The effect of clinical decision support systems on clinical outcomes in acute kidney injury: a systematic review and meta-analysis of randomized controlled trials

Abstract Objectives To determine whether clinical decision support systems (CDSS) for acute kidney injury (AKI) would enhance patient outcomes in terms of mortality, dialysis, and acute kidney damage progression. Methods The systematic review and meta-analysis included the relevant randomized controlled trials (RCTs) retrieved from PubMed, EMBASE, Web of Science, Cochrane, and SCOPUS databases until 21st January 2024. The meta-analysis was done using (RevMan 5.4.1). PROSPERO ID: CRD42024517399. Results Our meta-analysis included ten RCTs with 18,355 patients. There was no significant difference between CDSS and usual care in all-cause mortality (RR: 1.00 with 95% CI [0.93, 1.07], p = 0.91) and renal replacement therapy (RR: 1.11 with 95% CI [0.99, 1.24], p = 0.07). However, CDSS was significantly associated with a decreased incidence of hyperkalemia (RR: 0.27 with 95% CI [0.10, 0.73], p = 0.01) and increased eGFR change (MD: 1.97 with 95% CI [0.47, 3.48], p = 0.01). Conclusions CDSS were not associated with clinical benefit in patients with AKI, with no effect on all-cause mortality or the need for renal replacement therapy. However, CDSS reduced the incidence of hyperkalemia and improved eGFR change in AKI patients.


Introduction
Acute kidney injury (AKi) is a prevalent grave complication that impacts hospital admissions [1].AKi affects more than 50% of patients in the intensive care unit (icU) and up to 18% of general inpatients [2].Also, AKi has significant resource and financial ramifications since it raises the risk of in-hospital mortality, the onset of cardiovascular disease, and the development of chronic kidney disease [3].early identification and effective management of AKi are essential to promote renal recovery and avoid associated negative consequences [4].
Medication errors may be decreased significantly with the use of clinical decision support (cDS) in computerized provider order entry (cPOe) and electronic medical records (eMRs) [5,6].Serum creatinine (Scr) variations are used by the automated algorithm that generates the electronic AKi alert to detect AKi early [7].clinicians can immediately act when they receive notification of an AKi episode [7].Recently, automated alerts have become a powerful tool for influencing physician behavior [8].Randomized controlled trials (Rcts) in the hospital setting have demonstrated the effectiveness of alerts in reducing medication interactions [9].
A recent compilation of research findings from multiple studies on clinical decision support systems (cDSS) found that while only 30% of studies positively impacted patient outcomes, 57% affected practitioner behavior [10].Based on these considerations, this systematic review and meta-analysis of Rcts was conducted to determine whether an automated AKi alert would reduce AKi severity and improve clinical outcomes (mortality, hyperkalemia, progression of AKi, length of hospital stay, and others) in hospitalized patients.to the best of our knowledge, this study is the most up-to-date and thorough synthesis of the available data in this quickly developing field of study.the possible conclusions drawn from this review are expected to have significant implications.they may help close a considerable knowledge gap regarding the contribution of cDSS to AKi patients' health and may also direct future medical procedures and treatment plans.

Protocol registration
the cochrane Handbook for Systematic Reviews and Meta-Analyses [11] and the PRiSMA (Preferred Reporting items for Systematic Reviews and Meta-Analyses) statement [12] guided the conduct of this systematic review and meta-analysis.the protocol for this review has been registered and published in PROSPeRO with the following iD: cRD42024517399.

Data sources and search strategy
Up until 21st January 2024, O.A. and M.A.azm searched several databases, including PubMed (MeDliNe), eMBASe, web of Science, ScOPUS, and the cochrane central Register of controlled trials (ceNtRAl).No filters or search restrictions were applied.table S1 contains more information about the search strategy, including the keywords, search terms, and the search results.

