alert Systems for Patients With Acute Kidney Injury

This meta-analysis assesses the association between electronic alerts and patient survival, kidney outcomes, clinical practice patterns, and associated outcomes such as medical costs and hospital length of stay in patients with acute kidney injury.


Introduction
2][3] The introduction of electronic health record systems has enabled early detection of AKI through electronic alerts (e-alerts), considered potential interventions to reduce AKI-related complications and improve outcomes.[6][7] A 2012 study by Colpaert et al 8 using RIFLE (risk, injury, failure, loss of kidney function, and end-stage kidney disease) criteria showed that AKI e-alerts could enhance short-term renal outcomes and timely interventions.The 27th Acute Disease Quality Initiative consensus also highlighted that "AKI alerts driven by concrete criteria improve early detection and prompt AKI management." 9 Nevertheless, a 2017 published meta-analysis 6 and subsequent randomized clinical trials (RCTs) and non-RCTs, including Electronic Alerting for Acute Kidney Injury Amelioration (ELAIA)-1 10 and ELAIA-2, 11 questioned their impact on mortality.Despite assumptions about their efficacy in improving AKI outcomes and care, it remains uncertain whether AKI e-alerts, alone or with care bundles, are associated with lower mortality, AKI severity, or the need for kidney replacement or whether they impact clinical practices.
Given the lack of systematic analysis for several associated outcomes, an updated meta-analysis including recently published studies [10][11][12] is warranted.In the present study, we performed a systematic review and meta-analysis, incorporating subgroup analysis and trial sequential analysis using evidence-based medicine methods to assess the association between AKI e-alerts and patient survival, kidney outcomes, clinical practice patterns, and associated outcomes such as medical costs and hospital length of stay (LOS).

Literature Search Strategy
This systematic review and meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and checklist.We registered the protocol in PROSPERO (CRD42024527189).Two investigators (J.-J.C. and T.-H.L.) systematically and independently conducted a review of published data on outcomes in patients with AKI e-alerts.
A search of PubMed and Embase was performed on March 18, 2024, and the Cochrane Library was searched on March 20, 2024, to identify all relevant studies.Detailed search strategies, including search terms specific to each source, are provided in eTable 1 in Supplement 1.There were no limitations on language or article types.

Study Eligibility Criteria
After removing duplicates, titles and abstracts were screened by 2 reviewers (J.-J.C. and T.-H.L.) for relevance.Full texts of potentially relevant articles were then reviewed for eligibility.Inclusion criteria required studies to involve adults, compare AKI e-alert groups with non-e-alert groups, and report on any of the primary or secondary outcomes.For eligibility disagreements, a third reviewer (C.-H.C.) was consulted for consensus.Exclusions were made for duplicate cohorts, insufficient outcome data, or absence of a control group.

Data Extraction and Outcome Measurement
The 2 investigators (J.-J.C. and T.-H.L.) independently extracted data (author[s], publication year, design, location, AKI care bundle presence, sample size, AKI criteria, mean age, proportion of population that was female) and outcomes from each study.For binary outcomes, participant and event numbers were noted; for continuous outcomes, mean and SD were extracted or calculated from median (IQR).Discrepancies were resolved through discussion with a third investigator (P.-C.F.).

Outcomes
This systematic review and meta-analysis evaluated the differences between AKI e-alerts vs standard care or no e-alerts for patient outcomes or clinical practice patterns.Primary outcomes included mortality and dialysis after AKI (prioritizing 28-day or 30-day, then 60-day, 90-day, and in-hospital mortality and dialysis), AKI stage progression, and kidney recovery after AKI.Secondary outcomes were nephrologist consultations, post-AKI exposure to nonsteroidal anti-inflammatory drugs (NSAIDs), post-AKI angiotensin-converting enzyme inhibitor and/or angiotensin receptor blocker (ACEI/ARB) prescription, AKI documentation, post-AKI intravenous fluid prescription, hospital LOS, and medical costs.

Statistical Analysis
In the R meta package, the metabin and metacont functions were used for binary and continuous outcomes, respectively. 13We applied a random-effects model using the inverse variance method.
Between-study variance was estimated using the restricted maximum-likelihood estimator method, while the DerSimonian and Laird method estimated the 95% CI of the effect.We assessed the overall effect using pooled risk ratios (RRs) for binary outcomes and mean differences for continuous outcomes.Heterogeneity was evaluated with the I 2 statistic.Small study bias was examined using funnel plots and the Egger test via the metabias function. 14

Prespecified Subgroup Analysis
In our analysis, we differentiated studies as RCTs vs non-RCTs.We hypothesized that AKI e-alerts, combined with care recommendations or bundles, might be associated with patient outcomes.To explore this, we performed a subgroup analysis, dividing studies into those using e-alerts with AKI care bundles or recommendations and those using e-alerts alone.For studies reporting mortality outcomes over different time periods, we additionally conducted a subgroup analysis based on the specific time period.

