Effects and characteristics of clinical decision support systems on the outcomes of patients with kidney disease: a systematic review

ABSTRACT Objectives This systematic review was conducted to investigate the characteristics and effects of clinical decision support systems (CDSSs) on clinical and process-of-care outcomes of patients with kidney disease. Methods A comprehensive systematic search was conducted in electronic databases to identify relevant studies published until November 2020. Randomized clinical trials evaluating the effects of using electronic CDSS on at least one clinical or process-of-care outcome in patients with kidney disease were included in this study. The characteristics of the included studies, features of CDSSs, and effects of the interventions on the outcomes were extracted. Studies were appraised for quality using the Cochrane risk-of-bias assessment tool. Results Out of 8722 retrieved records, 11 eligible studies measured 32 outcomes, including 10 clinical outcomes and 22 process-of-care outcomes. The effects of CDSSs on 45.5% of the process-of-care outcomes were statistically significant, and all the clinical outcomes were not statistically significant. Medication-related process-of-care outcomes were the most frequently measured (54.5%), and CDSSs had the most effective and positive effect on medication appropriateness (18.2%). The characteristics of CDSSs investigated in the included studies comprised automatic data entry, real-time feedback, providing recommendations, and CDSS integration with the Computerized Provider Order Entry system. Conclusion Although CDSS may potentially be able to improve processes of care for patients with kidney disease, particularly with regard to medication appropriateness, no evidence was found that CDSS affects clinical outcomes in these patients. Further research is thus required to determine the effects of CDSSs on clinical outcomes in patients with kidney diseases.


Introduction
Given the growing prevalence of kidney diseases worldwide [1,2], taking measures to prevent their progression to endstage renal disease (ESRD) is crucial [3]. In 2018, WHO estimated annual mortalities from kidney diseases at 5-10 million people in the world [4]. Kidney disease is costly in terms of its treatment and follow-up, as it uses a major portion of healthcare resources [5,6]. Over 2% of the healthcare budget of highincome countries is annually allocated to providing critical care services for patients with kidney diseases [4,7].
The efficiency of clinical decision support systems (CDSSs) has risen in recent years [8]. CDSSs can support numerous clinical tasks and provide physicians with recommendations by matching patient characteristics with clinical databases. Integrating CDSS with electronic health records (EHR) and computerized provider order entry (CPOE) has extended the application of CDSS to patient monitoring, assisting physicians with diagnoses and providing reminders and alerts in case of any deviations in the results of laboratory tests, drug-drug interactions, prescription errors, and contraindications such as allergy [9].
Evidence suggests CDSS-based interventions improve outcomes in patients with chronic diseases [10][11][12]. The applications of CDSSs to kidney diseases include identifying high-risk patients [13], diagnosing and managing acute kidney injury (AKI) and determining its severity [14,15], diagnosing and managing chronic renal failure (CRF) [16], controlling blood pressure and anemia [17,18], monitoring the patient status, evaluating the kidney function to improve prescriptions and reduce prescribing errors [19][20][21][22][23], and estimating direct costs of medications [24]. CDSSs have also been used to assess and improve outcomes such as the quality and safety of medication prescription and consumption [25][26][27], frequency of dialysis use, length of stay and in-hospital mortality [28], attitudes of healthcare providers [29] and compliance with guidelines [30].
To summarize these findings, a systematic review is required. To the best of the authors' knowledge, a single systematic review has been conducted so far to evaluate the effects of CDSS on outcomes in patients with kidney disease. This systematic review focused on the effects of CDSSs on prescribing medications for kidney disease and revealed that the computerized and manual CDSSs were found to enhance the physician's performance in terms of prescription, e.g. frequency of appropriate dosing [31].
Although CDSS-based interventions can improve the outcomes of many diseases, no systematic review has yet assessed the effects of CDSSs on all the outcomes associated with kidney diseases, i.e. process-of-care outcomes and clinical outcomes. The present study thus summarized the results of studies on the effects of CDSSs on different outcomes in kidney diseases. The questions raised were: 'What are the effects of CDSSs on clinical and process-of-care outcomes in patients with kidney disease?' and 'What are the characteristics of the CDSSs used for these patients?.'

