Key Findings and Strength of Evidence

Reducing Readmissions and Mortality

Table 20 summarizes our key findings by intervention category, main outcomes (readmission rates, mortality, or the composite of all-cause readmission or death), and timing of measurement of outcomes (30 days; 3-6 months). It documents our results when data met the following three criteria: (1) sufficient evidence to grade the SOE and to draw a conclusion that evidence either supports benefit [+] or does not [-]; (2) insufficient (I) evidence to make a determination (e.g. only one trial reporting an outcome of interest); or (3) no included trials that reported an outcome (NR).

Table 20

Summary of key findings and strength of evidence, by intervention category and outcome.

Table 21 then presents more detailed results, including relative risks (RRs) and 95% confidence intervals (CIs). We also present numbers needed to treat (NNTs, when applicable) for comparisons that included at least one trial reporting an outcome of interest.

Table 21

Summary of key findings and strength of evidence for transitional care interventions: Readmission, composite endpoint, and mortality.

Outcomes at 30 Days

We found very little evidence on whether interventions reduce 30-day readmission rates. Most trials reported rates over 3 to 6 months.

The high-intensity home-visiting trial showed efficacy in reducing 30-day all-cause readmission and the composite endpoint (both low SOE).⁵⁵ Despite having only this trial, this intervention and other trials in the home-visiting category also consistently reduced readmission rates measured over 3 to 6 months. Because of those robust findings, we considered high-intensity home-visiting programs to be efficacious in reducing all-cause readmission and the composite outcome at 30 days.

Evidence was insufficient to determine whether several other intervention categories reduced 30-day all-cause readmission rates, because we found no trials or only a single trial (none of which showed efficacy).

Outcomes at 3 to 6 Months

For outcomes measured over 3 to 6 months, we found evidence of efficacy (high SOE) for improving at least one of our primary outcomes for home-visiting programs, STS, and MDS-HF clinic interventions. Specifically, home-visiting programs reduced all-cause readmission rates (high SOE), HF-specific readmission rates (moderate SOE) and the composite outcome (moderate SOE). STS interventions reduced HF-specific readmission rates (high SOE) and mortality (moderate SOE) but not all-cause readmission (moderate SOE for no benefit). MDS-HF clinic interventions reduced all-cause readmission rates (high SOE) and mortality (moderate SOE) but not HF-specific readmission (moderate SOE for no benefit).

For these outcomes, we calculated NNTs when data on a given intervention or outcome permitted (Table 21). For all-cause readmission, NNTs were 9 for home-visiting programs and 8 for MDS-HF clinic interventions. For mortality, NNTs were 33 for home-visiting programs, 27 for STS, and 18 for MDS-HF clinic interventions. An NNT of 9 means, for example, that 9 hospitalized patients with HF would need to receive a home-visiting program following discharge (rather than usual care) to prevent one additional person from being readmitted over 3 to 6 months.

Our meta-analyses did not find telemonitoring, nurse-led clinic interventions, or primarily educational interventions to be efficacious for any primary outcomes. Evidence was insufficient to determine the efficacy of the following interventions in reducing readmission rates or mortality: primary care clinic interventions, peer support interventions, and cognitive training interventions (for people with HF and coexisting mild cognitive impairment).

Some experts have cautioned that inappropriate focus on reduction of readmission rates could negatively affect patient care and perhaps mortality. However, we found no evidence of such an effect. Specifically, no interventions reduced readmission rates but increased mortality.

Findings in Relation to What Is Already Known

The 2009 update of the American Heart Association/American College of Cardiology (AHA/ACC) guidelines addressed postdischarge HF interventions. ^{25 25} These guidelines focused on the importance of discharge planning, emphasizing written discharge instructions or educational material targeted to the patient or caregiver at discharge. The AHA/ACC guidelines also recommended that “postdischarge systems of care, if available, should be used to facilitate the transition to effective outpatient care for patients hospitalized with heart failure.”

The 2010 Heart Failure Society of America guidelines are similar; their guidance emphasizes specific components of discharge planning. The Society does not give any specific guidance on optimal transitional care interventions aimed at preventing early readmissions for patients with HF.

As our results make clear, certain intervention categories (or combinations of components) have better efficacy than others for HF patients. For instance, categories of interventions that were delivered face-to-face (i.e., home-visiting and MDS-HF clinic interventions) reduced all-cause readmission rates, whereas other intervention categories did not. In general, intervention categories that included higher-intensity interventions, with education on self-care delivered over repeated visits through the transition from hospital to home (i.e., home-visiting programs, STS, MDS-HF clinic interventions) reduced all-cause readmission or mortality, whereas categories with lower-intensity intervention (i.e., primarily educational interventions, nurse-led HF clinic interventions) did not. Guidelines that focus on written discharge instructions and early outpatient followup may not provide sufficient guidance on optimal strategies to reduce readmission and mortality.

