Effectiveness of Quality Improvement Strategies for the Management of CKD

Abstract

Background and objectives

Quality improvement interventions have enhanced care for other chronic illnesses, but their effectiveness for patients with CKD is unknown. We sought to determine the effects of quality improvement strategies on clinical outcomes in adult patients with nondialysis-requiring CKD.

Design, setting, participants, & measurements

We conducted a systematic review of randomized trials, searching Medline and the Cochrane Effective Practice and Organization of Care database from January of 2003 to April of 2015. Eligible studies evaluated one or more of 11 prespecified quality improvement strategies, and prespecified study outcomes included at least one process of care measure, surrogate outcome, or hard clinical outcome. We used a random effects model to estimate the pooled risk ratio (RR; dichotomous data) or the mean difference (continuous data).

Results

We reviewed 15 patient-level randomized trials (n=3298 patients), and six cluster-randomized trials (n=30,042 patients). Quality improvement strategies reduced dialysis incidence (seven trials; RR, 0.85; 95% confidence interval [95% CI], 0.74 to 0.97) and LDL cholesterol concentrations (four trials; mean difference, −17.6 mg/dl; 95% CI, −28.7 to −6.5), and increased the likelihood that patients received renin-angiotensin-aldosterone system inhibitors (nine trials; RR, 1.16; 95% CI, 1.06 to 1.27). We did not observe statistically significant effects on mortality, cardiovascular events, eGFR, glycated hemoglobin, and systolic or diastolic BP.

Conclusions

Quality improvement interventions yielded significant beneficial effects on three elements of CKD care. Estimates of the effectiveness of quality improvement strategies were limited by study number and adherence to quality improvement principles.

Podcast

This article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2017_09_06_CJASNPodcast_17_10.mp3

Introduction

CKD is a common condition that affects approximately 10% of adults in North America (1). Patients with CKD require multifaceted care to target risk factors for disease progression (including hypertension and proteinuria) and complications (including cardiovascular and bone disease). Despite evidence and clinical practice guidelines recommending a number of preventative and therapeutic strategies (2), many patients with CKD do not receive optimal care (3,4). For example, renin-angiotensin-aldosterone system blockade in patients with diabetic kidney disease is not utilized in a substantial proportion of patients (5). This quality of care gap has been identified among primary care providers and nephrologists (4,5).

Quality improvement (QI) interventions, whose objective is the continual strengthening of health system performance, may help overcome some of these care gaps in the CKD population. QI strategies can target the health care system, health care provider, and/or patient, and have improved care processes and outcomes for several chronic illnesses (6,7). For example, a meta-analysis of 94 randomized controlled trials in patients with diabetes mellitus found that QI strategies were associated with reductions in glycated hemoglobin (HbA1c), BP, and LDL cholesterol (8).

Although clinicians and policymakers devote substantial time and effort to identifying optimal patterns of care for patients with CKD, it remains unclear which QI interventions are effective in the CKD population. To address this knowledge gap, we conducted a systematic review and meta-analysis to synthesize the available literature on QI strategies for patients with CKD, and to describe the effect of QI interventions on clinical outcomes.

Materials and Methods

We reported this systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines (9), with the exception of prospective protocol publication and registration. We fully support protocol publication and registration in our future research to increase transparency and decrease unplanned study duplication.

Data Sources and Searches

We used a strategy developed with a health informatics specialist to search Medline (January 2003–April 2, 2015) and the Cochrane Effective Practice and Organization of Care (EPOC) database (January of 2003 to April of 2015). We reviewed the bibliographies of identified articles to locate additional eligible studies. We restricted our final search strategy to adult studies (≥19 years of age) and applied no language restrictions. We adjusted the strategy as necessary for searching the Cochrane EPOC database. The Medline search strategy is provided in Supplemental Figure 1.

Study Selection

The target population was adult patients with nondialysis-requiring CKD, which we defined using the Kidney Disease Improve Global Outcomes criteria (2). We excluded studies that focused exclusively on patients receiving dialysis or kidney transplant recipients (ESRD) because we anticipated these patient populations would respond to different QI interventions.

We included randomized clinical trials that used patient-level randomization or cluster- randomization. We excluded unpublished conference abstracts. Because of anticipated heterogeneity in QI approaches, we required QI interventions to assess one or more of 11 predefined QI strategies on the basis of a taxonomy adapted from the EPOC group that has been previously described (Supplemental Figure 2) (8,10). Eligible QI strategies targeted most components of health care, including the health care system (e.g., team changes), health care providers (e.g., reminders), and patients (e.g., self-management).

