Chronic disease management interventions for people with chronic kidney disease in primary care: a systematic review and meta-analysis

Abstract

Background

Primary care providers manage the majority of patients with chronic kidney disease (CKD), although the most effective chronic disease management (CDM) strategies for these patients are unknown. We assessed the efficacy of CDM interventions used by primary care providers managing patients with CKD.

Methods

The Medline, Embase and Cochrane Central databases were systematically searched (inception to November 2014) for randomized controlled trials (RCTs) assessing education-based and computer-assisted CDM interventions targeting primary care providers managing patients with CKD in the community. The efficacy of CDM interventions was assessed using quality indicators [use of angiotensin-converting enzyme inhibitor (ACEI) or angiotensin receptor blocker (ARB), proteinuria measurement and achievement of blood pressure (BP) targets] and clinical outcomes (change in BP and glomerular filtration rate). Two independent reviewers evaluated studies for inclusion, quality and extracted data. Random effects models were used to estimate pooled odds ratios (ORs) and weighted mean differences for outcomes of interest.

Results

Five studies (188 clinics; 494 physicians; 42 852 patients with CKD) were included. Two studies compared computer-assisted intervention strategies with usual care, two studies compared education-based intervention strategies with computer-assisted intervention strategies and one study compared both these intervention strategies with usual care.

Compared with usual care, computer-assisted CDM interventions did not increase the likelihood of ACEI/ARB use among patients with CKD {pooled OR 1.00 [95% confidence interval (CI) 0.83–1.21]; I² = 0.0%}. Similarly, education-related CDM interventions did not increase the likelihood of ACEI/ARB use compared with computer-assisted CDM interventions [pooled OR 1.12 (95% CI 0.77–1.64); I² = 0.0%]. Inconsistencies in reporting methods limited further pooling of data.

Conclusions

To date, there have been very few randomized trials testing CDM interventions targeting primary care providers with the goal of improving care of people with CKD. Those conducted to date have shown minimal impact, suggesting that other strategies, or multifaceted interventions, may be required to enhance care for patients with CKD in the community.

INTRODUCTION

Chronic kidney disease [CKD; defined as a glomerular filtration rate (GFR) <60 mL/min/1.73 m²] is a significant public health problem [1] affecting an estimated 2.9 million Canadian adults [2] and increases the risk of cardiovascular disease and all-cause mortality [3]. Clinical practice guidelines recommend early recognition of CKD, modification of risk factors associated with CKD progression (e.g. hypertension, diabetes and proteinuria) and management of CKD-related complications (e.g. cardiovascular disease, anemia and bone disease) [4–8]. However, translating guidelines into practice has been challenging [9, 10]. Data suggest that chronic disease management (CDM) strategies (defined as ongoing and proactive follow-up of patients) are effective in improving care for patients with diabetes and thus may prove helpful in addressing the evidence-to-practice gap in CKD care [11]. However, evidence to date has shown that CDM strategies are not equally effective for improving patient care among various chronic conditions [12]. Although current CKD practice guidelines recommend the development of disease management programs that optimize community management of CKD, they do not provide a consensus on the optimal strategy for CKD management [4]. As such, decision makers and key stakeholders have limited guidance for CKD management programs.

Several studies have attempted to determine how interventions targeting primary care providers affect the management of patients with CKD, but these found considerable variability in efficacy on patient outcomes and processes of care [13, 14]. Furthermore, a recent systematic review found that despite the emphasis on early management of CKD, a substantial evi- dence–practice gap persists regarding the most effective strategies for managing patients with CKD in the community [15].

We therefore conducted a systematic review and meta-analysis of randomized controlled trials (RCTs) to evaluate the efficacy of primary care physician-targeted CDM interventions meant to improve outcomes and processes of care for people with CKD compared with usual care.

MATERIALS AND METHODS

Data sources and searches

We performed a systematic review and meta-analysis according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses for the conduct of meta-analyses of RCTs [16]. Medline via OVID (1946–November 2014), Embase via OVID (1947–November 2014) and the Cochrane Central Registry of Controlled Trials (CENTRAL; no date restriction) were systematically searched (Supplementary data, Table S1) for relevant RCTs. Relevant text words and subject headings were used to capture computer-assisted and education-based interventions including electronic prompts, computer decision support systems, audit and feedback, academic detailing and guideline education. The search was not limited by language. Two reviewers (L.G. and C.J.) manually searched reference lists of all included articles, systematic reviews and meta-analyses for additional eligible articles. Finally, experts in the field were contacted to obtain information about additional ongoing or unpublished studies.