Eligibility criteria
A Population, intervention, comparison, and Outcomes (PicO) criterion was used to include Rcts: population (P): inpatient adults aged 18 years or older with AKi; intervention (i) cDSS for AKi alert and care bundle; control (c): usual care; and outcomes (O): primary outcomes: all-cause mortality and renal replacement therapy, secondary outcomes: hyperkalemia (serum potassium >5.5 mmol/l), estimated glomerular filtration rate (eGFR) change (eGFR at AKi minus the lowest eGFR within the time of randomization (from admission to the discharge), if the patient died, eGFR was considered to be equal to zero, re-hospitalization, creatinine change (serum creatinine between hospital admission and highest serum creatinine within time of randomization), number of patients received non-steroidal anti-inflammatory drugs (NSAiDs), aminoglycosides, fluids, angiotensin-converting enzyme inhibitors (Acei) and angiotensin ii receptor blockers (ARBs), contrast, AKi on discharge, nephrologist consultation, progression of AKi, renal recovery (total; <1.2 times of baseline serum creatinine, partial; ≥1.2 times and 1.5 times of baseline serum creatinine), renal ultrasonography within 2 days, length of hospital stay (days), and hospital cost.
the exclusion criteria included studies that were conducted on animals, pilot projects, unpublished study protocols, different types of observational studies, such as cohorts, case-control, cross-sectional, case series, and case reports, single-arm clinical trials, in vitro experiments on tissues and cultures, book chapters, editorials, press articles, publications that only contained abstracts or posters, and studies published in languages other than english.

Study selection
the covidence online software was used to conduct the review process.After removing any duplicate records, four reviewers independently completed them (M.t.ani, M.t.hat, O.B., and A.B.).Full-text screening was used to review the complete texts of the records that initially satisfied the eligibility requirements.Any disagreements were settled through conversations involving (O.A.).

Data extraction
Using a pre-made extraction sheet, four reviewers (M.t.ani, M.t.hat, O.B., and A.B.) separately extracted the following data: summary characteristics (study design, country, total participants, registry number, number of centers, AKi alert details, control, main inclusion criteria, follow-up duration, and primary outcome); baseline characteristics (number of participants in each group, age, gender, basal metabolic index, smoking, hospital admissions at intensive care unit (icU), AKi grade, comorbidities (diabetes, hypertension, liver disease, chronic obstructive pulmonary disease (cOPD), cerebrovascular disease, congestive heart failure, chronic kidney disease); efficacy and safety data (primary and secondary outcomes) as previously described.Any disagreements were resolved through discussion with the first author (O.A.).

Risk of bias and certainty of evidence
Using the cochrane RoB2 tool, four reviewers (M.t.ani, M.t.hat, O.B., and A.B.) independently evaluated the quality of the studies included in the research [13].the risk of bias arising from the randomization process, the risk of bias resulting from deviating from the intended intervention, the risk of bias resulting from missing outcome data, the risk of bias in measuring the outcome, and the risk of bias resulting from choosing the published results were among the domains that were assessed.if there were any conflicts, the reviewers discussed them and reached a consensus.
For the certainty of evidence, we followed the Grading of Recommendations Assessment, Development, and evaluation (GRADe) recommendations [14,15], considering inconsistency, imprecision, indirectness, publication bias, and risk of bias. the evaluation was carried out for each outcome, and the decisions were justified and documented.Any discrepancies were settled through discussion.

Statistical analysis
RevMan v5.4 was used to conduct the statistical analysis.For continuous outcomes, we used the mean difference (MD) with a 95% confidence interval (ci), and for dichotomous outcomes, we used the risk ratio (RR).we performed both the chi-square and i-square tests to evaluate heterogeneity, where the chi-square test detects the presence of heterogeneity, and the i-square test evaluates its degree.we used the random-effects model when there was significant heterogeneity (I 2 > 50%) and the common-effect model when heterogeneity was not significant (I 2 < 50%).the i-square was interpreted as follows by the cochrane Handbook (chapter nine): heterogeneity is moderate for 30-60%, substantial for 50-90%, and considerable for 75-100%.it is not significant for 0-40%. in case of significant heterogeneity, a leave-oneout sensitivity analysis was conducted to investigate the source of heterogeneity.

Search results and study selection
Five thousand and three hundred studies were screened and assessed based on their titles and abstracts during the search process.Forty-two articles moved on to full-text screening once (2287) duplicates and (2971) irrelevant studies were excluded.Finally, we included 10 Rcts (Figure 1).

Characteristics of included studies
ten randomized clinical trials involving 18,355 adults with AKi were included in our analysis, comprising 10,191 (55.5%) men and 8164 (44.5%) women.the studies were conducted in china, the USA, Germany, and the UK. the follow-up period ranged from seven days to one year.the included participants' comorbidities, baseline characteristics, and details of the included Rcts are detailed in tables 1 and 2.

Risk of bias and certainty of evidence
Seven Rcts had a low risk of overall bias, and three Rcts showed some concerns mainly attributed to the unclear risk of reporting bias due to the randomization process and deviations from the intended intervention (Figure 2).certainty of evidence is outlined in a GRADe evidence profile (table 3).

