Integrated disease management interventions for patients with chronic obstructive pulmonary disease

Annemarije L Kruis; Nynke Smidt; Willem JJ Assendelft; Jacobijn Gussekloo; Melinde RS Boland; Maureen Rutten‐van Mölken; Niels H Chavannes

doi:10.1002/14651858.CD009437.pub2

Integrated disease management interventions for patients with chronic obstructive pulmonary disease

Authors' declarations of interest

Version published: 10 October 2013 Version history

https://doi.org/10.1002/14651858.CD009437.pub2

Collapse all Expand all

Abstract

available in

Background

In people with chronic obstructive pulmonary disease (COPD) there is considerable variation in symptoms, limitations and well‐being, which often complicates medical care. To improve quality of life (QoL) and exercise tolerance, while reducing the number of exacerbations, a multidisciplinary program including different elements of care is needed.

Objectives

To evaluate the effects of integrated disease management (IDM) programs or interventions in people with COPD on health‐related QoL, exercise tolerance and number of exacerbations.

Search methods

We searched the Cochrane Airways Group Register of trials, CENTRAL, MEDLINE, EMBASE and CINAHL for potentially eligible studies (last searched 12 April 2012).

Selection criteria

Randomized controlled trials evaluating IDM programs for COPD compared with controls were included. Included interventions consisted of multidisciplinary (two or more health care providers) and multi‐treatment (two or more components) IDM programs with a duration of at least three months.

Data collection and analysis

Two review authors independently assessed trial quality and extracted data; if required, we contacted authors for additional data. We performed meta‐analyses using random‐effects modeling. We carried out sensitivity analysis for allocation concealment, blinding of outcome assessment, study design and intention‐to‐treat analysis.

Main results

A total of 26 trials involving 2997 people were included, with a follow‐up ranging from 3 to 24 months. Studies were conducted in 11 different countries. The mean age of the included participants was 68 years, 68% were male and the mean forced expiratory volume in one second (FEV1)% predicted value was 44.3% (range 28% to 66%). Participants were treated in all types of healthcare settings: primary (n = 8), secondary (n = 12), tertiary care (n = 1), and in both primary and secondary care (n = 5). Overall, the studies were of high to moderate methodological quality.

Compared with controls, IDM showed a statistically and clinically significant improvement in disease‐specific QoL on all domains of the Chronic Respiratory Questionnaire after 12 months: dyspnea (mean difference (MD) 1.02; 95% confidence interval (CI) 0.67 to 1.36); fatigue (MD 0.82; 95% CI 0.46 to 1.17); emotional (MD 0.61; 95% CI 0.26 to 0.95) and mastery (MD 0.75; 95% CI 0.38 to 1.12). The St. George's Respiratory Questionnaire (SGRQ) for QoL reached the clinically relevant difference of four units only for the impact domain (MD ‐4.04; 95% CI ‐5.96 to ‐2.11, P < 0.0001). IDM showed a significantly improved disease‐specific QoL on the activity domain of the SGRQ: MD ‐2.70 (95% CI ‐4.84 to ‐0.55, P = 0.01). There was no significant difference on the symptom domain of the SGRQ: MD ‐2.39 (95% CI ‐5.31 to 0.53, P = 0.11). According to the GRADE approach, quality of evidence on the SGRQ was scored as high quality, and on the CRQ as moderate quality evidence. Participants treated with an IDM program had a clinically relevant improvement in six‐minute walking distance of 43.86 meters compared with controls after 12 months (95% CI 21.83 to 65.89; P < 0.001, moderate quality). There was a reduction in the number of participants with one or more hospital admissions over three to 12 months from 27 per 100 participants in the control group to 20 (95% CI 15 to 27) per 100 participants in the IDM group (OR 0.68; 95% CI 0.47 to 0.99, P = 0.04; number needed to treat = 15). Hospitalization days were significantly lower in the IDM group compared with controls after 12 months (MD ‐3.78 days; 95% CI ‐5.90 to ‐1.67, P < 0.001). Admissions and hospital days were graded as high quality evidence. No adverse effects were reported in the intervention group. No difference between groups was found on mortality (OR 0.96; 95%CI 0.52 to 1.74). There was insufficient evidence to refute or confirm the long term effectiveness of IDM.

Authors' conclusions

In these COPD participants, IDM not only improved disease‐specific QoL and exercise capacity, but also reduced hospital admissions and hospital days per person.