The interventions in the included trials were heterogeneous and could probably be categorized using a variety of approaches. We classified them in a manner that we believe is both descriptive and informative; other approaches to categorization could lead to different conclusions about the efficacy (or lack of it) for primary or other outcomes.

Other reviews have highlighted the difficulty in classifying trials into distinct categories. For example, we classified one trial by Rainville et al.⁷¹ as STS as did a 2011 Cochrane review,³² but a 2012 Cochrane review classified the same trial as case management, grouping it with trials that assessed home-visiting programs.³¹

The results of our current review are, in general, consistent with those in previous reviews of this literature. Unlike earlier reviews, however, we focused on shorter-term outcomes (6 months or less), with enrollment during or shortly following an admission for HF). By contrast, some other reviews included outcomes up to 12 months (or longer) following an index hospitalization or did not exclude studies in which patients had had no recent index hospitalization.^31,102,103 For these reasons, our conclusions may differ. For example, a recent meta-analysis and network meta-analysis found no statistically significant effect of remote monitoring interventions on mortality or all-cause readmission up to 1 year, but it combined STS and telemonitoring interventions.¹⁰² At outcome timings over 3 to 6 months, we found that STS was associated with a mortality benefit while telemonitoring interventions were not.

Other reviews have used different categorization strategies (as mentioned above); a 2009 Cochrane review found that “case-management” interventions (home-visiting programs and telephone support) were associated with a reduction in all-cause mortality at 12 months (but not at 6 months) and a reduction in readmissions related to HF at 6 months and 1 year.³¹ One recent review of interventions to reduce 30-day readmissions excluded interventions that were disease specific; that is, those authors did not include interventions tailored to HF patients or other chronic diseases.³³ They found that most studies tested multicomponent discharge bundles. Common components of effective interventions included postdischarge telephone calls and patient-centered discharge instructions—e.g., facilitation of patient engagement in the transition of care individually tailored for the patient's health and social circumstances.³³

Applicability

Most trials included adults with moderate to severe HF. The mean age of subjects was generally in the 70s; very few trials enrolled patients who were, on average, either younger or older. We did not find evidence to confirm or refute whether treatments are more or less efficacious for many other subgroups, including groups defined by sex, racial or ethnic minorities, people with higher severity of HF, type of HF (e.g., diastolic vs. systolic), and patients with certain coexisting conditions. Included trials commonly excluded patients who had end-stage renal disease or severe or unstable cardiovascular disease (e.g., recent myocardial infarction [MI]). Usual-care groups of included trials experienced rates of readmission that are comparable to those found in a Medicare population: the median all-cause readmission rate at 30 days among patients receiving usual care was 19 percent (intraquartile range [IQR] 14 percent to 22 percent); over 3 to 6 months, the rate was 43 percent (IQR 35 percent to 41 percent).

Although all included studies enrolled patients with HF, some degree of population heterogeneity across intervention categories may impact our findings (e.g., variation in HF severity, etiology, or number of comorbidities). The majority of trials included patients with moderate to severe HF based on the NYHA classification; most trials did not report whether patients had a preserved versus reduced ejection fraction. The interventions included are applicable only to patients who are discharged to home. Whether interventions would benefit patients who are discharged to another institution (e.g., assisted living facility) remains unclear.

Few trials used a risk stratification scheme to determine inclusion eligibility. Two trials assessing a home-visiting program (by the same author) specified inclusion or exclusion criteria based on risk of readmission. One of these trials included only patients with one “risk factor” for early readmission⁵⁷ (prior history of HF, four or more hospitalizations for any reason in the preceding 5 years, or HF precipitated by either an acute MI or uncontrolled hypertension); the other trial excluded patients deemed “low risk.”⁵⁸ Neither trial used a validated tool to assess risk.

Two other trials, one assessing a MDS-HF clinic intervention and one assessing an STS intervention, included patients deemed “high risk for readmission” but did not define how they assessed risk.^65,87 One trial testing a primarily educational intervention excluded patients who were judged to have a “noncardiac illness likely to increase 6-month mortality or hospitalization risk.”⁹⁵ How such inclusion criteria based on risk affect the applicability of results remains unclear. A systematic review of validated readmission risk prediction models found that most perform poorly.¹⁰⁴

Approximately one-half of the home-visiting programs were conducted in the United States; the others were conducted in Australia, the United Kingdom, and various European countries. One of three trials assessing MDS-HF clinic was conducted in the United States; the other two were conducted in Taiwan and Canada; despite the potential differences in settings and health care systems, results were similar across the three trials.