Sources of data and search strategies
A comprehensive systematic search was conducted in the electronic databases of Medline (via PubMed), Scopus, Cochrane, ISI, and Embase to identify relevant studies published until 14 January 2020. The literature search was last updated on 21 November 2020. The search strategy included a combination of keywords and MeSH terms related to CDSS and also kidney disease in the title and abstract ( Table 1). The reference lists of both the included studies and similar reviews were also reviewed to ensure that all the relevant studies had been found.

Eligibility criteria
The selection criteria were determined according to the population, intervention, comparison, outcomes, and study design (PICOS) format [32]. The study inclusion criteria were: (1) Patients with kidney disease and their caregivers, such as physicians or other healthcare providers, were included in the study; (2) electronic CDSSs were used to care for patients with kidney disease; a CDSS was defined as an information system designed to improve clinical decisions made by physicians and other healthcare providers [33]; and (4) The study design was RCT, with a control group receiving standard/usual care without using any CDSS.
Descriptive studies without a comparison group, quasiexperimental studies, reviews, case reports, letters to the editor, study protocols, conference proceedings, and dissertations; and studies published in languages other than English were excluded.

Data extraction
Two reviewers independently screened the titles and abstracts of all the studies identified through the electronic searches. The full-text of any studies deemed potentially relevant was retrieved and assessed for final inclusion. Potential disputes were resolved by consulting with the third reviewer.
The data extracted from the included studies using a structured form consisted of details of the study such as author names, country and year of publication, study design, and study population; intervention details such as its duration and setting; descriptions of intervention and control groups; intervention functionalities; and the measured outcomes. The first reviewer extracted the data, which was then independently checked and approved by the second and third reviewers.

Synthesis and analysis of data
No meta-analysis was conducted because of the variability and heterogeneity of the outcomes in the included studies. A narrative synthesis of the evidence was carried out based on the classification of the outcomes and the interventions' characteristics and effects. Similar to a systematic review conducted by Goldzweig et al. (2015) [34], the characteristics of the interventions were classified according to IT design, data entry source, and implementation characteristics. Similar to a systematic review conducted by Jeddi et al. (2017) [35], the outcomes were classified as clinical outcomes and process-ofcare outcomes. The effects of the interventions were also classified as statistically significant positive, no effect (not statistically significant) and mixed effects (both positive and without an effect).

Study selection
As shown in the PRISMA flow diagram (Figure 1), a total of 6552 records remained after the removal of duplicates from 8722 extracted records. After reviewing the titles and abstracts according to the eligibility criteria, 17 records were selected for full text assessment, four of which were later excluded given that their intervention [37] or study designs [17,24,38] mismatched the inclusion criteria. Two others were also excluded owing to applying CDSS to the control group and measuring the effects of complementary interventions on CDSSs [39,40]. Finally, a total of 11 studies met all inclusion criteria [41][42][43][44][45][46][47][48][49][50][51]. Table 2 presents the general characteristics of the included studies. The first study was published in 2009, and the most recent in 2020. The studies included seven (63.6%) from the United States [43][44][45]47,48,50,51], one each from Canada [42], China [49], and Germany [46] and one was jointly conducted in eight European countries (Bulgaria, Croatia, Germany, Italy, Latvia, Poland, Romania and Serbia and Montenegro) [41]. The duration of the interventions varied from 3 months to 2 years (median: 12 months, Q1:7, Q3:15). The sample sizes of the studies ranged from 248 to 32,917 patients (median: 854, Q1:494, Q3:4010) and 30 to 514 physicians (median: 80, Q1:36, Q3:333.5). In total, six studies involved adult patients with chronic renal failure as the study population, one study involved physicians, and four studies involved both patients and physicians.