PICOs

Population

Intervention

Comparison

Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Plain language summary

available in

Integrated disease management for chronic obstructive pulmonary disease

Background

Chronic obstructive pulmonary disease (COPD) is a chronic respiratory (lung), disabling disease which affects a lot of people worldwide and causes millions of deaths every year. People with COPD suffer differing levels of impairment, daily complaints/symptoms and number of exacerbations.

Different health care providers, such as doctors, nurses and physiotherapists, typically provide different components of care (for example medication, self management and education, exercise training) to people with COPD. The aim of an integrated disease management (IDM) program is to establish a program of different components of care in which different health care providers are co‐operating and collaborating to provide efficient and good quality care.

Review question

We wished to determine the effect of such a program on quality of life, exercise tolerance and the number of exacerbations. We have chosen these outcomes as they are most important for people with COPD.

What we found

We evaluated 26 studies in 2997 people with COPD. Overall the evidence found was of high to moderate quality. The trials were conducted in 11 different countries. The average age of participants was 68 years, 68% of participants were men and the severity of COPD on average was severe (according to lung function measures). Some of the trials took place in GP clinics and some in hospitals. Overall, the studies were of good to moderate methodological quality.

People who participated in an IDM program had better quality of life and improved their exercise tolerance after 12 months. Furthermore, in participants treated with such a program, the number of hospital admissions related to exacerbations decreased and the total number of hospital days was reduced by three days. We found no evidence of an effect on mortality.

The results support an IDM program for people with COPD to optimize quality of life and exercise tolerance.

This plain language summary is up‐to‐date as of April 2012.

Authors' conclusions

Implications for practice

This meta‐analysis provides evidence for the efficacy of integrated disease management (IDM) programs of at least three months duration for chronic obstructive pulmonary disease (COPD) patients, for up to 12 months follow‐up. We found positive effects on disease‐specific quality of life and exercise capacity in studies containing an exercise program, suggesting that exercise training is an important element in an IDM program. Long‐term effects are still unclear, as only a few studies evaluated these. The magnitude of improvement in disease‐specific quality of life was clinically relevant, especially using the Chronic Respiratory Questionnaire (CRQ).

We calculated that seven hospital admissions related to respiratory problems can be prevented for every 100 patients treated with IDM for three to 12 months, giving to a number needed to treat of 15 patients to prevent one being admitted. Furthermore, hospitalisation decreased by three days in patients treated with IDM compared to controls. This is of utmost importance, as hospitalizations contribute to the highest burden and costs in patients with COPD. The effects of IDM on the total number of patients suffering at least one exacerbation still remain unclear. It is possible that patients who have learned from education and have an action plan may recognize exacerbations at an early stage and can start medical treatment directly. It is therefore likely that further worsening of health status and hospital admissions can be prevented in these patients.

Implications for research

The following issues could be assessed if authors are planning future trials regarding the effectiveness of IDM:

Study quality: Overall, studies included in this systematic review were of moderate quality, as not all aspects of risk of bias were appropriately addressed. Therefore, there is a need for future trials to report a proper description of the processes of randomisation and data collection. Preferentially, a study protocol including measured outcomes should be published in advance to minimize selection and reporting bias.
More detailed description of intervention: A detailed description of the precise nature of the intervention is important, in order to be able to determine in the future which components, duration and intensity of a program are most effective. Ideally, we wish to determine which combination of health care providers and which components are most effective in IDM programs.
Consensus on reporting common outcomes: Given the huge variation in outcome measures and follow‐up time points, we strongly recommend consensus on the reporting of common outcomes, such as change from baseline in health‐related quality of life, in order to be able to combine more results in future meta‐analyses. We advise future trial authors to measure at least one of the following outcomes: quality of life, exercise tolerance or exacerbation‐related outcomes.
Adequate power calculation and methods of analysis: two cluster‐randomized controlled trials introduced noteworthy bias due to inadequate methods of analysis, not taking the clustering into account (Rea 2004; Wood‐Baker 2006) and loss to follow‐up of clusters (Rea 2004). Therefore, we recommend performing a proper power calculation beforehand and, if needed, adjusting this calculation for intra‐cluster effects (Guyatt 2011; Higgins 2011).

Finally, given the heterogeneity of interventions, there is a need to reach consensus on which interventions are likely to yield the best results when applying integrated care programs for COPD.

