Article Text

Download PDFPDF

Does case management for patients with heart failure based in the community reduce unplanned hospital admissions? A systematic review and meta-analysis
  1. A L Huntley,
  2. R Johnson,
  3. A King,
  4. R W Morris,
  5. S Purdy
  1. Centre of Academic Primary Care, School of Social and Community Medicine, University of Bristol, Bristol, UK
  1. Correspondence to Dr AL Huntley; alyson.huntley{at}


Objectives The aim of this systematic review of randomised controlled trials (RCTs) and controlled trials (non-RCTs, NRCTs) is to investigate the effectiveness and related costs of case management (CM) for patients with heart failure (HF) predominantly based in the community in reducing unplanned readmissions and length of stay (LOS).

Setting CM initiated either while as an inpatient, or on discharge from acute care hospitals, or in the community and then continuing on in the community.

Participants Adults with a diagnosis of HF and resident in Organisation for Economic Co-operation and Development countries.

Intervention CM based on nurse coordinated multicomponent care which is applicable to the primary care-based health systems.

Primary and secondary outcomes Primary outcomes of interest were unplanned (re)admissions, LOS and any related cost data. Secondary outcomes were primary healthcare resources.

Results 22 studies were included: 17 RCTs and 5 NRCTs. 17 studies described hospital-initiated CM (n=4794) and 5 described community-initiated CM of HF (n=3832). Hospital-initiated CM reduced readmissions (rate ratio 0.74 (95% CI 0.60 to 0.92), p=0.008) and LOS (mean difference −1.28 days (95% CI −2.04 to −0.52), p=0.001) in favour of CM compared with usual care. 9 trials described cost data of which 6 reported no difference between CM and usual care. There were 4 studies of community-initiated CM versus usual care (2 RCTs and 2 NRCTs) with only the 2 NRCTs showing a reduction in admissions.

Conclusions Hospital-initiated CM can be successful in reducing unplanned hospital readmissions for HF and length of hospital stay for people with HF. 9 trials described cost data; no clear difference emerged between CM and usual care. There was limited evidence for community-initiated CM which suggested it does not reduce admission.

  • systematic review
  • meta-analysis
  • case management
  • hospital admission

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Strengths and limitations of this study

  • High-quality systematic review.

  • Interventions examine nurse-led multicomponent care of patients with heart failure.

  • Focus on use of resources specific to heart failure.

  • Community-initiated case management trials were limited in quantity and were mostly of low quality.

  • Lack of cost data in most trials.


Applying current prevalence figures to population estimates suggests that more than 550 000 individuals (more than 308 000 men and slightly fewer than 250 000 women) in the UK are living with heart failure (HF).1 Quality and Outcome Framework (QOF) data supports this: in 2012/2013, just over 480 000 patients were recorded as having HF.2 The average age of patients with HF in general practice in the UK is 77 years.3

Prior to 1990, 60–70% of patients died within 5 years of diagnosis, and admission to hospital with worsening symptoms was a regular and recurrent event.4–6 Effective treatment has improved care, with a relative reduction in hospitalisation in recent years of 30–50%, and smaller but significant decreases in mortality.4–6

More than £6.8 billion was spent on treating all cardiovascular disease within the National Health Service (NHS) in England in 2012/2013 with 63% of these costs coming from within secondary care and 21% within primary care. Within secondary care, non-elective inpatient admittance for cardiovascular disease, that is, emergency admissions, had the greatest expenditure with £1925 million.1

Case management (CM) is the process of planning, coordinating and reviewing the care of an individual. We used the definition cited by the King's Fund in the UK ‘A collaborative process of assessment, planning, facilitation, care coordination, evaluation, and advocacy for options and services to meet an individual's and family's comprehensive health needs through communication and available resources to promote quality cost-effective outcomes’.7 The NHS has used less-intensive approaches than the traditional US model, for example, through the use of nurses to support older people and those with long-term conditions at home.8 In this review, we have focused on CM based on nurse coordinated multicomponent care of patients which is applicable to the primary care-based health systems such as that in the UK.