Trial Sequential Analysis and Sensitivity Analysis
To determine whether the primary outcome conclusions of our meta-analysis were premature, we performed trial sequential analysis (TSA) using TSA software, version 0.9.5.10 beta. 15A more detailed description is found in eAppendix 1 in Supplement 1.
Considering that the traditional DerSimonian and Laird method might underestimate betweenstudy heterogeneity and the relatively small number of enrolled studies, we performed sensitivity analyses for binary outcomes using the Hartung-Knapp method and beta-binomial bayesian metaanalysis.The beta-binomial bayesian meta-analysis was conducted using R software and the JAGS (Just Another Gibbs Sampler) program, version 4.3.2(GNU General Public License).Additionally, we conducted further TSA including only RCTs for both primary and secondary outcomes that showed associations in the conventional meta-analysis.

Risk of Bias and Certainty of Evidence Assessment and Confidence
We assessed the risk of bias using RoB 2.0 (a revised tool to assess risk of bias in randomized trials) 16 and ROBINS-I tool (Risk of Bias in Nonrandomized Studies of Interventions) 17 for included RCTs and non-RCTs, respectively.Two independent reviewers (J.-J.C. and T.-H.L.) assessed the bias according to each domain, and the disagreements between the reviewers were resolved by discussion with another author (P.-C.F.).The quality of evidence was evaluated based on the guidelines of the GRADE (Grades of Recommendation, Assessment, Development, and Evaluation) Working Group. 18,19

Search Results and Study Characteristics
A flowchart of the literature search is provided in eFigure 1 in Supplement 1).The electronic database search identified 189 potentially eligible studies from PubMed, 98 from Embase, and 42 from the Cochrane Library.After removing duplicate articles, the remaining 259 articles were screened.After screening the titles and abstracts, the full texts of 34 studies were reviewed to assess their eligibility.

Subgroup Analysis
Subgroup analysis was performed by dividing enrolled studies into e-alerts in combination with an AKI care bundle or recommendation and those studies without.For AKI progression, studies with AKI e-alerts combined with AKI care bundle or recommendation had a lower RR compared with the non-e-alert groups (RR, 0.85 [95% CI, 0.77-0.93];P = .03for subgroup difference test) (eFigure 6 in Supplement 1).For the other 3 primary outcomes (mortality, dialysis, and kidney recovery) (eFigures 7-9 in Supplement 1) and most secondary outcomes (eFigures 10-15 in Supplement 1), there was no significant subgroup heterogeneity detected.Regarding post-AKI ACEI/ARB exposure, AKI e-alerts combined with an AKI care bundle were associated with lower RR (0.78 [95% CI, 0.70-0.88];P = .002for subgroup difference test) (eFigure 16 in Supplement 1).For studies reporting mortality outcomes over different time periods, there was no significant subgroup difference (eFigure 17 in Supplement 1).

Trial Sequential Analysis and Sensitivity Analysis
A trial sequential analysis on mortality indicated that e-alerts were unlikely to be associated with a 10% risk reduction (eFigure 18 in Supplement 1).For AKI stage progression, TSA indicated a premature conclusion (eFigure 19 in Supplement 1).For dialysis, TSA show a true-positive finding with sufficient sample size (eFigure 20 in Supplement 1) and an uncertain result regarding kidney recovery (eFigure 21 in Supplement 1).Trial sequential analysis also supported the results from conventional analysis regarding nephrologist consultations, AKI documentation, and reduced post-AKI NSAID exposure (eFigures 22-24 in Supplement 1).
Including only RCTs in the TSA, e-alerts showed a true-positive finding for dialysis, NSAID exposure, and consultation (eTable 4 in Supplement 1).Other outcomes were premature, inconclusive, or ineffective.Sensitivity analysis using the Hartung-Knapp method still showed a significantly increased the RR for dialysis.The beta-binomial bayesian meta-analysis also showed a significantly lower RR for NSAID exposure after AKI (eTable 5 in Supplement 1).

Publication Bias and Certainty of Evidence
The funnel plot for all primary and secondary outcomes are provided (eFigure 25 in Supplement 1).
There was no significant asymmetry observed in the funnel plots.The Egger tests were performed for outcomes with more than 10 studies and found no publication bias for mortality (Egger P = .13),dialysis (Egger P = .63),or nephrologist consultation (Egger P = .26).
The overall certainty of evidence (CoE) varied from moderate to very low.We summarized the results of CoE assessment in eTable 6 in Supplement 1.The detailed reasons for downgrading are provided in eTable 6 in Supplement 1 and eAppendix 3 in Supplement 1.We also summarized the results and CoE assessment (Table 2).