Summary of findings

Open in table viewer

Summary of findings for the main comparison. Integrated disease management compared to control for patients with chronic obstructive pulmonary disease

Integrated disease management compared to control for patients with chronic obstructive pulmonary disease
Patient or population: patients with chronic obstructive pulmonary disease Settings: 8 studies in primary care, 12 studies in secondary care, 1 study in tertiary care, 5 studies in both primary and secondary care Intervention: integrated disease management Comparison: control
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Disease management
Quality of life measured on the SGRQ (St George's Respiratory Questionnaire) total score. Scale from: 0 to 100. Lower score indicates improvement Follow‐up: 3 to 12 months	The mean change in the SGRQ (total score) ranged from 3.4 lower to 6.24 higher	The mean SGRQ (total score) in the intervention groups was 3.71 lower (5.83 to 1.59 lower)	MD ‐3.71 (‐5.83 to ‐1.59)	1425 (13 studies)	⊕⊕⊕⊕ high²	MCID = ‐4 points, lower score means improvement
Quality of life measured on the CRQ dyspnoea domain Scale from: 0 to 7. Higher score indicates improvement Follow‐up: 3 to 12 months	The mean change in the CRQ (dyspnoea domain) ranged from 0 to 0.2 lower	The mean CRQ dyspnoea domain in the intervention groups was 1.02 higher (0.67 to 1.36 higher)	MD 1.02 (0.67 to 1.36)	160 (4 studies)	⊕⊕⊕⊝ moderate¹	MCID = 0.5 points Results on the other domains of the CRQ (fatigue, emotion, mastery) were also all statistically and clinically relevant
Functional exercise capacity 6‐minute walking distance (6MWD) Follow‐up: 3 to 12 months	The mean change in the 6MWD ranged from 38 lower to 36 higher	The mean functional exercise capacity in the intervention groups was 43.86 higher (21.83 to 65.89 higher)	MD 43.86 (21.83 to 65.89)	838 (14 studies)	⊕⊕⊕⊝ moderate³	MCID = 35 meters. Sensitivity analysis did show there was inconsistency in the effect. After removing low‐quality studies, the MD was 15.15 meters (95% CI 6.37 to 23.93, P < 0.001)
Respiratory‐related hospital admissions Follow‐up: 3 to 12 months	27 per 100	20 per 100 (15 to 27)	OR 0.68 (0.47 to 0.99)	1470 (7 studies)	⊕⊕⊕⊕ high
Hospital days per patient (all causes) Follow‐up: 3 to 12 months	The mean change in hospital days ranged from 1.6 to 11.9 higher	The mean number of hospital days per patient (all causes) in the intervention groups was 3.78 lower (5.9 to 1.67 lower)	MD ‐3.78 (‐5.9 to ‐1.67)	741 (6 studies)	⊕⊕⊕⊕ high
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; IDM: integrated disease management; MCID: minimal clinically important difference; MD: mean difference; OR: odds ratio
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹We downgraded one as there was considerable risk of bias in two studies on allocation concealment and two studies did not blind the outcome assessor. ²We did not downgrade due to risk of bias, as studies contributing more than 2.7% to the meta‐analysis had a low risk of bias. Sensitivity analysis on high‐risk studies did not change the effect or significance of the effect. ³We downgraded one as all included studies were of moderate to low quality. If we removed studies which had high or unclear risk of bias on allocation concealment, the effect decreased to 15 meters.

Background

Description of the condition

Chronic obstructive pulmonary disease (COPD) is a heterogeneous, systemic condition characterized by restricted airflow which is not fully reversible. It is a major cause of morbidity, due to the ageing of the world's population and the continued use of tobacco and exposure to indoor biomass pollution. The prevalence of COPD is expected to increase substantially in the coming decades (Lopez 2006; GOLD 2009). According to the World Health Organization (WHO), COPD will be the third leading cause of death in 2020 (Lopez 2006; WHO 2008). Given the rise in prevalence, COPD has important financial consequences, with high reported direct costs (healthcare resources, medication prescriptions) and indirect costs (absence from paid work, consequences of disability) (Britton 2003).