Study selection

We included cluster RCTs of adult patients with CKD managed by primary care providers in the community. We focused on cluster RCTs since health services interventions related to the assessment of delivery of care modes such as computer-assisted and education-based decision support systems are often tested at the level of the practice rather than the individual patient [17]. We limited the CDM interventions to physician-targeted interventions only (computer-assisted and education-based), as we were interested in the impact of interventions targeting primary care providers on patient outcomes. Usual care (defined as the typical care received consisting of assessments and treatments considered necessary by the family doctor) was the comparator of interest. Studies were restricted to trials that measured quality indicators or clinical or process of care outcomes. Studies that included patients with end-stage renal disease (those treated with renal replacement therapy or kidney transplant) were excluded. A third reviewer (M.J.) was consulted in cases where discrepancies existed between reviewers.

Data extraction and quality assessment

Information from articles selected for inclusion in the systematic review and meta-analysis was extracted into an Excel spreadsheet (Microsoft, Redmond, WA, USA) independently by two reviewers (L.G. and C.J.). Data abstracted included the study setting (country, primary care or community setting, cluster design), intervention details (computer-assisted or education-based, description, duration), comparator(s), participant demographics (number of clinics, physicians and patients; mean patient age; sex proportions; comorbidities) and outcome events as described below. Risk of bias of all eligible articles was independently assessed using the Cochrane Collaboration Tool for Assessing Risk of Bias [18]. Each study was assigned a risk of bias score from low to high based on sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting and other potential threats to validity as outlined in the tool.

Outcomes

Data were collected on (i) the proportion of patients using angiotensin-converting enzyme inhibitors (ACEIs) or angiotensin receptor blockers (ARBs), (ii) the proportion of patients who reached a prespecified blood pressure (BP) target, the mean systolic blood pressure (SBP), the proportion of patients who had a measurement of proteinuria and the mean estimated glomerular filtration rate (eGFR) at the end of the study.

Data synthesis and analysis

For binary outcomes, individual study outcomes were expressed as odds ratios (ORs) with 95% confidence intervals (CIs) calculated from the event numbers extracted from each trial. In the case where event numbers were not provided, the reported OR was extracted. ORs were calculated with the proportion of events as the numerator and proportion of nonevents as the denominator. Continuous outcomes were expressed as weighted mean differences calculated using end-of-trial mean values, standard deviations and treatment arm size. To account for relative variability within and between randomized clusters, appropriate adjustments were made using the intraclass correlation coefficient [19] prior to pooling. A random effects model was used to calculate summary estimate ORs or weighted mean differences of the outcomes of interest. Heterogeneity was assessed with the I² statistic, expressed as the percentage of variability across studies attributable to heterogeneity beyond chance. Publication bias was assessed using the Begg's test with graphical representation using funnel plots of the natural logarithm of the OR against its standard error. A two-sided P-value <0.05 was considered statistically significant. All analyses were performed using STATA, version 13 (StataCorp, College Station, TX, USA)[20].

RESULTS

Search results and characteristics of included studies

The literature search yielded 1405 articles, which were screened by reviewers with 98.4% agreement (κ = 0.51) to identify those for full-text review (Figure 1). Of the 15 studies reviewed in full, 5 RCTs were eligible for inclusion [14, 21–24]. Four studies [14, 21, 22, 24] provided data appropriate for meta-analysis while one study [23] was described qualitatively due to differences in reporting methods. A total of 23 unique patient and processes of care outcomes were reported in the four trials included in the meta-analysis. However, only five outcomes (ACEI/ARB use, achieving target BP, proteinuria measurement, mean SBP and mean eGFR) were reported in two or more studies.

FIGURE 1.

Flow diagram of studies that were considered for inclusion.

We observed substantial variability in the components of the CDM interventions assessed across the included studies. Of the four studies eligible for inclusion in the meta-analysis, one study compared the use of two 15-min education sessions combined with real-time automated electronic medical record (EMR) alerts for patients with eGFR <45 mL/min/1.73 m² (computerized prompt intervention) with the use of only two 15-min education sessions (education-based intervention) [24]. The second study compared an education intervention consisting of a lecture on CKD guidelines with a computer-assisted intervention (access to a web-based CKD registry combined with a lecture of CKD guidelines) [22]. A third study compared an enhanced management-based laboratory eGFR prompt with usual care (standard eGFR laboratory prompt) [14]. The final study compared an education-based intervention consisting of audit-based education, including feedback and training at data quality workshops, printed aids and target patient lists compared with usual care. In addition, this study compared a computer-assisted intervention (academic detailing, printed information including CKD guidelines, and access to an information website) with usual care [21].