We required study outcomes to fit within a prespecified framework of clinically relevant outcomes that were felt to be of significance to health care providers and/or patients (Supplemental Table 1). Study outcomes had to include at least one process of care measure (e.g., proportion of patients taking renin-angiotensin-aldosterone system inhibitors), a relevant surrogate outcome (e.g., BP), or a clinical outcome (e.g., dialysis).

Data Abstraction

From each study, we collected data on study details (e.g., population, setting, randomization of individuals or clusters, type of QI intervention), the characteristics of participants (e.g., mean age, sex, CKD stage), the outcomes that were ascertained, and study results.

Two authors (S.A.S. and Z.H.) independently scanned titles and abstracts from the initial search for preliminary selection, blinded to author and institution. We reviewed selected full-text papers in more detail for determination of eligibility and classification of QI strategies using the aforementioned framework. We used the Cochrane Risk of Bias tool to assess the risk of bias in individual studies (11). We also contacted authors of trials with missing information for clarification before inclusion. We resolved all discrepancies by discussion or involvement of a third reviewer, which primarily involved distinguishing case management from team changes and risk of bias assessment for incomplete outcome data and selective outcome reporting.

Statistical Analyses

We qualitatively synthesized the results of all included studies, focusing on the patient population, study design, details of the QI intervention, and outcomes.

Because many of the cluster trials that were included analyzed results at the patient-level rather than the cluster-level (i.e., unit of analysis issue), we calculated an effective sample size for each cluster trial by use of the intracluster correlation coefficient (ICC) (12,13). For one trial that did not report an ICC (14), we imputed an ICC of 0.57 on the basis of the results of other included trials. These methods have been used by others when combining results from cluster-randomized and patient-randomized trials (8,15).

To ensure that we maintained study independence, we included a maximum of two groups in our analysis even if trials included more than two groups. This restriction applied to one trial, which compared audit and feedback versus clinician education versus usual care (16); in this case, we excluded the results for the clinician education group because it was not the authors’ primary QI intervention of interest.

We imputed unreported SDs and converted data presented as median/interquartile range to mean/SD, using established methods (15,17,18). We used a DerSimonian and Laird random effects model to estimate the pooled risk ratio (RR; dichotomous data) or the mean difference (MD; continuous data) across the included trials. We assessed the consistency of results across studies by use of forest plots, and statistical heterogeneity using the Cochrane Q test (significance set at 0.01) and I² values (19). Publication bias was assessed visually using a funnel plot. We performed all analyses using Review Manager version 5.3 (The Cochrane Collaboration) (20).

Results

Our search strategy yielded 12,554 unique citations. We excluded 12,504 citations on the basis of screening of title and/or abstract because of nonrelevant interventions, ineligible outcomes, or a lack of randomization. This left 50 papers for full-text review. We subsequently excluded 31 studies that did not fulfill our inclusion criteria because of exclusive recruitment of patients with ESRD (n=9), the absence of a CKD subgroup in studies focusing on diabetes mellitus or hypertension (n=7; 312 total patients with CKD, with two studies not reporting the number of patients with CKD), outcome reporting that did not satisfy our prespecified criteria (n=5; primarily measured knowledge uptake and retention), lack of randomization (n=3), nonprimary publication (n=3; reviews or protocols), duplicate or companion publications (n=3), and interventions deemed not to be QI strategies (n=1). We identified two additional studies through bibliography review. This process yielded 21 studies for further synthesis and analysis (Figure 1), which included 15 patient-level randomized trials (3298 patients) (21–36) and six cluster-randomized trials (30,042 patients) (14,16,37–40).

Figure 1.

PRISMA flow diagram of included studies.

Study Characteristics

Most trials focused on patients with stage 3–5 CKD, tested two QI strategies per study, followed patients for 12 months, and evaluated effects on BP, eGFR, and/or HbA1c (Tables 1 and 2). Proteinuria was an outcome in five out of 21 (24%) included studies. Primary outcomes were prespecified for 20 out of 21 (95%) studies, with the most common being BP (five out of 21, 24%), kidney function (four out of 21, 19%), and death (two out of 21, 10%). Power calculations were provided for 16 out of 21 (76%) studies, and 17 out of 21 (81%) were analyzed using the intention-to-treat principle. Outcomes that were infrequently studied included all-cause hospitalization (n=2) (25,26), anemia management (n=1) (21), dialysis modality selection (n=1) (32), transplantation decisions (n=1) (23), and vascular access creation (n=0).

Table 1.