Optimal management of COPD is complex, as it is a multi‐component disease. Clinical, functional and radiological presentation varies greatly from patient to patient, despite having a similar degree of airflow limitation (Wedzicha 2000; GOLD 2009; Agusti 2010). Evidence suggests that the previous 2007 Global Initiative for Chronic Obstructive Lung Disease (GOLD) classification of disease severity, solely based upon the degree of airflow limitation, is a poor predictor of other important negative features of COPD (Agusti 2010; Burgel 2010).

Health‐related quality of life (HRQoL) and exercise tolerance may be more important to people with COPD than the more traditional measure of lung function. This is because COPD has a profound impact on HRQoL and exercise tolerance, even in those with modest airflow limitation (Engstrom 1996). Furthermore, impaired HRQoL (Domingo‐Salvany 2002; Fan 2002; Martinez 2006) and exercise tolerance (Gerardi 1996; Pinto‐Plata 2004) have been associated with an increased risk of mortality (Cote 2009).

In addition, some people are more prone than others to episodes of acute exacerbations, which are an important cause of morbidity, mortality, hospital admission and impaired health status (Seemungal 1998; Wedzicha 2000; Calverley 2003). Although exacerbations become more severe and occur more frequently with increased severity of COPD, this is not always the case. There is some evidence for a 'frequent‐exacerbation' phenotype (or group of people) that exacerbate more often than would be expected given their 'severity' as predicted by lung function testing (Hurst 2010).

Episodes of exacerbations are often not reported by patients to health care providers (Seemungal 2000). An important reason for patients' delay in reporting an increase in symptoms to their doctor is the fear of being sent to hospital. This passive behavior can eventually lead to a respiratory crisis, indeed necessitating urgent referral. In order to break through the self reinforcing negative spiral of dyspnoea, deconditioning and social deprivation doctors need to collaborate with their patients, with a focus on self management skills: "if symptoms increase, you need to let us know rapidly to prevent further worsening" (Chavannes 2008). In viewing COPD as a disease process with a clinical, heterogeneous picture of progressive deterioration, an integrated system of care could be built on a disease management model. Ideally, it is based on active self management to slow down progression of the disease, including daily self care, patient‐physician collaboration and exacerbation management. Information should be tailored to the person's needs, knowledge level and clinical profile and be accessible by the patient when they need it most (Tiep 1997; Bourbeau 2013).

Description of the intervention

In the last decade, the concept of integrated disease management (IDM) was introduced as a mean of improving quality and efficiency of care. IDM interventions are aimed at reducing symptoms and avoiding fragmentation of care, while containing costs. Therefore, IDM programs are generally believed to be cost‐effective, but the available evidence is inconclusive. Several systematic reviews have shown positive results, at least for some outcomes of chronic IDM, in people with chronic heart failure (Gonseth 2004; Roccaforte 2005), diabetes (Norris 2002; Knight 2005; Pimouguet 2010) and depression (Badamgarav 2003; Neumeyer‐Gromen 2004).

However, there is no consensus in the literature about the definition of IDM. Several definitions have been proposed since the introduction of the concept 'disease management'. In order to facilitate the communication between researchers, policy makers and IDM program leaders, Schrijvers proposed a definition, based on earlier reported definitions (Care Continuum Alliance; Dellby 1996; Epstein 1996; Ellrodt 1997; Zitter 1997; Weingarten 2002; Faxon 2004): "Disease management consists of a group of coherent interventions designed to prevent or manage one or more chronic conditions using a systematic, multidisciplinary approach and potentially employing multiple treatment modalities. The goal of chronic disease management is to identify persons at risk for one or more chronic conditions, to promote self‐management by patients and to address the illness or conditions with maximum clinical outcome, effectiveness and efficiency regardless of treatment setting(s) or typical reimbursement patterns" (Schrijvers 2009). In addition, Peytremann‐Bridevaux and Burnand added more elements, adapting the definition as follows: "Chronic disease prevention and management consists of a group of coherent interventions, designed to prevent or manage one or more chronic conditions using a community wide, systematic and structured multidisciplinary approach potentially employing multiple treatment modalities. The goal of chronic disease prevention and management is to identify persons with one or more chronic conditions, to promote self‐management by patients and to address the illness or conditions according to disease severity and patient needs and based on the best available evidence, maximizing clinical effectiveness and efficiency regardless of treatment setting(s) or typical reimbursement patterns. Routine process and outcome measurements should allow feedback to all those involved, as well as to adapt the programme" (Peytremann‐Bridevaux 2009).