A previous systematic review and meta-analysis shows that CM is not effective in reducing unplanned hospital admissions for the general older/elderly population.9 However, limited data suggest that CM for patients with HF is promising.10 This current review aimed to (1) identify the evidence of the effectiveness and related costs of CM interventions for patients with HF predominantly based in the community and (2) to better understand the potential success of CM by examining the components of tested interventions.



Databases and registries

A search strategy was developed using keywords for the electronic databases according to their specific subject headings or searching structure. The search strategy was run from 1985 to 2012 in the OVID databases—MEDLINE, Excerpta Medica Database (EMBASE), Cumulative Index to Nursing and Allied Health Literature (CINAHL) and PsycINFO on 2 July 2014 (see online supplementary appendix 1). The search strategy was modified to search internet sites such as the Agency for Healthcare Research and Quality (AHRQ) and the King's Fund. A pragmatic update of these searches was conducted on 20 November 2015 using the full search strategy and run in MEDLINE and MEDLINE in process only.

Other sources

Once the included papers were determined, both backwards (reference list of paper) and forwards citation searching (via Google Scholar) was performed to identify any other potentially relevant studies. All authors of included studies in the field were contacted with data queries and to identify additional relevant studies.

Eligibility criteria

Randomised controlled trials (RCTs) and other controlled studies (non-RCTs, NRCTs; controlled trials, controlled before and after studies, analytic cohorts, comparative studies) were included as determined by our eligibility criteria. We were aware from our previous work that not all community-based studies were randomised and felt it was important to be more inclusive in order to understand why CM may work for HF. CM interventions needed to be initiated either while as an inpatient or on discharge from acute care hospitals including the emergency department (ED), or in the community, and then continue on in the community. Only studies including adults with HF in Organisation for Economic Co-operation and Development (OECD) countries were included as the outcomes were more likely to be comparable for synthesis, and relevant to the UK situation.11 Studies were included as long as one of the outcomes of interest was unplanned hospital (re)admissions, ED attendance, length of hospital stay (LOS) as well as related costs of the interventions. Other outcomes of interest were primary healthcare resources, for example, general practitioner visits, visits to other primary care health professionals or services and prescriptions. Studies written in any language were considered if there was an English abstract available.

Reference management and study selection

EndNote and Excel were used to manage the references. Duplicates were removed from the EndNote file. References underwent a two-stage process of screening using the inclusion and exclusion criteria by two reviewers independently (ALH, AK, RJ). First, a screen of titles and abstracts (if abstract available) and second screening of the full paper was conducted. Where there was continued disagreement between reviewers about including or excluding a paper, a third reviewer made the final decision (SP or RJ).

In addition to the included quantitative intervention papers, we identified relevant reviews from the search. Any potentially relevant conference proceedings were followed up, first by searching in MEDLINE to see if the study had been published. If the study was not published, the authors were contacted where possible to check if the studies were likely to be published within the work frame of this review.

Data extraction and assessment of risk of bias

Data were extracted into a custom-designed table which included description of trial type, participants, intervention, controls, outcome measures and results. Based on the Kings Fund definition of CM, we devised taxonomy of intervention components8 (table 1). As part of this data extraction process, the intervention and control treatments were also described by their component parts, for example, monitoring signs and symptoms using the framework of the CM definition.

Table 1

Components of CM interventions

Quantitative data concerning the outcomes of interest were extracted into the Cochrane Revman software. The Cochrane risk of bias tool was used to record trial bias for RCTs and the Effective Practice and Organisation of Care (EPOC) risk of bias tool was used for NRCTs.12 These processes were performed by one author and checked by a second (ALH, AK). Any disagreements were resolved by discussion and if necessary with a third author (RJ, SP).

Data analysis

Trials were divided as previously described by Huntley et al9 into hospital-initiated CM and community-initiated CM. Where there were data from three or more studies, effect sizes were calculated and presented in forest plots as rate ratios ((re)admissions) or mean differences (LOS) using Revman software. If the heterogeneity of the combined data was >50%, a random-effects model was used for analysis.