Characteristics of included studies and patients

Characteristics	Patient Trials (N=15; 3298 Patients)	Cluster Trials (N=6; 30,042 Patients)
Median no. of patients (25th–75th percentile)	130 (68–316)	515 (185–4278)
Median no. of clusters (25th–75th percentile)	Not applicable	20 (4–77)
Median age in yr (25th–75th percentile)	60 (57–67)	72 (67–75)
Men, n (%)	1908 (58)	11,290 (38)
CKD stage, n (%)a
Stage 1	4 (27)	1 (17)
Stage 2	4 (27)	1 (17)
Stage 3	10 (67)	6 (100)
Stage 4	11 (73)	4 (67)
Stage 5	9 (60)	2 (33)
Country, n (%)
Australia	2 (13)	0
Canada	5 (33)	1 (17)
China	1 (7)	0
Denmark	1 (7)	0
Mexico	0	1 (17)
Netherlands	1 (7)	1 (17)
New Zealand	1 (7)	0
Taiwan	1 (7)	0
United Kingdom	2 (13)	1 (17)
United States	1 (7)	2 (33)
Longest duration of follow-up in mo (25th–75th percentile)	12 (6–24)	12 (12–21)
Median number of QI strategies per study (25th–75th percentile)	2 (2–3)	2 (2–2)
Types of QI strategies tested, n (%)a
Case management	7 (47)	1 (17)
Team changes	4 (27)	1 (17)
Electronic patient registries	0	1 (17)
Facilitated relay of information to clinicians	1 (7)	0
Continuous quality improvement	0	0
Audit and feedback	0	1 (17)
Clinician education	0	5 (83)
Clinician reminders	1 (7)	2 (33)
Patient self-management	10 (67)	0
Patient education	12 (80)	0
Patient reminders	4 (27)	0
Administrators of QI interventions, n (%)a
Primary care physician	0	6 (100)
Nephrologist/specialist	3 (20)	1 (17)
Nurse	9 (60)	1 (17)
Pharmacist	0	0
Dietician	5 (33)	0
Psychologist	0	0
Social worker	2 (13)	0
Otherb	4 (27)	0
Study outcomes, n (%)a
Mortality	8 (53)	1 (17)
Dialysis	8 (53)	0
Cardiovascular events	5 (33)	2 (33)
Systolic BP	5 (33)	4 (67)
Diastolic BP	5 (33)	3 (50)
Use of ACE inhibitor/ARB	4 (27)	6 (100)
Use of aspirin	3 (20)	1 (17)
Use of statin	3 (20)	3 (50)
Use of vitamin D	2 (13)	0
Use of phosphate binders	3 (20)	0
eGFR	9 (60)	4 (67)
Proteinuria	3 (20)	2 (33)
HbA1c	7 (47)	1 (17)
LDL cholesterol	4 (27)	1 (17)
Smoking cessation	4 (27)	2 (33)
Body mass index	3 (20)	2 (33)
Assessment for proteinuria	0	3 (50)
Referral to a nephrologist	0	3 (50)

QI, quality improvement; ACE, angiotensin-converting enzyme; ARB, angiotensin receptor blocker; HbA1c, glycated hemoglobin.

^aRefers to the number of trials that included these characteristics.

^bOther category includes health educators, health care assistants, lay health workers, and patients’ peers.

Table 2.