The study characteristics, interventions and outcomes of the five studies included in the qualitative synthesis are summarized in Table 1. Overall, study size ranged from 94 to 504 207 patients and between 30 and 354 physicians from up to 93 clinic settings. Mean patient age ranged from 62 to 78 years. Males represented the minority in four of the five studies (range 34–45%) [14, 21, 23, 24] and one study reported 95% of patients being male [22]. Studies were conducted in the USA, Canada, the UK and Mexico and were published between 2008 and 2013. All studies identified were published in English.

Table 1.

Characteristics of included studies

Author (year)	Country of origin	Inclusion criteria	Intervention (category)	Comparator	Unit of randomization	Total no. of patients (no. of clusters)	Mean patient age (years)	Patient gender (% male)	Outcomes reported	Timeline
Cortés-Sanabria et al. (2008)	Mexico	Primary health care units, patients with type 2 diabetes and CKD	6-months education based on theory-practice model	Usual care	Clinic	94 (2)	62.0	43.5	Clinical competence of physicians; BP; BMI; smoking cessation; alcohol cessation; glucose; cholesterol; albuminuria; eGFR; use of antihypertensives, antidiabetics, statins, NSAID use	6-months intervention; outcomes assessed at enrollment, 6- and 12-month time points
Abdel-Kader et al. (2011)	USA	CKD patients (eGFR <45 mL/min/1.73 m2) in the 12 months prior to their visit and had never been evaluated by a nephrologist	Two 15-min education sessions (education related) + real-time automated EMR alerts (EMR related) for patients with eGFR <45 mL/min/1.73 m2	Two 15-min education sessions (education related)	Physician practice	248 (30)	65.3	37.7	EMR order for nephrology consultation; Albuminuria or proteinuria; ACEI/ARB, NSAID use; documentation of CKD; achievement of target BP; BP; eGFR; Hb; bicarbonate; calcium; phosphorus; PTH	12-month intervention; outcomes assessed 1 year before and 1 year after (exceptions ACEI/ ARB assessed at onset and after
Drawz et al. (2012)	USA	Primary care clinics, CKD patients (eGFR <60 mL/min/1.73 m2 based on two readings between 90–730 days previous	Access to web-based CKD registry (EMR related) + lecture on CKD guidelines (education related)	Lecture on CKD guidelines (education related)	Patients	781 (N/A)	71.0	95.2	PTH measurement; achievement of target BP; phosphorous; proteinuria; Hb measurement; use of ACEI/ARB, diuretica	12-month intervention; outcomes assessed 1 year before and 1 year after
Manns et al. (2012)	Canada	Primary care practices, elderly (>66 years old) CKD patients defined by eGFR <60 mL/min/ 1.73 m2 with diabetes or proteinuria	Enhanced eGFR laboratory prompt (EMR related)	Standard eGFR laboratory prompt (usual care)	Clinic	5444 (90)	78.1	44.8	ACEI/ARB prescription; cholesterol lowering medication; new class antihypertensive medication; nephrologist consultation; albuminuria measurement; Lipid measurement; Hb A1C measurement	12-month intervention; outcomes assessed within 1 year of first prompt
de Lusignana et al. (2013)	UK	Primary care clinics, CKD patients (eGFR <60/mL/min/1.73 m2) based on two readings at least 90 days apart	Audit-based education (education related) involved feedback and training at data quality workshops, printed aids, target patient lists. Guidelines and prompts (EMR related) involved academic detailing, printed information including CKD guidelines, and access to an information website	Usual care	Clinic	504 207 (93)	75.0	33.9	Reduction in SBP over time; incident cases of cardiovascular disease; eGFR	2-year intervention; outcomes assessed between earliest and latest measurements

Outcomes reported in studies included continuous variables, expressed as means, and categorical variables, expressed as numbers or proportions.

KDOQI, Kidney Disease Outcomes Quality Initiative; N/A, not applicable; BP, blood pressure; Hb, hemoglobin; TG, triglyceride; SCr, serum creatinine; SBP, systolic blood pressure; PTH, parathyroid hormone; ACEI/ARB, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker; EMR, electronic medical record; NSAID, nonsteroidal anti-inflammatory drug; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate.

^aOutcomes reported as ORs.

Study quality of included studies was assessed and is presented in Table 2. Allocation concealment is of concern in cluster randomized trials, as all clusters are typically randomized at once [19]; however, this was included as a key indicator within the Cochrane Collaboration tool for assessing risk of bias. One study [14] did not contain any components with a high risk of bias. There was one study each with a high risk of bias in one, two or three of the key indicators. De Lusignana et al. [21] was assessed with the highest risk of bias in three indicators: incomplete outcome data, selective outcome reporting and other sources of bias. Cortes-Sanabria et al. [23] was rated with low or uncertain risk of bias in all components but was excluded from meta-analysis and Table 2 as previously mentioned.