Across most included trials, the majority of patients were prescribed an angiotensin-converting enzyme inhibitor (ACEI) or angiotensin receptor blocker (ARB) (when information was reported). By contrast, the percentages of patients across trials who were prescribed beta-blockers at discharge varied widely across trials.

Whether “usual care” in trials published during the early 1990s is comparable to current practice is questionable. In general, trials did not report details of usual care; for example, they typically did not specify whether followup was scheduled soon after discharge or whether patients were receiving additional services such as home health care.

The trials we examined were conducted in a mix of inpatient settings. These settings included academic medical centers, Department of Veterans Affairs hospital settings, and community hospitals. Our findings are, therefore, generally applicable to many hospital settings in the United States.

Implications for Clinical and Policy Decisionmaking

Few trials reported readmission rates within 30 days following a HF hospitalization. Data based on Medicare claims suggest that 35.2 percent of 30-day readmissions are for HF; the remainder are for diverse indications (e.g., renal disorders, pneumonia, arrhythmias, and septicemia or shock).⁵ Whether certain interventions that reduce readmissions at 3 and 6 months would also be effective in reducing earlier readmissions remains uncertain.

We found moderately strong evidence for interventions that provided relatively frequent in-person monitoring following discharge—specifically, home-visiting programs and MDS-HF clinic interventions. The one trial that showed efficacy for reducing 30-day all-cause readmission provided frequent, in-home monitoring that began within 24 hours of discharge.⁵⁵ These findings suggest, in general, the utility of such face-to-face monitoring, whether delivered at home or at an appropriate clinic.

Interventions that did not show efficacy for all-cause readmissions tended to focus more narrowly on HF self-care alone; this finding was true, for instance, for both STS and primarily educational interventions. For reducing all-cause readmissions, focusing on HF self-care training or weight monitoring alone does not appear sufficient.

Current clinical practice in the care of adults with HF after hospitalization varies greatly across the country, by hospital settings, and for other factors. A recent telephone survey of 100 U.S. hospitals found wide variation in education, discharge processes, care transition, and quality-improvement methods for patients hospitalized with HF.¹⁷ As mentioned in the introduction to this review, readmission rates vary by both geographic location and insurance coverage.⁶

Nevertheless, our findings provide some guidance to quality-improvement efforts that try to reduce readmissions for people with HF. Specifically, systems or providers aiming to implement programs to improve transitional care for patients with HF may wish to start with the multicomponent interventions that had meaningful effects on both all-cause readmission and mortality. In this regard, we found moderately strong evidence for home-visiting programs and MDS-HF clinic interventions for reducing all-cause readmission rates; in designing quality-improvement efforts, administrators may do well to give these approaches the greatest consideration.

We did not prespecify a magnitude of effect (i.e., a specific reduction in relative risk) that clinicians or policymakers should consider a meaningful change in readmission rates. The percentage of readmissions that are preventable may differ across settings and patient populations that are not comparable with patients enrolled in the trials we examined. However, as mentioned above, the 30-day readmission rates of included trials are similar to those seen in Medicare populations.

Finally, we did not find direct evidence on whether certain types of interventions increase or decrease caregiver or self-care burden, but these issues are meaningful for patients and their families. Thus, clinicians should consider the inherent demands or expectations that some transitional care interventions may place on caregivers and patients (e.g., burden of transportation). In recommending or selecting a transitional care intervention, administrators may want to weigh the goal of minimizing additional burden that a candidate intervention might pose.

Limitations of the Comparative Effectiveness Review Process

The interventions in the included trials were quite diverse; they probably could be categorized using a variety of classification schemes or conceptual models. As explained previously, we classified them in a manner that we believe is both descriptive and informative; it accords with how numerous experts conceptualize these highly varied programs. Nonetheless, we acknowledge that other approaches to categorization could lead analysts to different conclusions.

Other reviews have highlighted the difficulty in classifying transitional care programs, particularly those aiming at specific kinds of outcomes or targeting specific types of patients, into distinct (mutually exclusive) categories. Using different categorizations for systematic reviews can place similar (or exactly the same) trials into different groupings (as noted above concerning the Rainville et al. trial). The effects of such classification decisions on findings, especially those relying on pooled data from eligible trials, are unpredictable.

We used the term “transitional care” broadly; generally, we were guided by Coleman's definition as “a set of actions designed to ensure the coordination and continuity of health care as patients transfer between different locations or different levels of care within the same location” (p. 30).²⁴ The included interventions are diverse in terms of whether they aimed to coordinate care at the provider level or focused more on strategies to transfer care back to the patient (e.g., through self-care training for managing HF). We did not include or exclude trials based on any specific set of components; for that reason, included trials assessed diverse interventions. We chose to cast a broad net to be able to examine a comprehensive set of strategies to reduce readmission rates, lower mortality, improve quality of life, or influence other patient-centered outcomes, on the grounds that doing so would be useful to stakeholders in different settings (hospitals, outpatient clinics, or others).