Discussion
This systematic review and meta-analysis highlights 4 key findings.First, AKI e-alerts may be unlikely to be associated with a 10% reduction of risk for mortality in patients with AKI, a finding supported by TSA.Second, AKI e-alerts might be associated with lower RR of AKI progression, but more research is needed to support this conclusion.Third, AKI e-alerts were linked to increased dialysis events.Fourth, AKI e-alerts seem to be associated with different clinical practices (eg, more nephrologist consultations and AKI documentation and less post-AKI NSAID exposure).

Figure 2 .
Figure 2. Association of Acute Kidney Injury (AKI) Electronic Alerts (e-Alerts) With Nephrologist Consultation and Nonsteroidal Anti-Inflammatory Drug (NSAID) Exposure

Figure 3 .
Figure 3. Association of Acute Kidney Injury (AKI) Electronic Alerts (e-Alerts) With Hospital Length of Stay (LOS), Medical Costs (B), and AKI Documentation (C)

Table 1 .
Baseline Characteristics of Included Studies reported mortality outcomes for different time periods, they were prioritized in the following order: 28 days or 30 days, then 60 days, 90 days, and finally in-hospital mortality and/or dialysis.
Abbreviations: AKI, acute kidney injury; Cr, creatinine; KDIGO, Kidney Disease: Improving Global Outcomes; NA, nonapplicable; NR, not reported; RCT, randomized clinical trial; RIFLE, risk, injury, failure, loss of kidney function, and end-stage kidney disease; UOP, urine output.SI conversion factor: To convert creatinine to μmol/L, multiply by 88.4. a Most studies did not provide overall SD of the mean age of enrolled participants; RCTs might only provide the IQR of each group.Some provide overall median age and IQR; therefore, the SD was calculated from the IQR.b If a study c Calculated from the mean and SD of 2 groups.
RCT indicates randomized clinical trial; RR, risk ratio.Diamonds indicate heterogeneity; different marker sizes, weights.
RCT indicates randomized clinical trial; RR, risk ratio.Diamonds indicate heterogeneity; different marker sizes, weights.
JAMA Network Open.2024;7(8):e2430401.doi:10.1001/jamanetworkopen.2024.30401(Reprinted) August 27, 2024 7/14 Downloaded from jamanetwork.comby guest on 09/01/2024 MD indicates mean difference; RCT, randomized clinical trial; RR, risk ratio; and USD, US dollars.Diamonds indicate heterogeneity; different marker sizes, weights.Search Strategy and Result eTable 2. Reasons for Excluding Full-Text Screening Studies eTable 3. Inclusion and Exclusion Criteria and Care Bundle or Suggestions for Enrolled Studies eTable 4. Trial Sequential Analysis With Only Enrolled Randomized Clinical Trials eTable 5. Sensitivity Analysis eTable 6. Summary of Certainty of Evidence Assessment eFigure 1. PRISMA Flow Diagram eFigure 2. Version 2 of the Cochrane Risk-of-Bias Tool for Randomized Trials (RoB 2.0) Assessment of Included Studies and Summary eFigure 3. Risk of Bias in Nonrandomized Studies of Interventions (ROBINS-I) Assessment of Included Studies and Summary eFigure 4. Forest Plot Illustrating the Association of Acute Kidney Injury (AKI) Electronic Alerts (e-Alerts) With Dialysis and Kidney Recovery after AKI eFigure 5. Forest Plot Illustrating the Association of Acute Kidney Injury (AKI) Electronic Alerts (e-Alerts) With ACEI/ARB Prescription and Fluid Prescription After AKI eFigure 6. Subgroup Analysis for AKI Progression eFigure 7. Subgroup Analysis for Mortality eFigure 8. Subgroup Analysis for Dialysis eFigure 9. Subgroup Analysis for Kidney Recovery eFigure 10.Subgroup Analysis for Nephrologist Consultation eFigure 11.Subgroup Analysis for NSAID Exposure After AKI eFigure 12. Subgroup Analysis for Hospital Length of Stay eFigure 13.Subgroup Analysis for Medical Costs eFigure 14.Subgroup Analysis for AKI Documentation eFigure 15.Subgroup Analysis for Fluid prescription eFigure 16.Subgroup Analysis for ACEI/ARB exposure eFigure 17.Subgroup Analysis for Mortality With Different Follow-Up Period eFigure 18. Trial Sequential Analysis for Mortality eFigure 19.Trial Sequential Analysis for AKI Progression eFigure 20.Trial Sequential Analysis for Dialysis eFigure 21.Trial Sequential Analysis for Kidney Recovery eFigure 22. Trial Sequential Analysis for Nephrologist Consultation eFigure 23.Trial Sequential Analysis for NSAID Exposure After AKI eFigure 24.Trial Sequential Analysis for AKI documentation eFigure 25.Funnel Plots eAppendix 1. Supplemental Method for Trial Sequential Analysis eAppendix 2. Quality of Included Studies eAppendix 3. Certainty of Evidence Assessment for AKI Progression