Table 2.

Risk of bias assessment of included trials using the Cochrane Collaboration Tool for Assessing Risk of Bias

	Sequence generation	Allocation concealment	Blinding	Incomplete outcome data	Selective outcome reporting	Other potential biases
Abdel-Kadar et al. [24]	Low	?	High	Low	Low	?
Drawz et al. [22]	High	High	Low	Low	Low	?
Manns et al. [14]	Low	Low	?	Low	Low	Low
De Lusignana et al. [21]	Low	?	Low	High	High	High

Patient clinical outcomes and processes of care

ACEI/ARB use

Two studies provided sufficient information to compare the effects of computer-assisted interventions with usual care on the proportion of patients using ACEI/ARBs [14, 21]. The proportion of patients using ACEI/ARBs did not differ between computer-assisted interventions and usual care [pooled OR 1.00 (95% CI 0.83–1.21)] (Figure 2) (I² = 0.0%, P-value = 0.60). Three studies [21, 22, 24] provided data sufficient to compare ACEI/ARB use between education-based and computer-assisted interventions; similarly, the proportion of patients using ACEI/ARBs did not differ [pooled OR 1.12 (95% CI 0.77–1.64)] (Figure 2) with no evidence of heterogeneity in the magnitude of effect across the included studies (I² = 0.0%, P = 0.87).

FIGURE 2.

Forest plot of studies reporting the odds of ACEI/ARB use in CKD patients between computer-assisted CDM interventions and usual care and education-based and computer-assisted CDM interventions using random effects analysis.

BP target

Three studies [21, 22, 24] reported the proportion of patients achieving a prespecified BP target (130/80 or 140/80 mmHg). The proportion was similar for education-based compared with computer-assisted interventions [pooled OR 1.11 (95% CI 0.90–1.37)] (Figure 3), with no evidence of heterogeneity across studies included (I² = 0.0%, P = 0.86).

FIGURE 3.

Forest plot of studies comparing education-based to computer-assisted CDM interventions on the odds of CKD patients reaching a specified BP target and having a proteinuria assessment using random effects analysis.

Proteinuria assessment

Two [22, 24] studies provided sufficient information to compare the effects of education-based interventions with computer-assisted interventions on the proportion of patients having a proteinuria measurement (binary measure). The proportion of patients with a proteinuria assessment did not differ [pooled OR 0.87 (95% CI 0.41–1.84)] (Figure 3), with moderate heterogeneity across the included studies (I² = 63.7%, P = 0.09).

Mean change in SBP

The mean difference in SBP postintervention between the education-based and computer-assisted interventions was included as an outcome in two [21, 24] studies identified. The mean difference in SBP did not differ [weighted mean difference (WMD) −0.59 mmHg (95% CI −2.80–1.61)] across the interventions, with no evidence of heterogeneity across the included studies (I² = 0.0%, P = 0.76) (Figure 4).

FIGURE 4.

Forest plot of studies comparing education-based to computer-assisted CDM interventions on the mean weighted difference of SBP and eGFR in CKD patients using random effects analysis.

Mean change in eGFR

Two studies [21, 24] reported the mean change in eGFR for patients within the education-based interventions compared with computer-assisted interventions. There was no difference in the mean eGFR between the intervention types [WMD −0.32 mL/min/ 1.73 m² (95% CI −2.37–1.73)], with no evidence of heterogeneity (I² = 0.0%, P = 0.89) (Figure 4).

Publication bias

Publication bias could not be assessed due to inconsistency of data reporting in the included studies.

DISCUSSION

Our systematic review assessing CDM interventions targeting primary care providers who care for CKD patients in the community identified a critical lack of studies, with only five relevant RCTs and with only four eligible for inclusion in the meta-analysis. When compared with usual care, computer-assisted interventions had no effect on ACEI/ARB use among CKD patients. A head-to-head comparison of education-based versus computer-assisted CDM interventions also found no effect on any of the patient outcomes or processes of care. However, these findings are limited by the considerable lack of evidence for all CDM intervention types targeting primary care providers managing patients with CKD.

