Article Text
PDF
Abstract
Objective To estimate the direction and magnitude of effect and quality of evidence for hospital-based heart failure (HF) quality improvement interventions on process of care measures and clinical outcomes among patients with acute HF.
Review methods We performed a structured search to identify relevant randomised trials evaluating the effect of in-hospital quality improvement interventions for patients hospitalised with HF through February 2017. Studies were independently reviewed in duplicate for key characteristics, outcomes were summarised and a qualitative synthesis was performed due to substantial heterogeneity.
Results From 3615 records, 14 randomised controlled trials were identified for inclusion with multifaceted interventions. There was a trend towards higher in-hospital use of ACE inhibitors (ACE-I; 57.9%vs40.0%) and beta-blockers (BBs; 46.7%vs10.2%) in the intervention than the comparator in one trial (n=429 participants). Five trials (n=78 727 participants) demonstrated no effect of the intervention on use of ACE-I or angiotensin receptor blocker at discharge. Three trials (n=89 660 participants) reported no effect on use of BB at discharge. Two trials (n=419 participants) demonstrated a trend towards lower hospital readmission up to 90 days after discharge. There was no consistent effect of the quality improvement intervention on 30-day all-cause mortality, hospital length of stay and patient-level health-related quality of life.
Conclusions Randomised trials of hospital-based HF quality improvement interventions do not show a consistent effect on most process of care measures and clinical outcomes. The overall quality of evidence for the prespecified primary and key secondary outcomes was very low to moderate, suggesting that future research will likely influence these estimates.
Trial registration number CRD42016049545.
- heart failure
- quality and outcomes of care
- systemic review
Statistics from Altmetric.com
Introduction
Heart failure (HF) is an end-stage manifestation of cardiovascular disease and a growing cause of global cardiovascular morbidity and mortality.1 2 HF affects an estimated 26 million people worldwide and is estimated to have cost $108 billion globally in 2012.3 At 40 years of age, the lifetime risk of developing HF is one in four in the USA and even higher in some race/ethnic subgroups.4 The prevalence of HF in the USA is expected to increase 46% from 2012 to 2030 leading to greater than 8 million patients with HF and approximately 50% of patients diagnosed with HF will die within 5 years.5 6
Given the high mortality and morbidity rates of patients with HF and disparities in the use of guideline-recommended therapy,7 8 quality improvement initiatives have been developed to improve clinical outcomes. These quality improvement initiatives have largely targeted inpatient HF management such as the American Heart Association’s Get With the Guidelines programme, which has been associated with improvements in processes of care and lower 30-day readmission rates.9 Large-scale, hospital-based registries with associated quality improvement tools, such as the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF) and the Acute Decompensated Heart Failure National Registry (ADHERE), have also demonstrated the use of inpatient process-of-care improvement tools are associated with better quality of care as defined by use of guideline-recommended therapy, adherence to performance measures and shorter length of hospitalisation.7 10 However, whether these relationships are causal or are driven by patient-level, provider-level or institutional-level confounders is less certain but appears possible.
To understand which components of these multifaceted quality improvement interventions are effective, several research teams have led randomised trials evaluating the effect of HF quality improvement interventions to overcome the potential confounding and bias inherent in non-randomised studies. Most hospital-based quality improvement interventions include admission and discharge checklists, personalised site performance feedback and patient education.11 These interventions may be particularly useful in low-income and middle-income countries (LMICs) where the quality of HF care remains poor,12 and adherence to guideline-recommended therapy is suboptimal.13 The objective of this systematic review is to estimate the direction and magnitude of effect and quality of evidence for randomised trials of hospital-based HF quality improvement interventions on process of care measures and clinical outcomes among patients with acute HF.
Methods
Literature search and study selection
This review evaluated the effect of in-hospital heart failure quality improvement interventions in hospitalised patients with heart failure compared with usual care on the primary outcomes of in-hospital mortality, in-hospital and discharge medical therapy and hospital readmissions up to 90 days after discharge (online appendix 1). We followed guidelines published by the Cochrane Collaboration to synthesise the effects of multiple interventions,14 and the prespecified protocol was prospectively registered.15
Supplemental material
We searched multiple bibliometric databases for published literature from date of inception to 6 February 2017. We also searched ClinicalTrials.gov for records of ongoing trials and unpublished studies on 13 February 2017. An experienced information specialist (MAB) performed all searches. We contacted study authors of included trials when necessary to identify additional information we might have missed. We used the reference section of published trials that met our inclusion criteria as an added resource to identify other trials. Details of the search methods are provided in online appendix 2.