Details of included randomized trials

Study/Country/Age, yr (Men, %)	Setting	Stage of CKD	No. of Patients (No. after Adjusting for Clustering)/Duration of Follow-up, mo	QI Intervention	Primary Outcome	Secondary Outcomes
						BP/ACE and ARB Use	Kidney Function	Other
Patent-Randomized Trials
Barrett et al. (21)/Canada/C=67 (44); QI=67 (45)	Primary care in an urban center without previous nephrology contact	3–4	474/24	Case management	None (pilot study)	Yesa/Yes	eGFRb, Dialysis	Mortality
				Team changes				Cardiovascular events
				Self-management				Use of ASA, statins, vitamin D, phosphate binders
								Laboratory results (HbA1c, LDL, Hb, iron indices, Ca, PO4, PTH, bicarbonate)a
								Smoking cessation
Blakeman et al. (22)/United Kingdom/C=72 (41); QI=72 (42)	Primary care in underserviced communities	3	436/6	Self-management	BP and quality of life	Yesa/No	No	Active engagement in life
				Patient education				Outpatient PCP use
Boulware et al. (23)/United States/C=60 (41); QI=59 (40)	Nephrology practices (academic and community in an urban city)	3–5	130/6	Case management	Behaviors related to living donor kidney transplantation	No/No	No	None
				Patient education
Campbell et al. (24)/Australia/C=71 (63); QI=70 (49)	Predialysis clinic (single center)	4–5	56/3	Case management	Body cell mass	No/No	eGFR	Protein/energy intake
				Self-management				Weight
				Patient reminders				Serum albumin
								C-reactive protein
Chan et al. (25) China/C=65 (67); QI=65 (66)	Diabetes clinics from public hospitals in Hong Kong	3–4	205/24	Case management	Death, dialysis, or serum Cr ≥5.65 mg/dl	Yes/Yes	eGFR, Dialysis	Cardiovascular events
				Team changes				Hospitalizations
				Self-management				Laboratory results (HbA1c, lipids)
				Patient reminders
Chen et al. (26)/Taiwan/C=69 (56); QI=68 (56)	Predialysis clinic (single center)	3–5	54/12	Case management	eGFR and number of hospitalizations	No/No	eGFR, Dialysis	Mortality
				Self-management				CKD knowledge
				Patient education
				Patient reminders
Degen (27)/Canada/C=53 (77); QI=56 (27)	Predialysis clinic (two academic hospitals)	4–5 with PO4>4.5 mg/dl	24/3	Patient education	Serum phosphate	No/No	eGFR	Use of phosphate binders
				Patient reminders				Body mass index
								Body weight
								Protein/energy intake
								Laboratory results (Ca, albumin)
Devins et al. (28)/Canada/C=57 (69); QI=60 (52)	Predialysis clinics at a tertiary care center	5	297/18	Patient education	Dialysis	No/No	Dialysis	Mortality
				Self-management
Devins et al. (29)/Canada/C=53 (62); QI=51 (63)	Predialysis clinics in an urban center	5	335/240	Patient education	Dialysis	No/No	Dialysis	Mortalityc
				Self-management
Gaede et al. (30)/Denmark/C=55 (70); QI=55 (79)	Specialized diabetes center (single center)	1–2 with MA	160/96	Case management	Cardiovascular outcomes	Yes/Yes	eGFR, Dialysis Proteinuria	Mortality
				Team changes				Cardiovascular events
				Patient education				Use of ASA, statins, BP medication
								Body mass index
								Protein/energy intake
								Laboratory results (HbA1c, lipids)
								Smoking cessation
Hotu et al. (31)/New Zealand/C=60 (53); QI=63 (55)	Primary care and specialty clinics in urban center (Maori and Pacific patients)	3–4 with diabetes and HTN	65/12	Case management	BP	Yes/No	eGFR, Dialysis Proteinuria	Mortality
				Facilitated relay of information				Cardiovascular events
				Patient education				BP medication
				Patient reminders				Laboratory results (HbA1c, total cholesterol)
Manns et al. (32)/Canada/C=35 (64); QI=35 (65)	Predialysis clinic (single center)	3–4	70/12	Self-management	Intention to initiate self-care dialysis	No/No	No	Modality selection
				Patient education
MASTERPLAN: Peeters et al. (33); van Zuilen et al. (34)/The Netherlands C=59 (68); QI=59 (67)	Nine hospitals with a nephrology department	2–4	788/72	Team changes	Composite of death from myocardial infarction, stroke, or cardiovascular disease	Yes/Yes	eGFR, Dialysis	Mortality
				Patient education				Use of ASA, statin, vitamin D, phosphate binders, BP medication
							Proteinuria	Laboratory results (HbA1c, LDL, Hb, PO4, PTH)
								Body mass index
								Sodium intake
								Smoking cessation
Steed et al. (35)/United Kingdom/C=60 (75); QI=59 (68)	Two inner city hospital clinics	1–5 with diabetes and MA	124/3	Self-management	Quality of life, HbA1c	No/No	No	Smoking cessation
				Patient education
Williams et al. (36)/Australia/C=66 (56); QI=68 (56)	Outpatient diabetes and nephrology clinics (single center)	3–4	80/9	Self-management	Systolic BP	Yesb/No	eGFR	HbA1c
				Patient education				Medication adherence
Cluster-Randomized Trials
Abdel-Kader et al. (37)/United States/C=65 (41); QI=66 (35)	Primary care academic hospital, single center	3–4	248 (232)/12	Clinician education	Nephrology referral	Yes/Yes	eGFR	Proteinuria assessment
				Clinician reminders				CKD documentation
								Laboratory results (Hb, Ca, PO4, PTH, bicarbonate)
Cortés-Sanabria et al. (14)/Mexico/C=61 (39); QI=63 (48)	Two primary care units in Mexico	1–3	94 (67)/12	Clinician education	Physician competence and patient kidney function	Yesd/Yes	eGFR, Proteinuria	Use of ASA, statins
								Body mass index
								Total cholesterol
								Smoking cessation
Lusignan et al. (16)/United Kingdom/C=75 (33); QI=75 (34)	Primary care	3–5	23,311 (1457)/24	Audit and feedback	Systolic BP	Yes (systolic only)/Yes	eGFR	Mortality
				Clinician education				New cardiovascular disease
Drawz et al. (38)/United Stages/C=71 (96); QI=71 (95)	Primary care clinics (single center)	3–4	781 (NR)/12	Electronic patient registry	PTH measurement	Yesa/Yes	No	Assessment of proteinuria, Hb, and PO4
				Clinician education
Manns et al. (39)/Canada/C=78 (44); QI=78 (45)	Primary care practices	3–5 with diabetes or MA	5444 (2544)/26	Clinician education	Use of ACE/ARB	No/Yes	Dialysis	Composite outcome (death, dialysis, doubling of serum creatinine, and hospitalization for myocardial infarction, heart failure, or stroke)
				Clinician reminders				Use of statin, BP medication
								Proteinuria assessment
								Nephrology referral
Scherpbier et al. (40)/The Netherlands/C=72 (53); QI=74 (38)	Primary care practices part of an academic research network	3 with diabetes or HTN	164 (95)/12	Case management	BP	Yes/Yes	eGFR, Proteinuria	Use of statins
				Team changes				Functional health status
				Clinician education				Body mass index
								Body weight
								Nephrology referrals
								Laboratory results (HbA1c, lipids, Hb, Ca, PO4, PTH, albumin)
								Smoking cessation