How the intervention might work

There is great variation in the symptoms, functional limitations and degrees of psychological well‐being of COPD patients, as well as the speed of the progression of COPD towards more severe stages (Agusti 2010). This calls for a multi‐faceted response, including different elements (e.g. smoking cessation, physiotherapeutic reactivation, self management, optimal medication adherence) targeted at the patient, professional or organizational level. Therefore, IDM programs have been developed to improve effectiveness and economic efficiency of chronic care delivery (Norris 2003) by combining patient‐related, professional‐directed and organizational interventions (Wagner 2001; Lemmens 2009).

Why it is important to do this review

As health‐related quality of life, exercise tolerance and number of exacerbations are the most important patient‐related outcomes in COPD, the focus in this review will be on these primary outcomes.

Several systematic reviews have been published that evaluated the effect of IDM in COPD patients (Adams 2007; Niesink 2007; Peytremann‐Bridevaux 2008; Lemmens 2009; Steuten 2009). These reviews differ from our review in various ways. Adams' review focused solely on interventions which could be arranged according to the chronic care model of Wagner (Wagner 1996; Adams 2007). Furthermore, Adams included studies between 1966 and 2005. Since then, several studies focusing on IDM in COPD patients have been published. Niesink and colleagues evaluated the quality of life in COPD patients, but did not report outcomes of exacerbations or exercise tolerance. Furthermore, the authors decided not to perform a meta‐analysis; reasons for this were not clearly described (Niesink 2007). Peytremann‐Bridevaux performed a meta‐analysis and focused on quality of life, exacerbations and exercise tolerance. However, they did not take into account the differences in study design (randomised controlled trials (RCT) versus before/after uncontrolled studies) in their conclusions (Peytremann‐Bridevaux 2008). Lemmens' review examined the effectiveness of IDM in a mix of patients with COPD, asthma or both (Lemmens 2009). No subgroup analysis was performed for patients with COPD. Furthermore, conclusions were drawn irrespective of the study designs (i.e. RCTs, controlled clinical trials, quasi‐experimental, controlled before and after time studies and time series designs; Lemmens 2009). Steuten et al aimed to determine the cost‐effectiveness of COPD programs and the authors did not perform a meta‐analysis of clinical effects (Steuten 2009).

Overall, all reviews suggested some beneficial effects on health status. However, firm conclusions could not be made regarding the effectiveness of IDM, due to the large heterogeneity in the interventions, study populations, outcome measurements and methodological quality. The literature searches of the aforementioned reviews for relevant RCTs investigating the effectiveness of IDM for patients with COPD were carried out between December 2006 and May 2008. Since then, several studies have been published. Furthermore, none of the former published systematic reviews were carried out according to the latest methods for conducting a systematic review (Higgins 2011). Within the framework of The Cochrane Collaboration, we have systematically and comprehensively evaluated the effectiveness of IDM in people with COPD.

Objectives

To evaluate the effectiveness of IDM programs or interventions in people with COPD on health‐related quality of life, exercise tolerance and the number of exacerbations.

Methods

Criteria for considering studies for this review

Types of studies

We included only randomised controlled trials (RCTs) in which IDM programs or interventions were compared to controls in people with COPD. Cluster‐randomized trials were also eligible. There were no restrictions regarding the language of the paper.

Types of participants

People with a clinical diagnosis of COPD according to the GOLD criteria were included: people having chronic respiratory symptoms (i.e. coughing, sputum or dyspnoea) and a limited post‐bronchodilator forced expiratory volume in one second (FEV1) to forced vital capacity (FVC) ratio of < 0.7. Severity of airflow obstruction was classified using the GOLD stages of 2009 (GOLD 2009). All GOLD stages were accepted. Studies including participants with other diagnoses than COPD were only eligible if the results of participants with COPD were available separately.

Types of interventions

We included studies where the IDM intervention consisted of strategies to improve the care for participants with COPD, including organizational, professional, patient‐directed and financial interventions. We classified these according to the Cochrane Effective Practice and Organization of Care Group (EPOC) taxonomy of interventions (EPOC 2008), complemented with patient‐directed interventions (i.e. self management and education). Our definitive checklist consisted of the following components of the IDM intervention that could be scored:

Education/self management: i.e. education, self‐management, personal goals and/or action plan, exacerbation management
Exercise: i.e. (home) exercise training and/or strength and/or endurance training
Psychosocial: cognitive behavioral therapy, stress management, other psychological assessment and/or treatment
Smoking cessation
Medication: optimal medication/prescription of medication adherence
Nutrition: dietary intervention
Follow‐up and/or communication: structural follow‐up and/or communication, case management by nurses, optimal diagnosis
Multidisciplinary team: active participation and formation of teams of professional caregivers from different disciplines, revision of professional roles, integration of services, local team meetings
Financial intervention: fees/payment/grants for providing IDM.