Eligibility criteria
We included randomised trials of HF quality improvement interventions. We sought to include quasi-randomised (eg, interrupted time series), but there were none identified in our search. We excluded general, hospital-based quality improvement interventions. We included trials that tested the effect of a variety of individual or combined interventions including but not limited to: audit and feedback reporting systems, admission and discharge checklists, chart case management, patient educational or behavioural change materials, healthcare quality training that was directed at the hospital system, doctors, nurses or allied health professionals or information management systems with the goal of being inclusive in the type and target of intervention. We included trials evaluating the effect of these interventions among individuals admitted to hospitals for the management of acute HF.
Data extraction
Three authors (AA, EB and SGKY) independently screened abstracts, titles and full texts of retrieved studies in duplicate to identify studies to be assessed further. Two authors (AA and SGKY) independently extracted key data using a structured data extraction form and assessed risk of bias using the Cochrane Risk of Bias Tool. Discrepancies in risk of bias assessment were resolved through consensus or through review with a third review author (MDH).
Outcomes
We included a combination of process of care measures and clinical outcomes. The primary outcomes included: (1) rate of in-hospital mortality, (2) rates of in-hospital and discharge medical therapy and (3) rate of hospital readmission up to 90 days after discharge. The secondary outcomes included: (1) rates of 30-day and 90 day all-cause mortality, (2) hospital length of stay, (3) uptake of quality improvement intervention components, (4) rate of cardiac rehabilitation referral, (5) rate of implantable cardioverter defibrillator (ICD) placement to prevent sudden death in eligible patients, (6) patient-level health-related quality of life using a validated instrument and (7) patient-related direct costs. We created a summary of findings table reporting the effects of the intervention and the quality of evidence related to the primary and key secondary outcomes.
Quality assessment
Two authors (AA, SGKY) independently evaluated the quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework, which takes into account issues related to internal validity and external validity.16 Discrepancies were resolved through consensus or through review with a third review author (MDH).
Data synthesis
We planned to perform a meta-analysis, but there was substantial, unexplained heterogeneity across the studies, as well as differences in their presentation of intervention effects that precluded pooling and reporting of summary effect estimates. Thus, we present a narrative, qualitative synthesis.
Results
Search results
The initial search identified 3615 records after merging data from independent searches and removing duplicates (figure 1). After title and abstract screening, we evaluated 69 full-text articles and identified 14 studies for inclusion in the review.11 17–29 Details regarding studies that were excluded are reported in online appendix 3, and details regarding ongoing studies are reported in online appendix 4.
PRISMA flow chart of included studies. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.
Characteristics of the included studies
A description of the 14 included studies is provided in online appendix 5, and detailed characteristics of the included studies are provided in online appendix 6. We identified six cluster randomised control trials (cRCTs) in which the unit of randomisation was the hospital11 18 24–26 29 and eight RCTs in which the unit of randomisation was the individual.17 19–23 27 28 Eight studies were conducted in a single academic centre.17 19–23 27 28 Ten studies were conducted in the USA,11 18–20 22–24 26–28 and four studies were conducted in Sweden,17 Japan,21 Italy25 and Canada.29 Two cRCTs contributed the greatest number of participants (DeVore et al: 71 829 participants11; Tu et al: 17 544 participants29), and three RCTs included <200 participants.17 21 28 Participants were hospitalised patients with HF, and type or cause of HF was generally not reported. Two studies also included outpatients with HF, and disaggregated data specifically for hospitalised patients with HF were not available.19 20 Participants’ mean age ranged from 63.0 to 80.7 years, and the proportion of females in each study ranged from 1% to 63%. Two trials did not report age nor sex.18 24
Trial-specific risk of bias assessment is presented in figure 2, and detailed risk of bias assessment is presented in online appendix 6. Thirteen trials had low risk of selection bias based on reported methods of random sequence generation,11 17 19–29 but only four trials had low risk of selection bias based on reported methods of allocation concealment.19 20 22 27 All others had an unclear risk of selection bias based on reported methods of allocation concealment.11 17 18 21 23–26 28 29 Most trials (n=12) did not blind study personnel nor participants given the nature of the quality improvement intervention and were at high risk of performance bias.11 17 18 21–29 One trial had high risk of detection bias from lack of blinding of outcome assessors,23 four trials had low risk of detection bias17 18 20 28 and the remainder had an unclear risk of detection bias from reported methods of outcome assessment.11 19 21 22 24–27 29 Five trials had high risk of attrition bias from incomplete outcome data.19–21 23 24 Two trials had high risk of reporting bias from selective reporting based on previously published protocols,20 29 2 trials had low risk of reporting bias11 25 and 10 trials had unclear risk of reporting bias.17–19 21–24 26–28
Risk of bias assessment of included studies.