We conducted prespecified sensitivity analysis in response to the risk of bias assessment of studies, removing high risk of bias studies as appropriate; the results of both analyses are presented.13 We conducted prespecified subgroup analysis to explore the effects of CM duration (3, 6 and 12 months plus) on hospital admission and LOS. There was insufficient detail in trials to perform subanalysis by severity of HF or intensity of intervention.

Data were assessed narratively in respect of the components of interventions using the CM definition cited above as guidance8 (table 1). In addition, where possible, post hoc subgroup analysis was conducted in Revman in which interventions with components of interest were compared with those that did not have these components.


The systematic review yielded 22 studies with data published over 32 papers of which 17 were RCTs and 5 were NRCTs, all published in the English language14–45 (figure 1). No relevant studies were identified in a pragmatic update using the full search strategy run in MEDLINE and MEDLINE in process only in November 2015. Seventeen of these studies described hospital-initiated CM (n=4794),14 ,15 ,17 ,18 ,20–24 ,26–28 ,31 ,32 and five described community-initiated CM of HF (n=3832).38 ,42–45 The PRISMA checklist was used to ensure the quality of our systematic review manuscript.

Figure 1

PRISMA flow diagram. CCT, controlled clinical trial; CHF, chronic heart failure; CM, case management; EOL, end of life; RCT, randomised controlled trial.

Profile of patients

The range of female participants in the trials was 1–58%, but the majority of trials had relatively even gender divide (table 2). Comorbidity and multimorbidity were common. Eight of the 22 trials gave no detail on ethnicity of participants; in four studies, the triallists used white/non-white and English-speaking/non-English-speaking categories. In the remaining 10 studies, a fuller profile was described. Twelve of the 22 trials were conducted in the USA and the ethnicity profile reflected that including Spanish speaking/Hispanic, American Indian, black, African-American, Asian and white participants.

Table 2

Study characteristics of intervention studies

The majority of trials described the severity of HF using New York Heart Association (NYHA) classification. Twelve of the trials gave a breakdown of numbers or percentages in the I–IV classes with some trials only giving numbers of participants for the III and IV classes. In these trials, the percentage range of III and IV class patients was 6–98%. Four trials gave mean and median values of NYHA status, one trial used the all-patient refined-diagnosis related group (APR-DRG) severity of illness scale, and five trials did not describe disease severity.

Profile of interventions

The majority of studies (n=15) described the intervention being delivered by a case manager/specialist nurse with no specific mention of other health professionals, and the remaining seven studies described a case manager/specialty nurse working as part of a multidisciplinary team (table 2).

Figure 2

(A) Risk of bias of included randomised controlled trials. (B) Effective Practice and Organisation of Care (EPOC) risk of bias for non-randomised controlled trials.

All but two studies compared CM with usual care although the control group was not always described. The two remaining studies were comparative: one RCT comparing CM with specialist clinics and one RCT comparing CM with telemedicine plus CM.32 ,42

The duration of the CM interventions in the studies was 1–24 months with the majority having a 3-month or 6-month duration. The majority of studies were conducted face to face or a combination of in-person and by phone. Four interventions were conducted purely by phone.21 ,22 ,26 ,42 Outcomes were measured to match the total duration of intervention in the majority of studies. For many of the studies, the intensity of interventions was not stated explicitly. When intensity was described, it was always a tapered approach after an initial intensive period.