The Kidney Disease: Improving Global Outcomes (KDIGO) clinical practice guidelines for evaluation and management of CKD patients were developed to standardize assessment and care for these patients. Unfortunately, evidence suggests that many physicians are unfamiliar with the CKD guidelines, resulting in a significant barrier to uptake in practice [25–27]. Moreover, dissemination and implementation of guidelines into practice alone is insufficient to overcome the challenges of daily management of CKD [9], as patients with CKD often have numerous comorbid conditions, making their comprehensive management exceptionally challenging [10]. CDM strategies may provide a solution to this evidence-practice gap by designing an intervention to improve patient care. CDM interventions can be categorized into 11 unique strategies [28] (audit and feedback, case management, team changes, electronic patient registry, clinician education, clinician reminders, facilitated relay of clinical information to clinicians, patient education, promotion of self-management, patient reminder systems and continuous quality improvement), with each strategy varying in resource intensity [12]. Compared with usual care, CDM interventions aim to streamline these assessments and treatments. However, CDM interventions can be arduous and the different strategies require careful consideration of barriers to optimal care, differential effectiveness, resource intensity and ease of implementation prior to choosing a CDM strategy [12]. These challenges may contribute to the lack of effect of the interventions reported in our meta-analysis.

There are a number of potential reasons why these CDM interventions did not show an effect. Relevant CDM interventions are required to target patients in earlier stages of CKD; however, this may be challenging, as less than 1/10 of individuals with moderately decreased kidney function (stage 3 CKD) report awareness of their CKD [29]. Additionally, in the case of computer-assisted interventions (alerts and prompts), primary care providers may have been overwhelmed, or experienced alert fatigue, by the number of patients receiving a prompt, as identified by two of the studies in our review [14, 24]. Importantly, three features have been shown to be associated with effective clinical decision support: routine guidance as part of clinician workflow, providing recommendations rather than assessments, and provision of guidance at the time and location of decision making [30]. Only two [14, 24] of the four included studies appear to have considered these features in the development and execution of their CDM interventions. We hypothesize that the systematic inclusion of these three features would improve the efficacy of CDM interventions for primary care providers managing CKD patients.

Although this review found no effect of CDM interventions on patient outcomes and processes of care for CKD patients, several systematic reviews of CDM interventions in other settings have demonstrated efficacy, including in diabetic patients, where CDM interventions improved glycemic control [11, 28, 31]. Many of the trials included in the review of people with diabetes used multidisciplinary interventions, where different disease management strategies were combined. Regardless of CDM type, all three prior systematic reviews noted that the effectiveness of the CDM strategies was largely determined by two key components: the ability for case managers to adjust treatment and medications autonomously (without prior physician approval), as well as regular and high frequency of patient contact [11, 28, 31]. Given that diabetes and CKD are both chronic conditions and patient characteristics are very similar across the two patient groups, components of successful CDM strategies in diabetes may be adaptable to CKD disease management. The complexity of care for CKD patients, similar to diabetes patients, may require a standardized multidisciplinary approach [12]. Indeed, observational studies assessing the effectiveness of CDM interventions have shown promising results, suggesting favorable effects with CDM interventions on improving outcomes among patients with CKD [32–34], congestive heart failure [35, 36], chronic obstructive pulmonary disease [37] and other chronic diseases [38, 39]. The development of standardized definitions for CDM interventions and standardized reporting of universally clinically relevant patient outcomes and processes of care would significantly improve the relevance and impact of future knowledge in this field.

The results of our systematic review should be interpreted in light of the study limitations. We are primarily limited by the paucity of evidence in the field to date. Additionally, this review found considerable heterogeneity in the definitions of CDM intervention within the computer-assisted and education-based CDM categories, which was compounded by the inconsistencies in CKD definitions and outcomes reporting across the five trials. It should be noted that while we have reported on the effects of CDM interventions on ACEI/ARB use, the most frequently reported study outcome across the included studies, this outcome may have limited relevance in certain patient groups, including patients without proteinuria, or more generally in the elderly, in whom ACEI/ARB use is not indicated. Study outcomes with broader utility across the wider CKD population are needed.

In summary, the limited evidence to date suggests that computer-assisted CDM interventions targeting primary care providers managing patients with CKD in the community have no effect on patient outcomes and processes of care when compared with usual care. A similar result was found in head-to-head comparisons with education-based CDM interventions. Future research is required to further inform this question due to the current lack of evidence and considerable heterogeneity in the definition of CDM interventions included in this review.

AUTHORS' CONTRIBUTIONS

All authors contributed to critical revision of this article. L.G., C.J., B.H. and M.J. contributed to study design and data interpretation. L.G. and C.J. were responsible for data collection, analysis and manuscript preparation. L.G. and C.J. are graduate students, M.D. is a senior research associate, M.J. is a postdoctoral fellow and B.H. and B.M. are clinician scientists.

SUPPLEMENTARY DATA

Supplementary data are available online at http://ndt.oxfordjournals.org.