Interventions
In-hospital HF quality improvement interventions varied considerably across trials. Five trials tested multifaceted quality improvement interventions that included various combinations of HF specific patient education, HF guideline clinical staff education, admission order sets, in-hospital guideline-recommended medication reminders, discharge checklists, case management coordination of care, performance reports and postdischarge telephone calls.18 23–26 Interventions in five trials emphasised HF specific patient education on medication management including self-management of diuretics, maintaining a symptom diary, diet and tobacco and alcohol cessation.17 21 22 27 28 Two large cRCTs evaluated the specific effect of audit and feedback by sharing personalised quality improvement reports highlighting performance measures to each hospital site11 or via a publicly released report card.29 Two studies evaluated the effect of automated guideline-recommended medication reminders attached to echocardiography reports.19 20
Primary outcomes
Primary outcomes included rate of in-hospital mortality, rates of in-hospital and discharge medical therapy and rate of hospital readmission up to 90 days after discharge (table 1). Overall, mean rates of in-hospital mortality presented in three trials (n=75 164 participants) ranged from 3.4% to 5.6% in the intervention group compared with 3.4% to 15.4% in the comparator group.11 25 26 Two trials revealed no effect of the quality improvement intervention on in-hospital mortality. The smallest of the three trials reporting in-hospital mortality demonstrated a decrease from 15.4% (33 events/215 participants) in the comparator group to 5.6% (12 events/214 participants) in the intervention group.25 One trial (n=429 participants) measured rates of in-hospital medical therapy. In-hospital use of ACE inhibitors (ACE-Is) increased from 40.0% in the comparator group to 57.9% in the intervention group, and in-hospital use of beta-blockers (BBs) increased from 10.2% in the comparator group to 46.7% in the intervention group.25 Six trials (n=96 271 participants) measured rates of discharge medical therapy.11 18 23 25 26 29 Five out of six trials demonstrated no effect of the quality improvement intervention on use of ACE-I or angiotensin receptor blocker (ARB) at discharge. In the largest cRCT (n=71 829 participants), the estimated treatment effect (95% CI) of the intervention on ACE-I or ARB at discharge was 0.8% (95% CI −2.7% to 4.2%; p=0.67) between the intervention and control groups.11 The second largest cRCT (n=17 544 participants) revealed the absolute difference (95% CI) for intervention versus control on ACE-I or ARB at discharge for participants with left ventricular systolic dysfunction was 5.9% (95% CI 1.0% to 10.7%; p=0.02) higher in the intervention group.29 Three trials reported no effect of the quality improvement intervention on use of BB at discharge,11 23 29 and one trial reported no effect of the intervention on use of aldosterone antagonist at discharge.11 Of the three small trials reporting hospital readmissions up to 90 days after discharge (n=706 participants), two demonstrated an effect of the intervention on reducing hospital readmissions from 67% and 19% in the comparator to 37% and 7%, respectively.23 27 28
Summary of outcomes from included studies
Secondary outcomes
Secondary outcomes included rates of 30-day and 90-day all-cause mortality, hospital length of stay, uptake of quality improvement intervention components, rate of cardiac rehabilitation referral, rate of ICD placement to prevent sudden death in eligible patients, patient-level health-related quality of life using a validated instrument and patient-related direct costs (table 1). Prespecified outcomes that were not reported by any of the included trials include: uptake of quality improvement intervention components, rate of cardiac rehabilitation referral, rate of ICD placement to prevent sudden death in eligible patients and patient-related