Discussion
we observed comparable outcomes between cDSS and usual care regarding all-cause mortality and renal replacement therapy.cDSS was significantly associated with a decreased incidence of hyperkalemia and an increased change in eGFR.However, there was no significant difference between cDSS and usual care regarding re-hospitalization, creatinine change, AKi at discharge, renal recovery, progression of AKi, length of hospital stay, and hospital cost.
the previous Zhao et al. [16] meta-analysis encompassed seven studies and involved 32,846 participants.the results   concerning renal replacement therapy, length of hospital stay, and progression of AKi were similar.However, the results relating to mortality differed.the study highlighted that implementing cDSS reduced overall mortality (OR 0.86; 95% ci 0.75-0.99;p = 0.040, n = 5 studies), with significant heterogeneity.Notably, that outcome included only one Rct, Selby et al. [17].it is also essential to highlight that our meta-analysis did not include Selby et al. in some outcomes.However, it was included in Zhao et al. 's meta-analysis because the authors did not report crude data.interestingly, this Rct had the most significant number of participants (24,059) and reported comparable 30-day mortality results for the pooled analysis.However, the multilevel logistic regression results for mortality showed a 13% increase in mortality in winter compared to spring.there was also an increase in mortality with older age, a decrease in mortality by 14% in females, and a higher mortality associated with a higher charlson comorbidity score [17].
Moreover, it is also worth mentioning that, according to chávez-Íñiguez et al. [18], patients with AKi and persistent hyperkalemia were correlated with the need for kidney replacement therapy, while those with changes from normal potassium to hyperkalemia were associated with death. in our study, as highlighted in the following paragraphs, there was a reduction in the incidence of hyperkalemia.these results emphasize that mortality should not be viewed as the sole determinant of primary outcomes, instead considering all aspects of the patient's health.Also, in our study, renal replacement therapy did not differ between groups, irrespective of whether a cDSS was in place.this could potentially be attributed to the fact that patients who require renal replacement therapy present with the most severe form of AKi.consequently, integrating renal replacement therapy into the ongoing management of critically ill patients could lead to an escalation in complexity and cost [19].therefore, deciding to initiate a critically ill patient on renal replacement therapy must balance the increased bedside workload and resource utilization and the preferences for care expressed by the patient and their family [20].
A significant finding of this study was that the cDSS was notably associated with a decrease in the incidence of hyperkalemia and an increase in eGFR change.Simultaneously, cDSS was insignificantly linked with a decline in the number of patients who received NSAiDs and aminoglycosides.However, due to the limited number of studies, particularly for NSAiDs and aminoglycosides, it is challenging to determine whether the reduction in hyperkalemia and increase in eGFR were results only of not taking these medications.
According to the guidelines, NSAiDs should be avoided in people with eGFR < 30 ml/min per 1.73 m 2 , and prolonged use should be avoided in those with eGFR between 30 and 59 ml/min per 1.73 m 2 [21].Furthermore, NSAiDs are known to cause hyperkalemia [22], not to mention the nephrotoxic effect of aminoglycosides, which require dose adjustment in cases of renal failure [23].these results, however, indicate the usefulness of cDSS in the early detection of patients with higher risk of AKi and in tailoring their medication strategy.However, a comprehensive approach to the work process is essential.this perspective has been explored by Marcilly et al. [24] in their DeteciP project, which yielded several noteworthy observations.Preliminary findings from geriatric and cardiology units indicate that despite the overall similarities in work processes, the management of iatrogenic risks of hyperkalemia/AKi is influenced by the availability and location of physicians and clinical pharmacists.Pharmacists conduct pharmaceutical analyses of orders in a relatively similar manner across units, irrespective of the use of the pharmacist cDSS.However, the physical proximity between residents and pharmacists varies between units. in the cardiology clinic, the shared office space facilitates cooperation, whereas in the geriatric unit, the pharmacist's location outside the ward hinders collaboration.consequently, the authors suggest that implementing pharmacist cDSS may yield different outcomes in cardiology and geriatrics units for AKi/hyperkaliemia prevention or management due to these variations [24].therefore, future trials investigating cDSS may also explain the work process in which the cDSS was implemented for a holistic approach analysis.