Assessment of the risk of bias
The results of the quality assessment of the included studies are shown in Figure 2; 63.6% of the studies had specifically reported on their allocation sequence generation and 9.1% had reported allocation sequence concealment. 72.8% of the studies had reported adequate information about the blinding of the participants and the personnel, in 36.4% of which the blinding was done (low risk) and in the other 36.4% it was not done (high risk). Moreover, blinding of the outcome assessment was reported in 36.4% of the studies. Attrition was also discussed in 72.7% of the studies, all of which had a low risk of incomplete outcome data. Selective reporting was assessed in all studies based on the match between the predetermined outcomes and the reported outcomes, and 100% of the studies had a low selective reporting bias. Other sources of bias were reported in 45.5% of the included studies (high risk) (Supplementary Material 1).  Abbreviations: ACE, angiotensin-converting enzyme; ACEi, angiotensin-converting enzyme inhibitors; ADEs, adverse drug events; AKI, acute kidney injury; ARB, angiotensin receptor blocker; CDSS, clinical decision support system; CKD, chronic kidney disease; Clcr, creatinine clearance; DRAP, drug renal alert pharmacy; EBPG, European best practice guideline; eCDSS, electronic clinical decision support system; eGFR, estimated glomerular filtration rate; EMD, electronic medication dispenser; ESAs, erythropoiesis-stimulating agents; Hb, hemoglobin; HRC, hypochromic red cell; KDIGO, kidney disease: improving global outcomes; ORAMA, optimal renal anemia management assessment; pADEs, potential adverse drug events; PCPs, primary care physicians; RCT, randomized controlled trial; sCr, serum creatinine; TSAT, transferrin saturation. * In both patient groups.

Interventions description
According to Table 3, CDSSs were respectively integrated with EHR and CPOE in nine (81.8%) and eight (72.7%) of the included studies, and real-time feedback and recommended courses of action were proposed in ten (90.9%) of the interventions. Clinical training was reported in six (54.5%) of the included studies, and clinical staff entered data specifically for intervention in only one (9.1%) study. A total of ten (90.9%) of the included studies performed pilot testing or used an iterative process of development. In four (36.4%) of the included studies, information was presented about other implementation components such as the cost and time required and the use of the frameworks. A total of nine (81.8%) of the interventions had presented information on appropriateness or guidelines specifically tailored to the individual patient, often as a pop-up or alert. Some of these interventions also recommended alternative interventions (B category). Table 4 shows the effects of the interventions on the outcomes in the included studies, which comprised 10 clinical outcomes and 22 process-of-care outcomes. The intervention effects on 62.5% (n = 20) of the measured outcomes were not statistically significant, on 31.25% (n = 10) were statistically significant positive, and on 6.25% (n = 2) were mixed (both positive and without an effect). The effects of CDSS-based interventions on clinical outcomes were not statistically significant in all the studies. As the most frequently reported clinical outcomes, rates or results of laboratory tests, mortality, and blood pressure changes were respectively assessed in four [41,44,47,50], two [47,49] and two [44,51] studies.