direct costs. Two studies (n=17 681 participants) demonstrated no effect of the quality improvement intervention on 30-day all-cause mortality with the outcome ranging from 7.5% to 10.6% in the comparator to 7.1% to 9.6% in the intervention, respectively (p=1.00, p=0.26).28 29 Between two studies reporting hospital length of stay (n=3335 participants), the smaller trial showed a decrease in the mean (95% CI) hospital length of stay by 1 day in the intervention group from 11.4 (95% CI 10.5 to 12.3) days to 10.4 (95% CI 9.6 to 11.0; p=0.03) days in the comparator group.25 26 Three trials (n=3411 participants) assessed patient-level health-related quality of life using validated questionnaires.22 26 27 Only one trial demonstrated an improvement in patient-level health-related quality of life as assessed by the Chronic Heart Failure Questionnaire with mean (SD) score improving from 11.3 (16.4) in the comparator to 22.1 (20.8) in the intervention,27 but the other two trials demonstrated no difference.
Study quality assessment
Outcome-specific assessment of the quality of included studies using the GRADE framework is presented in table 2. The quality of evidence for the primary outcomes of in-hospital mortality and in-hospital medical therapy was low, and both outcomes were downgraded for inconsistency and imprecision.16 The quality of evidence for the primary outcome of discharge medical therapy was moderate, downgraded for inconsistency as the direction of effect was inconsistent and point estimates varied across studies. The quality of evidence was low for the secondary outcome of hospital readmission up to 90 days after discharge, downgraded for inconsistency and imprecision. The quality of evidence was low for the secondary outcomes of 30 day all-cause mortality and hospital length of stay. For the secondary outcome of patient-level health-related quality of life, the quality of evidence was very low, downgraded for inconsistency, imprecision, and study limitations.
Summary of findings
Discussion
Randomised trials of hospital-based HF quality improvement interventions do not show a consistent effect on process of care measures nor clinical outcomes including in-hospital mortality, discharge medical therapy, 30-day all-cause mortality, hospital length of stay and patient-level health-related quality of life. These interventions appear to increase utilisation of in-hospital ACE-I and BB and may reduce hospital admission up to 90 days after discharge. To our knowledge, this is the first systematic review of randomised trials to evaluate the effect of in-hospital HF quality improvement interventions compared with usual care on a range of outcomes of importance in hospitalised patients with HF.
The varied interventions studied in these RCTs have more modest effects on process of care measures than non-randomised studies suggest. During the 2-year OPTIMIZE-HF observational quality improvement registry with 259 participating US hospitals, the prescription rates of BB at discharge increased from 76.3% to 86.4% (p<0.001).10 However, in an cRCT of 147 hospitals participating in the Get With The Guidelines – Heart Failure quality improvement programme, there was no change in the already high rates of BB at discharge (96.8% in the intervention group vs 95.5% in the comparator group; estimated treatment effect: 1.05, 95% CI −1.17 to 3.27, p=0.36) as a result of personalised performance feedback.11 Furthermore, median rates of use of ACE-I in participants with left ventricular systolic dysfunction in the ADHERE registry were 72.0%, revealing a gap in adherence to established quality-of-care indicators during the period of data collection (July 2002 to December 2003).7 However, both subsequent observational HF registry and RCT data demonstrate little effect of quality improvement interventions on treatment rates of ACE-I or ARB at discharge for eligible participants,10 11 18 23 25 26 which may be driven by favourable temporal trends in medication use in the comparator groups.