QI, quality improvement; ACE, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; C, control; ASA, aspirin; HbA1c, glycated hemoglobin; Hb, serum hemoglobin; Ca, serum calcium; PO⁴, serum phosphate; PTH, parathyroid hormone; PCP, primary care physician; Cr, creatinine; MA, microalbuminuria; HTN, hypertension; NR, not reported.

^aPresented as a dichotomous outcome.

^bMissing before or after absolute values.

^cOutcomes presented as hazard ratios.

^dStratified on the basis of effectiveness of the intervention.

The two study types (patient-level randomized trials versus cluster-randomized trials) differed with respect to sample size, setting, population, study outcomes, and the types of QI initiatives tested. Specifically, patient-level randomized trials were smaller, situated in predialysis (47%) or diabetes clinics (13%), tended to include younger male patients, and evaluated hard clinical outcomes (e.g., mortality, dialysis, cardiovascular events). Cluster-randomized trials targeted primary care practices and mostly focused on process measures (e.g., use of renin-angiotensin-aldosterone system inhibitors). Patient-level randomized trials mainly targeted health system changes (case management and team changes) or patients (self-management and education), and the QI strategies usually involved nurses, dieticians, and/or health care educators. In contrast, cluster-randomized trials targeted clinicians with QI strategies aimed at education or clinical reminders. CKD registries (38), facilitated relay of communication (31), and performance audit with feedback (16) were each evaluated by a single study. No studies evaluated continuous QI strategies, and no study targeted or involved pharmacists (Table 2).

Most studies (n=18, 86%) included process measures linked to appropriate outcomes (e.g., reminder to prescribe renin-angiotensin-aldosterone inhibitors, use of renin-angiotensin-aldosterone system inhibitors, and BP). Seven patient-randomized trials (21,25,26,30,31,33,34,36) that reported processes and outcomes incorporated checklists, algorithms, or clinical care targets, whereas five cluster-randomized trials (16,37–40) linked processes to outcomes using information technology. Although most studies sufficiently described the QI interventions to facilitate replication (n=19, 90%) (14,16,21–28,30–37,39,40), few reported on pilot testing (n=5, 24%) (14,16,21,35,36) and intervention fidelity (i.e., the QI intervention was used as intended; n=5, 24%) (14,22,23,36,38). Only two studies (14,36) reported a QI intervention fidelity >75%. A small number of studies described adverse events or unintended consequences of the QI changes (n=4, 19%) (21,22,30,33,34), as well as patient (n=4, 19%) (21,22,27,36) or health care team (n=0) experiences with the QI intervention.

Synthesis of Results

The results of the individual studies are presented by outcome in forest plots (Figure 2). We did not identify enough studies of similar outcomes to stratify the results by QI strategy or randomization type.

Figure 2.

Results of meta-analyses for prespecified outcomes and process measures. 95% CI, 95% confidence interval; ACE, angiotensin-converting enzyme; ARB, angiotensin receptor blocker; IV, inverse variance; M-H, Mantel-Haenszel.