Our inclusion and exclusion criteria specified that included trials had to enroll patients during (or soon after) a hospitalization for HF and also had to measure a readmission rate over 6 months or less. We did not include readmission rates or mortality rates measured beyond 6 months; interventions that we did not find efficacious may or may not be beneficial in long-term disease management in patients with HF. Finally, publication bias and selective reporting are potential limitations. Although we searched for unpublished trials and unpublished outcomes, we did not find direct evidence of either of these biases. Many of the included trials were published before trial registries (e.g., clinicaltrials.gov) became available; had we been able to consult such registries, we would have had greater certainty about the potential for either type of bias.

Limitations of the Evidence Base

The evidence base was inadequate to draw conclusions for some of our questions or subquestions of interest. In particular, as described above, direct evidence was insufficient to permit us to draw any conclusions on comparative effectiveness of transitional care interventions.

In addition, evidence was quite limited for some outcomes (particularly readmissions within 30 days, utilization outcomes, and quality of life). Evidence was similarly insufficient to draw any definitive conclusions about whether any transitional care interventions are more or less efficacious in reducing readmissions or mortality based on patient subgroups defined by age, sex, race, ethnicity, socioeconomic status, disease severity, or coexisting conditions. We found just two eligible trials reporting information on different subgroups.

We identified little evidence on the potential harms of transitional care interventions, such as whether they increase caregiver burden or increase the rate of ER visits. None of the included trials measured caregiver burden, which is relevant given that health care interventions affect not only the health of the individual receiving the intervention but also the health of those close to the patient.

Many trials had methodological limitations introducing some risk of bias. Some trials did not clearly describe methods used for assessing utilization outcomes (e.g., readmissions, ER visits). Methods of handling missing data varied. Some trials did nothing to address missing data (i.e., analyzed only completers), but others conducted true intention-to-treat analyses and used appropriate methods of handling missing data. Some trials were limited in their ability to ascertain readmissions that were beyond a single institution.

Limitations also included inadequate sample sizes and heterogeneity of outcome measures across trials (specifically types of readmission rates). Reporting use of health services other than for the primary outcomes, such as ER visits, was variable across the included trials. This limited our ability to determine the effect of interventions on other measures of health services use (aside from readmission rates), such as an increase in ER visits or observational hospital stays.

Sometimes usual care and certain aspects of treatment interventions were not adequately described. Specifically, descriptions of whether (and how) interventions addressed medication management were often unsatisfactory. Categories of interventions that showed efficacy (e.g., the home-visiting and MDS-HF interventions) often included frequent visits with clinicians. Separating out individual components that are necessary from the overall type of interventions that showed efficacy was not possible. Moreover, some confounding components that were not described may be associated with efficacy as well, such as addressing social needs or optimizing HF pharmacotherapy.

Research Gaps

We identified important gaps in the evidence that future research could address; many are highlighted above. Of note, the gaps we highlight relate only to the key questions that we addressed in this systematic review; our suggestions to address them should not eliminate a wide range of potentially important research that falls outside the specified scope of this review. Table 22 summarizes the gaps and offers examples of potential future research that could help to close them.

Table 22

Evidence gaps for future research, by Key Question.

Also, we identified several methodological issues that increased the risk of bias for trials, and investigators could improve them in future research. Often trials only inadequately described methods of ascertaining outcomes relating to use of health care (e.g., readmissions, ER visits). Specifically, they did not say whether measurements were based on patient report, chart review, or some combination of measurements. Masking outcome assessments raised other concerns; for example, in some trials personnel delivering the intervention also appeared to be the primary staff for measuring health care utilization. Future trials should consider methods (such as blinded outcome assessments) that guard against measurement bias.

Conclusions

Few trials evaluating transitional care interventions for adults with HF reported 30-day readmission rates; one high-intensity home-visiting trial did reduce all-cause readmission and the composite endpoint over 30 days (low SOE). Strong evidence supported findings that three approaches improved at least one of our primary outcomes over 3 to 6 months: home-visiting programs, STS, and MDS-HF clinic interventions. Home-visiting programs reduced all-cause readmission, HF-specific readmission, the composite outcome, and mortality; STS interventions reduced HF-specific readmission and mortality; and MDS-HF clinic interventions reduced all-cause readmission and mortality. Based on current evidence, telemonitoring interventions and primarily educational interventions did not reduce readmission or mortality. Direct evidence was insufficient to conclude whether one type of intervention was better than any other type of intervention. Evidence was generally insufficient to determine whether the efficacy of interventions differed for subgroups of patients. The potential harms of transitional care interventions (e.g., higher caregiver burden) were generally not reported.