Risk of bias

The degree of risk of bias was starkly different between the RCTs and NRCTs. All five of the NRCTs were rated at high risk or unknown risk for most domains (figure 2A, B).35–37 ,44 ,45

The majority of the RCTs were rated at low risk for most domains with the exception of the domain of blinding of the participants and personnel which is not applicable to this type of intervention. Three RCTs were assessed as at high risk for at least one domain: both Hancock et al43 and Wade et al42 gave no description of the randomisation process or allocation concealment, Riegel et al21 was randomised at physician level and patients were chosen by physician preference. Four of the five community-initiated trials (two RCTs and two NRCTs) were assessed to be at high risk of bias, and in some studies did not present usable data.35–37 ,42

All the intervention studies reported unplanned hospital (re)admissions14–45 and 17 reported length of time in hospital.14 ,15 ,17 ,18 ,20–22 ,24 ,26–28 ,35 ,38 ,42–45 There were few data on A and E attendance and primary care resource use. However, only some of the data could be used in meta-analysis with the main reasons being that data were presented in different formats where neither CIs, SEs nor raw data were given. Owing to heterogeneity of data, all analyses were conducted using a random-effects model.

Unplanned HF (re)admissions data

Hospital-initiated CM

Thirteen of the hospital-initiated CM trials had data that could be used in a meta-analysis of which 12 were RCTs. The pooled data from the RCTs showed a rate ratio of readmissions of 0.74 (95% CI 0.60 to 0.92; p=0.008; I2=69%) in favour of hospital-initiated CM (figure 3A). A sensitivity analysis was conducted, removing Riegel et al21 (RCT with high risk of bias for randomisation domain); this had a minimal effect on the rate ratio and heterogeneity (0.77 (0.61 to 0.96); p=0.02; I2=68%).35 Subanalysis looking at 3-month, 6-month and 12–18-month data did not produce a clear time-related effect which is most likely due to heterogeneity within and between studies (figure 3B–D). There was one hospital-initiated CM trial which compared CM with specialist clinics which reported no differences in hospital readmissions between the two groups.32

Figure 3

Chronic heart failure (CHF) admissions data. CM, case management.

Community-initiated CM

Of the four community-initiated trials (two RCTs and two NRCTs) comparing admissions between CM with usual care, two reported no significant differences38 ,43 and two reported statistically significant reductions in favour of CM.44 ,45 One further trial compared CM, with telehealth and CM and reported no differences in admissions but data were not presented.42

Length of hospital stay

Hospital-initiated CM

Nine of the hospital-initiated CM trials had data that could be used in a meta-analysis of which eight were RCTs. The pooled data from the RCTs showed that mean LOS was reduced in the CM group compared with usual care mean difference (MD −1.28 days (−2.04 to −0.52); p=0.001; I2=63%; figure 4A). A sensitivity analysis was conducted removing Riegel et al21 which had an important effect on the rate ratio and heterogeneity (MD −1.76 (−2.29 to −1.23); p<0.00001; I2=14%).21 ,35 Subanalysis looking at 3-month, 6-month and 12–18-month data suggests this effect is short term (first 3 months) but a longer time-related effect was difficult to assess due to lack of data (figure 4B–D).

Figure 4

Chronic heart failure (CHF) length of hospital stay. CM, case management.

The one study comparing CM with specialist clinic care reported that CM patients accumulated 592 and clinic patients 547 all-cause hospitalisations (p=0.087) associated with 3067 vs 4410 days of hospital stay (p<0.01 for rate and duration of hospital stay).32

Community-initiated CM

Of the four community-initiated trials comparing CM with usual care, two did not report LOS,38 ,43 one reported median values in favour of CM44 and the remaining one reported a mean reduction in LOS45 (table 2). The one comparative trial between community-initiated CM, and telehealth and CM did not report any useful data.42

Intervention components

Fourteen intervention components were identified and grouped as per the CM definition in methods and prevalence determined for hospital-initiated and community-initiated CM studies with a usual care control group (see tables 1 and 2 and online supplementary appendix 2).7

Hospital-initiated CM

Data from individual studies which contained components of family involvement showed an overall reduction in hospital readmissions in comparison with usual care and a reduction in hospital readmissions observationally in comparison with interventions which did not contain these components (rate ratio of 0.56 (0.34 to 0.92); p=0.003). However, post hoc analysis comparing these studies, in which the component was present with those studies in which the component was absent, did not yield any statistically significant differences (p=0.15; see online supplementary appendix 2a). The same calculations for medication review; referral to other services; and assessment of home environment, CM meetings and patient-directed access did not indicate any specific effect of these components of hospital-initiated CM on rates of admission (see online supplementary appendix 2b–g) The majority of the interventions included education/self-management, and there were insufficient data from studies without this component to allow comparison.