Process Measures: Use of Angiotensin-Converting Enzyme Inhibitors and Angiotensin Receptor Blockers

Ten studies (14,16,21,25,30,33,37–40) reported on the use of angiotensin-converting enzyme (ACE) inhibitors or angiotensin receptor blockers (ARBs). Of these, we included nine in the meta-analysis (n=6107), as the author of one cluster-randomized study was unable to provide additional information necessary to calculate an effective sample size (38). QI interventions conferred a higher likelihood of being prescribed an ACE inhibitor or ARB compared with usual care (RR, 1.16; 95% confidence interval, [95% CI], 1.06 to 1.27; I²=81%) after a median follow-up of 24 months.

Surrogate Outcomes

BP.

Nine studies (14,16,25,30,31,33,36,37,40) reported on changes in systolic BP, and eight studies (14,25,30,31,33,36,37,40) reported on changes in diastolic BP. Of these, we included seven (n=2492) and six (n=1454) in the meta-analysis, respectively, as two authors were unable to provide absolute BP results for the entire study population (14,36). Median follow-up was 24 months for systolic BP and 18 months for diastolic BP. There was no significant change in systolic (MD, −2.8 mm Hg; 95% CI, −5.7 to 0.1; I²=63%) and diastolic BP (MD, −1.0 mm Hg; 95% CI, −2.6 to 0.6; I²=0%) compared with usual care.

eGFR.

Thirteen studies (14,16,21,24–27,30,31,33,36,37,40) reported on changes in eGFR. Of these, we included 12 studies (n=3263) in the meta-analysis because one study did not report the final absolute eGFR values (21). After a median follow-up of 12 months, QI strategies did not lead to a significant change in eGFR compared with usual care (MD, 1.0 ml/min per 1.73 m²; 95% CI, −0.6 to 2.6; I²=63%).

HbA1c.

Changes in HbA1c were reported in eight studies (21,25,30,31,33,35,36,40) with a median follow-up of 12 months. We included seven studies (n=1415) in the meta-analysis because one study reported HbA1c as a dichotomous outcome (21). There was no significant change in HbA1c among groups receiving QI interventions compared with usual care (MD, −0.30%; 95% CI, −0.70 to 0.11; I²=90%).

LDL Cholesterol.

Five studies (21,25,30,33,40) reported on changes in LDL cholesterol with a median follow-up of 48 months. We included four studies (n=1176) in the meta-analysis because one study reported LDL cholesterol as a dichotomous outcome (21). QI interventions led to a statistically significant decrease in LDL cholesterol compared with usual care (MD, −17.6 mg/dl; 95% CI, −28.7 to −6.5; I²=75%).

Clinical Outcomes

Mortality.

Nine studies (16,21,25,26,28–31,33) compared QI strategies versus usual care for the outcome of mortality. Of these, we included eight studies (n=3853) in the meta-analysis because mortality was reported only as a hazard ratio in one study (29). There was no significant decrease in mortality after a median follow-up of 24 months among groups receiving QI interventions compared with usual care (RR, 0.94; 95% CI, 0.72 to 1.23; I²=21%).

Dialysis.

Dialysis was reported as an outcome measure by eight studies (21,25,26,28–31,33), with seven (n=2000) included in the meta-analysis because dialysis was reported only as a hazard ratio in one study (29). QI interventions lowered the incidence of dialysis (RR, 0.85; 95% CI, 0.74 to 0.97; I²=0%) after a median follow-up of 24 months. The number needed to treat in order to prevent one dialysis episode was 32. When the largest study by Devins et al. (28) was excluded in a post hoc analysis, the result was no longer statistically significant (RR, 0.89; 95% CI, 0.69 to 1.14; Supplemental Figure 3).

Cardiovascular Events.

Seven studies (16,21,25,30,31,34,39) reported on cardiovascular events. Of these, we included six in the meta-analysis (n=3526), as one study included cardiovascular events in a composite end point along with dialysis (39). There was no significant difference in cardiovascular events in the QI intervention group relative to usual care (RR, 0.82; 95% CI, 0.64 to 1.04; I²=0%) after a median follow-up of 24 months.

Risk of Bias

There were important differences in the risk of bias among the studies, with no single study satisfying all six criteria (Table 3). Three studies satisfied five out of six criteria (30,36,39). Random sequence generation was adequately reported in 17 out of 21 (81%) studies, and 12 out of 21 (57%) adequately reported concealing the allocation sequence. None of the studies masked treatment allocation from patients and the health care team, but six (29%) studies blinded outcome assessors. Outcome data were deemed complete in ten out of 21 (48%) studies, only one of which was a cluster-randomized trial (39). Outcome reporting was sufficient in 15 out of 21 (71%) studies.

Table 3.