Discussion
This systematic review of eleven RCTs characterized CDSSs and investigated the effects of these systems on outcomes in patients with kidney disease. The outcomes were generally classified as clinical and process-of-care outcomes. The present results revealed that CDSSs had statistically significant positive effects on 45.5% of the process-of-care outcomes, while they did not have any statistically significant effect on the clinical outcomes. The characteristics of CDSSs investigated in the included studies comprised automatic data entry, real-time feedback, providing recommendations, and CDSS integration with CPOE.
The effects of CDSS on at least one clinical outcome investigated in half of the included studies were found to be insignificant, which is consistent with the results of other systematic reviews on the effects of CDSSs on clinical outcomes in different diseases, including asthma and cardiovascular diseases [52][53][54].
This finding can be attributed to the small sample size of the reviewed studies [9,10], the short duration of the interventions to reveal clinically important effects [9], the relative difficulties of implementing RCTs in real clinical settings, and the difficult measurement of direct clinical effects of CDSSs [54].
At least one process-of-care outcome has been measured in all the included studies. Although processes of care cannot directly affect the clinical status, they can mediate improvements in clinical outcomes [55].
The most frequently measured process-of-care outcomes in the included studies were medication-related (54.5%) and mainly included medication usage, medication appropriateness, medication errors, and drug discontinuation. The most significant and positive effect of CDSSs on these outcomes has been exerted on medication appropriateness (18.2%) as a key outcome in treating patients with kidney disease. CDSSs can improve prescriptions by providing drug allergy and drug-drug interaction alerts and the information required for standardizing clinical practice as well as updating pharmacotherapeutic information and dose adjustment calculations based on patient status [56]. Research suggests the acceptable effects of CDSSs on medication-related outcomes in patients with kidney disease given the significant healthcare improvements these systems cause in the patients. In line with the present results, the positive effects of CDSSs on medication-related outcomes in patients with chronic diseases are reported in other systematic reviews [31,57,58].
In caring for patients with kidney diseases, some of the features of CDSSs can be mentioned, such as automatic data entry through EHR, real-time feedback, providing recommendations, and integration with CPOE. A systematic review of  Several theory-based strategies were utilized to encourage provider uptake of the eCDSS and participant uptake of provider and/or pharmacist recommendations which classified using COM-B framework**** * Intervention Classification: 'A' interventions provided information only; 'B' interventions presented information on appropriateness or guidelines specifically tailored to the individual patient, often as a pop-up or alert. Some of these interventions also recommended alternative interventions, but did not include any barrier for the clinician to order the test; 'C' interventions in general were similar to 'B' interventions, but required the ordering clinician to justify with free text why they were overriding the decision support recommendation that a study was inappropriate (ie, a 'soft stop'). 'D' interventions included a 'hard stop,' meaning the intervention prevented the clinician from ordering a test contrary to the CDS determination of inappropriateness, until additional discussion with or permission obtained from another clinician or radiologist.
seventy RCTs [59]) found the factors that contribute to the successful implementation of CDSSs and improved performance of healthcare providers to include integration of automated decision support with the workflow of physicians, providing both evaluation results and recommendations, supporting decisions at the time and place of decision making, and using computerized decision support. The strengths of this study included pioneering the systematic review of the effects of CDSSs on all the outcomes associated with kidney diseases despite the diverse applications of CDSSs to kidney diseases and numerous studies in this context. As the other strength of this study, adopting a comprehensive search strategy reduced the drop-out rate. This systematic review also included RCTs as the only study design to ensure the quality of the results. The limitations of the present study comprised the exclusion of conference proceedings given the lack of access to their full texts; a number of studies might have been missed. Conducting a meta-analysis was also impossible given the diverse characteristics of CDSSs and outcomes.
In line with the other systematic reviews [60][61][62], this study found the majority of CDSSs to be implemented in hospitals rather than in primary care settings. This finding can be explained by the slower adoption of EHRs in primary healthcare settings. Given that patients with mild-to-moderate chronic kidney diseases are mostly taken care of in primary healthcare centers, it is recommended that CDSS be implemented and evaluated in these settings as well. The present study found CDSS to improve medication-related outcomes in patients with kidney disease, e.g. through positively affecting medication appropriateness. It is therefore recommended that clinicians use CDSS to improve medication-related outcomes in these patients.
Approximately two-thirds of the included studies were conducted in the US and the rest in Europe, which shows the priority given to using electronic systems in the health sector in developed countries. It is recommended that these systems be implemented in developing countries as well, and their effects be evaluated. Given the need for matching the intervention duration and sample size with the study objectives and outcomes, it is recommended that further RCTs be conducted with adequately long interventions and large samples. The number of studies on clinical outcomes is lower than that on process-of-care outcomes. Given that long-term outcomes have been rarely assessed in the literature, well-designed long-term prospective studies or well-designed retrospective cohort studies can provide more details on the effects of CDSSs on mortality and morbidity as clinical outcomes in patients with kidney disease.

Conclusion
Although CDSS may potentially be able to improve processes of care for patients with kidney disease, particularly with regard to medication appropriateness, no evidence was found that CDSS affects clinical outcomes in these patients. Further research is therefore needed to determine the effect of CDSSs on clinical outcomes in patients with kidney diseases.
This systematic review found that CDSSs improve process of care outcomes for patients with kidney disease, particularly in relation to medication-related outcomes, while having no statistically significant effect on clinical outcomes. CDSS features include automated data entry, real-time feedback, recommendation delivery, and CDSS integration with CPOE that contribute to the successful implementation of CDSSs and improved performance of healthcare providers in the care of patients with renal disease. According to the other interpretation of the results, conducting further RCTs with sufficiently long interventions and large samples in primary care settings, especially in developing countries, is required.

Funding
This paper was not funded.

Declaration of financial/other relationships
The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.