Community-initiated CM

There were insufficient data to conduct any subgroup analysis on any of the remaining components of hospital-initiated CM, community-initiated studies or the LOS data.

Outpatient healthcare resources

Only six of the included studies measured outpatient resource use. In some studies, outpatient resource data were all-cause and not HF-specific. In some studies, primary and secondary use was combined.23 ,24 ,35 ,38 ,42 ,45 Two of these studies also reported ED attendance.23 ,42 All but one of these studies reported no difference between intervention and control group for these measures with the exception of Lowery et al45 which showed a statistically significant greater use of outpatient resources in the usual care group (optional primary care visits 1 year 16.75 (13.62); 10.43 (9.6), p<0.001; 2 year 14.27 (11.98); 9.35 (9.97), p<0.001).


Nine of the 17 hospital-initiated trials described cost data (table 3). Of these, six reported no statistically significant difference between CM (3-month or 6-month duration) and usual care,17 ,18 ,22 ,24 ,26 ,35 and three reported costs in favour of CM although data from Stauffer et al37 was brief.15 ,32 One of these was 12–18 months32 and two were 3 months in duration. It was difficult from the intervention descriptions to determine their intensity. There were no cost data reported from the community-initiated trials.

Table 3

Available cost data from studies (n=9)


This systematic review confirms that hospital-initiated CM can be successful in reducing unplanned hospital readmissions, and reducing LOS in hospital in the short term for people with HF. There were only five community-initiated CM studies (three RCTs and two NRCTs) of which four were at high risk of bias. This limited evidence suggests no effect of community-initiated CM on hospital admissions. A minority of trials report cost comparisons with usual care and most of those show no difference. There were limited data on the effect of CM on other healthcare resources.

Many factors are likely to modify the effect of CM on use of emergency care seen in these studies. It is generally accepted that CM is more appropriate for people with severe HF and poorer general health. However, it was difficult to compare the health status of the study participants in hospital-initiated and community-initiated trials as in some studies there was little detail, others gave median and mean figures for NYHA status, and the presentation format and detail of comorbidities varied. All the included studies have been conducted within the past 12 years, so it is important to put these results in the context of overall improved treatment and reduction in hospital admissions since the early 1990s.4–6

Seventeen studies described hospital-initiated CM and five described community-initiated CM of HF, although often the participants were identified via hospital clinic records. Overall, the meta-analysis showed that CM reduced readmissions and hospital LOS. This may be explained by the fact that in most of the trials the participants were identified via hospital contact, and therefore were likely to have had a recent exacerbation of their HF and to be at increased risk of readmission in the postdischarge period. In addition, it is likely that interventions are acting at a time of highest risk as reflected by HF mortality in first year of diagnosis.4 Therefore, once they were assessed and given extra support, they were stable for a period of time. Previous work by Roland et al46 suggests that admission rates in people aged 65 with two or more emergency admissions in 12 months fall in subsequent years without any intervention and account for fewer than 10% of admissions in the following year, and thus effectiveness of admission avoidance schemes cannot be judged by tracking admission rates without careful comparison with a control group. The data from trials of community-initiated CM was lacking both in the number of studies, and the fact there were limited useable data that showed no effect on unplanned hospital admissions. It is likely that these patients were likely to be in more stable health.41

A metareview of a wide range of HF disease management programmes by Savard et al47 reports that nine previous systematic reviews (2001–2009) identified significant reductions in HF admissions with reductions in risk ranging from 30% to 56%. However, the authors caution that these reviews are limited by inadequate reporting in the population, setting, intervention and comparator components. They report that reviewers have not taken into account statistical, clinical and methodological heterogeneity in interventions.47 Our review focused specifically on CM avoiding some of these limitations and indicates a reduction in HF readmissions with hospital-initiated CM in the range of 10–30%.