Risk of bias

Study	Random Sequence Generation	Allocation Concealment	Blinding of Participants/Personnel	Blinding of Outcome Assessment	Incomplete Outcome Data	Selective Outcome Reporting
Patient-randomized trials
Barrett et al. (21)	Low	Low	High	High	Low	Low
Blakeman et al. (22)	Low	Low	High	High	Low	High
Boulware et al. (23)	Low	Low	High	High	High	Low
Campbell et al. (24)	Low	Low	Unclear	Unclear	High	Low
Chan et al. (25)	Low	Low	Unclear	Unclear	Low	Low
Chen et al. (26)	Low	Unclear	High	High	Low	Low
Degen (27)	Low	Low	Unclear	Unclear	High	Low
Devins et al. (28)	Low	Unclear	High	Low	High	High
Devins et al. (29)	Low	Unclear	High	High	High	High
Gaede et al. (30)	Low	Low	High	Low	Low	Low
Hotu et al. (31)	Unclear	Unclear	High	Unclear	High	Low
Manns et al. (32)	Low	Low	High	Unclear	Low	Low
MASTERPLAN: Peeters et al. (33); van Zuilen et al. (34)	Low	Low	High	Low	Low	Low
Steed et al. (35)	High	Unclear	Unclear	High	Low	High
Williams et al. (36)	Low	Low	High	Low	Low	Low
Cluster-randomized trials
Abdel-Kader et al. (37)	Low	Unclear	High	High	High	Low
Cortés-Sanabria et al. (14)	High	Unclear	Unclear	Unclear	High	High
Lusignan et al. (16)	Low	Low	High	Unclear	High	Low
Drawz et al. (38)	Low	Unclear	High	Low	High	Low
Manns et al. (39)	Low	Low	Unclear	Low	Low	Low
Scherpbier et al. (40)	Unclear	Unclear	High	High	High	High

Low, low risk of bias; high, high risk of bias; unclear, unclear risk of bias.

Publication Bias

No evidence of publication bias for the included outcomes was suggested by visual inspection of the funnel plots.

Discussion

In this systematic review of 11 QI interventions designed to improve care for patients with nondialysis-requiring CKD, we identified that QI interventions significantly reduced dialysis incidence, increased the use of ACE inhibitors and ARBs, and reduced LDL cholesterol concentrations. There were no significant differences observed in mortality, cardiovascular events, eGFR, HbA1c, and systolic or diastolic BP. This study is the first to demonstrate that QI initiatives have the potential to improve CKD care in several areas, while also identifying gaps in the application of QI principles to inform future QI work and research studies.

Our ability to identify an effect for several QI strategies on different outcomes may have been limited by high study heterogeneity. Factors that contributed to heterogeneity included the patient populations, descriptions of control groups, practice patterns, and treatment targets on the basis of different clinical practice guidelines. The duration of follow-up also varied between studies, which is particularly important in QI and CKD research because QI strategy adoption takes time, and the effects of interventions on lifestyle and risk factor modification may require years for their results to modify surrogate and hard outcomes. Process measures are easier to affect quickly with QI (41), which may explain the observed 16% improvement in the use of ACE inhibitors and ARBs. Surrogate outcomes with longer durations of follow-up (i.e., >18 months) included LDL cholesterol and systolic BP, with QI strategies significantly lowering LDL cholesterol and approaching statistical significance for systolic BP. On the other hand, diastolic BP, eGFR, and HbA1c all had much shorter durations of follow-up, which may explain their smaller observed effects. Reducing mortality, cardiovascular events, or dialysis incidence requires an even longer period of time. Therefore, the effect of QI strategies on dialysis incidence should be interpreted cautiously, despite being driven by studies of longer duration (28,30,33), two of which were of high quality (30,33).

The nature of the QI intervention also contributes to its effectiveness. Of the interventions examined, those that simultaneously targeted health systems (e.g., case management and team changes) and patients (e.g., self-management and education) appeared most effective. However, we were unable to quantify this result given the small number of studies identified for each individual QI strategy. Nevertheless, the observed benefits of case management and team changes are not surprising, as the effectiveness of these strategies has been demonstrated for other chronic conditions, such as diabetes mellitus and heart failure (8,42). These two improvement strategies provide clinicians with a starting point for their own QI initiatives that target patients with CKD.