As IDM includes different components mentioned above, delivered by different healthcare disciplines, the RCT studies had to include:

at least two components of interventions as mentioned above;
active involvement of at least two different categories of healthcare providers; and
a minimum duration of the IDM intervention of three months.

In all studies, we determined the dominant component of the program.

We compared IDM versus controls (varying from usual care or no treatment to single interventions, mono‐disciplinary interventions).

Types of outcome measures

Primary outcomes

Health‐related quality of life (HRQoL), as reported by one of the following questionnaires: a validated disease‐specific questionnaire, e.g. Clinical COPD Questionnaire (CCQ; van der Molen 2003; Kocks 2006), Chronic Respiratory Questionnaire (CRQ; Guyatt 1987), St. George's Respiratory Questionnaire (SGRQ; Jones 1991; Jones 2005), COPD Assessment Test (CAT; Jones 2009) or a generic questionnaire, e.g. Short Form‐36 (SF‐36; Ware 1992), Euro Qol‐5D (EQ‐5D; EuroQol Group 1990)).
Maximal or functional exercise capacity, as reported by one of the following outcomes: the peak capacity measured in the exercise laboratory using an incremental exercise test defined according to the results of timed walk tests e.g. 6‐ or 12‐minute walk test (Redelmeier 1997) or shuttle run test (Singh 1992)).
Exacerbation‐related outcomes, as reported by one of the following: time to first exacerbation, number of exacerbations, duration and/or severity, and measured by reporting of symptoms, antibiotics or prednisolone prescriptions and/or hospital admissions or hospital days related to exacerbations.

Secondary outcomes

Clinical outcomes

Dyspnea, as measured by the Medical Research Council (MRC) Dyspnea Scale (Bestall 1999) or Borg score (Borg 1970).
Survival (mortality).
Lung function (FEV1, FVC).
Depression, as measured by the Hospital Anxiety and Depression Scale (HADS) (Zigmond 1983) or the Beck Depression Inventory (BDI) score (Beck 1961).

Process‐related outcomes

Co‐ordination of care, e.g. accessibility of care, participation rate in the disease management program, satisfaction of health care providers and participants with regard to the program, or the extent to which disease management was implemented, from the perspective of the patient (PACIC; Glasgow 2005) and the caregiver (Bonomi 2002).

We evaluated outcomes at the following endpoints: a) short‐term (12 months or less); b) long‐term (longer than 12 months) follow‐up, if possible.

Search methods for identification of studies

Electronic searches

We identified trials using the Cochrane Airways Group Register of trials, the Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library, MEDLINE, EMBASE and CINAHL. The search was performed without language restrictions, using the highly sensitive Cochrane Collaboration search strategy, which aims to identify all randomised controlled trials (Lefebvre 2009). We used specific MeSH headings and additional keywords to identify all RCTs on IDM in COPD patients. As IDM programs were first described in 1990, our search was restricted to publications from 1990 onwards. The complete search strategies for the database searches are provided in the appendices (MEDLINE Appendix 1; EMBASE Appendix 2; CINAHL Appendix 3; CENTRAL Appendix 4; Airways Register Appendix 5). The search has been conducted up to April 2012. We ran an update search on 12 April 2013, but the results have not been fully incorporated: nine studies have been added as 'ongoing studies' and three studies have been added as 'studies awaiting classification'.

Searching other resources

In order to identify all possible studies, we carried out an additional search for systematic reviews in the Cochrane Database of Systematic Reviews. We screened reference lists of included RCTs and systematic reviews for potential studies for this review. To identify ongoing or new studies, we searched databases of ongoing studies, including ClinicalTrials.gov and other relevant registers.