Strength and limitations
Our study demonstrates several strengths.First, the comprehensive literature search covered five databases until 21st January 2024, to ensure that relevant studies are included, minimizing selection bias.Second, the study adheres to a robust research design by including only Rcts.third, including ten Rcts with over 18,355 adults enhances the statistical power of the analysis.
Still, this meta-analysis was subject to a few limitations.First, we primarily reported the pooled analysis, and it was not possible to conduct a subgroup analysis based on the cDSS form mentioned above or based on follow-up.this was mainly due to the various implemented cDSS forms and inconsistent follow-up periods.Still, the pooled analysis for primary outcomes showed homogeneous results.Second, some outcomes showed heterogeneity, such as AKi at discharge, length of hospital stay, number of patients who received NSAiDs, nephrologist consultation, and number of patients who received fluids.therefore, we performed a sensitivity analysis and resolved all heterogeneous results except for nephrologist consultation and length of hospital stay.third, our analysis relies on pooled data from diverse studies, each with inherent limitations.these limitations are shared across the included studies.Fourth, we could not appropriately assess asymmetry in funnel plots due to the small number of studies (<10) included in our analysis.

Implications for future research
the decision to implement a cDSS in a healthcare facility should be based on a process gap that needs to be filled or supported.in this study, we have provided several forms of cDSS, including systems for recognizing high-risk patients with AKi for clinical pharmacist intervention [25], early warning systems for a rise in serum creatinine concentration, immediate consultation with a specialist, and issuing a patient kidney passport [26].we also included a care bundle involving fluid administration, potentially nephrotoxic medication discontinuation, and nephrology consultation [27].each form of cDSS was designed for specific purposes, and its usefulness should be limited within the scope of these purposes.For instance, cDSS were generally observed to improve AKi care delivery.However, improving AKi care delivery does not necessarily  translate exclusively to lower mortality [17].Still, as observed in our analysis, it can also lead to improved clinical functions, such as eGFR and reduced unwanted side effects, such as hyperkalemia.
Finally, while the hospital cost seems insignificantly higher with cDSS, the length of hospital stay was insignificantly shorter.However, considering the large number of patients with AKi, i.e., ~750,000 patients per year in the United States [28], this could account for a substantial health-economic benefit.we recommend that healthcare facilities conduct a cost-effectiveness analysis before implementing the cDSS system, considering direct operational and maintenance costs and indirect costs associated with administrative functions for healthcare professionals.

Conclusion
cDSS did not prevent all-cause mortality or decrease the need for renal replacement therapy in patients with AKi but reduced the incidence of hyperkalemia and increased eGFR.
tailoring cDSS to the specific needs of healthcare facilities can further amplify its impact.However, additional research is essential to ensure long-term sustainability and successful real-world implementation, especially when standardizing cDSS across multiple facilities.

Figure 1 .
Figure 1.PRiSMa flow chart of the screening process.

Figure 2 .
Figure 2. Quality assessment of risk of bias in the included trials.a schematic representation of risks (low = red, unclear = yellow, and high = red) for specific types of biases of each of the studies in the review.

Figure 3 .
Figure 3. Forest Plot of the primary outcomes (all-cause mortality and renal replacement therapy), RR: risk ratio, Ci: confidence interval.

Figure 4 .
Figure 4. (a) Forest plot of the hyperkalemia; (B) forest plot of the eGFR; (C) forest plot of the re-hospitalization; (D) forest plot of the creatinine change.RR: risk ratio; Ci: confidence interval.

Figure 5 .
Figure 5. (a) Forest plot of the number of patients who received nSaiDs; (B) forest plot of the number of patients who received aminoglycosides; (C) forest plot of number of patients who received aCei and aRBs; (D) forest plot of number of patients received contrast.RR: risk ratio; Ci: confidence interval.

Table 1 .
Summary characteristics of the included RCTs.

Table 2 .
Baseline characteristics of the participants.
na na na: not available; SD: standard deviation; BMi: body mass index; iCu: intensive care unit; aKi: acute kidney injury; CDSS: clinical decision support systems: COPD: chronic obstructive pulmonary disease.

Table 3 .
GRaDe evidence profile.a wide confidence interval that does not exclude the risk of appreciable harm/benefit with a low number of events.c a wide confidence interval that does not exclude the risk of appreciable harm/benefit.d Wu et al. had an overall concern of bias with 31.7% of pooled analysis weight.
Ci: confidence interval; MD: mean difference; RR: risk ratio.a Thomas et al. had an overall concern of bias with 65% of pooled analysis weight.b e I 2 > 75%.f Wu et al. had an overall concern of bias with 27% of pooled analysis weight.g I 2 > 50%.