Although our search strategy was comprehensive and included a broad range of QI interventions and outcomes, we still found important CKD processes and outcomes to be missing. Specifically, few trials evaluated QI interventions that targeted anemia, bone disease, vascular access, RRT selection (including preemptive transplantation), and hospitalizations. These knowledge gaps are concerning, especially because in many cases, evidence-based targets exist [e.g., anemia (43), vascular access (44), preemptive transplantation (2)] and/or QI initiatives have proven successful in other patient populations [e.g., anemia algorithms (45), Fistula First Initiative (46), alternative hospitalization programs (47)]. We also noted some QI strategies were rarely evaluated, especially in the health care system domain. This knowledge gap is important because QI strategies that target the health care system appear to be the most effective (8,42). Although we identified many studies on case management and team changes, no high-quality trials evaluated electronic patient registries, facilitated relay of information, or continuous QI. Moreover, only one study (16) assessed performance audit and feedback, a QI strategy with a strong evidence base (48). Another missing QI component was pharmacist involvement, whose engagement in QI has improved outcomes for patients with hypertension (49) and diabetes (50), and reduced hospitalizations, ESRD, and BP in uncontrolled studies for patients with CKD (51).

Published guidelines exist on the reporting and assessment of QI interventions, which emphasize describing the intervention in sufficient detail to facilitate replication, establishing whether the observed outcomes are actually due to the intervention, and identifying unintended consequences/adverse events (52,53). Most studies adequately described the QI initiative, but only five studies reported intervention fidelity (14,22,23,36,38), the proportion of the time that the intervention actually occurs as intended, and only two studies achieved fidelity >75% (14,36). Failure to achieve high fidelity makes it difficult to determine whether a lack of an effect is due to poor QI implementation of an effective intervention or an intervention that is inherently ineffective. Therefore, QI studies with higher fidelity may help extend the benefits we observed to other CKD outcomes. Health system change is also complex, and even though all improvement requires change, change does not necessarily lead to improvement (54). However, only four studies (21,22,30,33,34) described potential adverse effects on the system (e.g., opportunity and financial costs), and the same number reported patient experiences with the change (21,22,27,36). No studies reported feedback from the health care personnel involved in the QI change, which makes it difficult to determine the sustainability of a QI intervention. Although no single study followed all of these QI principles, three studies that used team changes and algorithms/checklists also measured opportunity costs (i.e., added clinic time) (21,30,33,34). The addition of a mechanism to measure fidelity (e.g., electronic checklist, intervention adherence audit, or process measurement) would have further strengthened these studies.

Future QI projects and evaluations should consider these three examples, along with targeting changes to the health care system rather than changes that only target the individual (clinician or patient). Although education or academic detailing may be a necessary component of QI, they are usually not sufficient to sustain long-term change (55). System changes are more difficult to work around, reinforce beneficial care processes, may make care easier for clinicians, and are associated with improved outcomes (8,56,57). Clinicians and health researchers are encouraged to utilize the provided framework (Supplemental Figure 2) or expand upon it, as long as the change targets the health care system and the main cause of their quality of care problem. The health care system change should then be tested on a small scale to ensure fidelity, efficacy, and satisfaction before evaluation in a large interventional study (58).

Strengths of our meta-analysis include the use of both an established search strategy and operational definitions for QI interventions to facilitate comparisons to other chronic illnesses. We also assessed studies on the basis of recognized standards of QI methodology, such as the Standards for Quality Improvement Reporting Excellence guidelines (52).

Our systematic review has several limitations. First, we could not include data from seven trials that reported CKD subgroups, despite our attempts to contact authors directly. These studies involved only a few hundred patients with CKD, so are unlikely to affect our results. Second, we were unable to stratify our results by QI strategy or baseline risk factors (e.g., CKD stage, systolic BP >140 mm Hg) because of the small number of eligible trials for each outcome, preventing formal comparison of the effectiveness of different QI strategies, analysis of whether certain QI strategies are more effective for different outcomes, and determination of whether QI strategies are more effective for patients not currently meeting evidence-based targets. This gap in the literature affects the generalizability of our results; for example, we could not identify the stage of CKD at which QI strategies may reduce dialysis incidence. Third, heterogeneity was substantial for several QI strategies and follow-up durations, as described above. Fourth, some patients in the usual care group may have crossed over to the QI intervention group, resulting in treatment contamination. However, most trials performed an intention-to-treat analysis, so contamination would bias our results toward no effect. Fifth, our search only identified randomized studies, and so would not identify quasi-experimental evaluations that are common in QI research.

In summary, we found that QI interventions in patients with nondialysis-requiring CKD have a favorable effect on dialysis incidence, the use of renin-angiotensin-aldosterone system blockade, and LDL cholesterol concentrations. Estimates of QI strategy effectiveness were limited by study number and adherence to QI principles. The number of tested QI initiatives for patients with CKD was found to be far behind other chronic illnesses, such as diabetes mellitus (8) and heart failure (42,59). Given the prevalence of CKD and associated consequences and costs, additional rigorously tested QI interventions have the potential to inform practice and materially improve the public health.

Disclosures

None.