It can be challenging to assess which elements of multicomponent interventions, including audit and feedback on performance metrics, standardised HF checklists for admission and discharge, clinical pathways and patient education, influence heterogeneous process of care measures and clinical outcomes. Two large cRCTs demonstrated no effect of release of performance indicators either via personalised feedback to the hospital or publicly released report cards on quality of care measures.11 29 The trials did not report process changes at the hospital level implemented as a result of the performance feedback; hence, there may be substantial heterogeneity across hospitals in their local quality improvement activities. The trials that focused primarily on patient education were small in size (n=38 to 282 participants) and may have been underpowered to demonstrate a potential effect on measured outcomes.17 21 22 27 28
Notably, there were no RCTs evaluating in-hospital HF quality improvement interventions from LMICs. The overall baseline quality of care in the USA, Sweden, Japan and Italy at the time of these trials may have been too high to demonstrate a significant effect of these interventions because of background improvements in the quality and safety of in-hospital HF care. Countries with lower quality of HF care including lower baseline use of guideline-recommended therapy may have a greater opportunity to benefit from in-hospital HF quality improvement interventions. For example, in contrast to baseline rates of ACE-I or ARB at discharge in ADHERE (72.5%), OPTIMIZE-HF (84% in first quarter) and GWTG-HF (93.9% in control), the Trivandrum Heart Failure Registry in India revealed lower rates of participants with left ventricular systolic dysfunction prescribed ACE-I at discharge (41.0%).7 10 11 13 The Acute Coronary Syndrome Quality Improvement in Kerala (ACS QUIK) cluster randomised, stepped wedge trial evaluated the effect of an in-hospital quality improvement toolkit for participants with acute myocardial infarction in India and demonstrated modestly higher rates of discharge medical therapy among the intervention group. This suggests potential gains in process of care measures with hospital-based HF quality improvement interventions may be possible in this setting.30 Whether such gains would translate into improvements in clinical outcomes, which were not observed in ACS QUIK, remains uncertain.
This systematic review has several strengths. First, this review includes only RCTs to better examine the potential effect of quality improvement interventions. Second, the prespecified published protocol prior to initiation of the systematic review guided the search strategy and evidence synthesis to minimise the risk of bias in the review process.15 Third, title screening, data extraction and quality assessments were performed in duplicate to minimise error or bias in the review process.
This review also has important limitations. First, HF type was not consistently reported across studies, and hence, we have limited ability to comment on the effect of the interventions in patients with HF with preserved ejection fraction versus HF with reduced ejection fraction. Second, there is substantial heterogeneity in the types of in-hospital quality improvement interventions included in this review. Third, pooling of summary estimates was not performed due to substantial and unexplained heterogeneity across the studies. We did not include non-randomised studies during which time there have been temporal improvements in process of care measures and clinical outcomes; however, randomised studies are better suited to assess the causal relationships between complex interventions and outcomes.
Conclusions
This systematic review demonstrates that randomised trials of in-hospital quality improvement interventions do not have a consistent effect on process of care measures and clinical outcomes in hospitalised participants with HF. The overall quality of evidence for the prespecified primary and key secondary outcomes was very low to moderate, suggesting that future research will likely influence these estimates. This review demonstrates the gaps in the evidence base for in-hospital quality improvement interventions for acute HF and suggests future areas of inquiry including the need for trials in LMIC settings where baseline quality of care may be lower.
References
Footnotes
Contributors AA, EB, MAB, AFH, DP and MDH conceived and designed the study. MAB did the scientific literature search. AA, EB, SGKY and MDH screened abstracts and full texts. AA, SGKY and MDH performed data extraction, risk of bias and graded the quality of evidence. All authors contributed to the data analysis and critically revised the manuscript. AA is the study guarantor.
Funding AA received funding from the Fogarty International Center of the National Institutes of Health, Duke Global Health Institute and Duke Hubert-Yeargan Center for Global Health for this research. Research reported in this publication was supported by the Fogarty International Center and National Institute of Mental Health of the National Institutes of Health under Award Number D43TW010543. The authors are independent from the funders.
Disclaimer The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Competing interests MDH receives funding from the World Heart Federation to serve as its senior programme advisor for the Emerging Leaders programme, which is supported by Boehringer Ingelheim and Novartis with previous support from BUPA and AstraZeneca. MDH also receives support from the American Heart Association, Verily and AstraZeneca for work unrelated to this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All data are freely available within the appendices in the data supplement. No additional data available.
Patient consent for publication Not required.