Wakefield et al48 in 2013 looked at common components of a range of HF care programmes focusing mainly on disease management and education investigated in RCTs, and 10/35 of the discussed studies were included in our review. They described patient education, symptom management by health professionals and by patients, and medication adherence strategies as the most commonly occurring elements of care. A literature review by Jaarsma et al49 looked at 70 ‘home care’ controlled studies (mostly RCTs) which encompassed 9 of our included CM studies covering a wide range of approaches such as telemedicine, hospital at home and health buddies for patients with HF. They identified a multidisciplinary team, continuity of care, care plans, optimising titration of medication, education/counselling of patients and caregivers and increased access as important. Unfortunately, we had insufficient data to perform subanalysis on the component of education/self-management.

Previous systematic reviews have investigated the role of the lay caregiver in HF patient management.50–52 These suggest that better relationship quality and communication were related to reduced mortality, increased health status and less distress, and improved patient self-care outcomes. Our review adds to this evidence base by suggesting that more family involvement in CM may also reduce unscheduled readmissions.

Education about HF and about its pharmacological and non-pharmaceutical treatment has been well reviewed both as an individual approach and as part of complex interventions, and is considered to be essential for improving many patient outcomes.49 ,53 ,54 A recent mixed-method study suggests asking patients with HF to write down their learning needs before the education increases their chances of receiving education based on their individual needs.55 Qualitative interviews with health professionals caring for patients with HF suggest that communication with, and education by specialist nurses facilitated by continuity of care is essential to good care of patients with HF. The authors also highlight the role of the specialist nurse in multidisciplinary team communication and functioning, essentially describing the role of the specialist nurse as a case manager.56

Our review of CM suggests that the evidence for its cost-effectiveness is lacking with most studies that have performed cost comparisons with usual care show no advantage. Previous work by de Bruin et al57 looked at cost-effectiveness of disease management for a range of chronic conditions and concluded that the data are most positive for HF with five out of the eight included studies showing cost-effectiveness.

Strengths and limitations

The contribution of our high-quality systematic review to the above is that we have focused on CM which is based on nurse coordinated multicomponent care of patients which is applicable to the primary care-based health systems such as that in the UK. We have focused on HF (re)admissions and LOS as opposed to all-cause data which many of the previous reviews have used.

By examining the components of CM, we have a profile of the components most likely to lead to the success of CM of patients with HF in terms of reducing (re)admissions and hospital LOS. Our review has highlighted the potential importance of family involvement albeit in post hoc analysis.

The limitations of this review are that majority of the community-initiated CM studies were of low quality with the exception of one low risk of bias RCT, and provided limited evidence. While funnel plot analysis was not appropriate with our data, we acknowledge that there may be publication bias on this topic.58 This was counteracted by the fact that the hospital-initiated studies comprised of predominantly community-based CM. There is a lack of cost data and analysis in the included papers. This point needs to be emphasised for future trials. It is possible that cost-effectiveness will be more likely with intervention for patients with more severe HF.


Hospital-initiated CM reduces unplanned hospital admissions, and LOS for people with HF in the short term. Cost data are limited. There was limited evidence for community-initiated CM which suggested it does not reduce hospital admission. Further research is needed to determine the individual components of CM that contribute to reduced admissions.



  • Contributors ALH is the main systematic reviewer, and worked across all stages of the review from inception to completed draft. RJ is the cardiology and primary care expertise, and worked on screening, selection of studies, commenting on analysis, and development and checking of final document content. AK is the second reviewer, involved in screening, selection, data checking and commenting on developing and final document content. RWM is the statistical expertise, advising on data analysis and commenting on the developing and final document content. SP is the primary care and admission avoidance expertise, and advised throughout project, third reviewer for screening process and commenting on the developing and final document content.

  • Funding This project was funded by the National School of Primary Health Care project no.238.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Full data extraction tables and data analysis files are available on request.