Data collection and analysis

Selection of studies

Two review authors (AK and NS) independently assessed the title and abstract of all identified citations. We excluded all trials that were not randomised controlled trials or in which participants had no diagnosis of COPD. All studies excluded by the first two review authors because of the nature of the intervention were double‐checked by a third review author (NC). Furthermore, if there was any doubt, we retrieved the full‐text article and examined it for inclusion eligibility. Disagreements were discussed in a consensus meeting.

Data extraction and management

We collected the following information from included studies in our review: 1) the study design (i.e. randomisation method, sample size, blinding); 2) participant characteristics (i.e. diagnosis COPD according to GOLD criteria, age, sex); 3) interventions (i.e. setting, number of professionals involved, elements of IDM program/intervention, frequency and duration of intervention); 4) outcome measures and timing of outcome assessment; 5) results (i.e. loss to follow‐up, outcomes). The outcome data were extracted by one author (AK) and checked by another (NC) using a standardized data extraction form. In case of missing data, we contacted the authors of these studies for additional information or clarification.

Assessment of risk of bias in included studies

Two of us (AK and NC) independently assessed risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), according to the following items:

Allocation sequence generation
Concealment of allocation
Blinding of participants and health care providers, in relation to the intervention
Blinding of outcome assessment
Incomplete outcome data
Selective outcome reporting

As cluster‐randomized trials were also considered for inclusion, we added the following design‐related criteria for these types of studies:

Recruitment bias (i.e. individuals are recruited after the clusters have been randomised)
Baseline imbalance between groups (i.e. the risk of baseline differences can be reduced by using stratified or pair‐matched randomisation of clusters)
Loss of follow‐up of clusters (i.e. missing clusters and missing outcomes for individuals within clusters may lead to a risk of bias in cluster‐randomized trials)
Methods of analysis adequate for cluster‐randomized controlled trials (i.e. taking clustering into account in the analysis) (Higgins 2011)

We judged all items as high, low or unclear risk of bias. We resolved disagreements in a consensus meeting.

Measures of treatment effect

We analyzed the results of the studies using RevMan 5, using random‐effects modeling. We used forest plots to compare results across trials. The results were related to the minimal clinically important difference (MCID).

We expressed the results of each RCT as risk ratios (RR) with corresponding 95% confidence intervals (95% CI) for dichotomous data, and mean difference (MD) or standardized mean difference (SMD) for continuous data, depending on the similarity of outcome measurement scale (i.e. MDs are used when all studies use the same outcome measurement scale and SMDs when studies use different outcome measurement scales). We summarized data in a meta‐analysis only if the data are clinically and statistically sufficiently homogenous. If the meta‐analysis led to statistically significant overall estimates, we transformed these results (pooled estimate of RR, MD or SMD) back into measures which are clinically useful in daily practice. We planned to use the number needed to treat for an additional beneficial outcome (NNTB) and the absolute and/or relative improvement on the original units in order to report these as the final results of the review.

Unit of analysis issues

In case of a unit of analysis error occurrence in cluster‐randomized controlled trials, we adjusted for the design effect by reducing the size of the trial to its "effective sample size" (Rao 1992). The effective sample size of a single intervention group in a cluster‐randomized trial is its original sample size divided by a quantity called the 'design effect'. The design effect is 1+ (M‐1)* ICC, where M is the average cluster size and ICC is the intra‐cluster correlation coefficient. For dichotomous data, both the number of participants and the number experiencing the event were divided by the design effect. For continuous data, only the sample sizes were reduced; means and standard deviations remained unchanged (Higgins 2011).

Dealing with missing data

In case of missing data, we planned to contact the authors for additional information about the missing data for individuals. We sent a reminder if we did not receive a response. Secondly, we planned to assume the missing values to have a poor outcome. For continuous outcomes (i.e. health‐related quality of life, exercise capacity) and dichotomous outcomes (i.e. mortality), we planned to calculate the effect size (SMD, MD, RR) based on the number of participants analyzed at the time point. If the number of participants analyzed is not reported for each time point, we planned to use the number of randomised participants in each group at baseline. We planned to perform sensitivity analysis to investigate whether our assumptions have been reasonable (i.e. comparing results using number of participants analyzed with number of participants randomised).

Assessment of heterogeneity

We measured clinical and statistical heterogeneity using the I² statistic (Higgins 2011). A P value of less than 0.10 or an I² value greater than 50% indicates substantial heterogeneity. In case of heterogeneity, we assessed studies, if possible, with respect to:

control group: a) no treatment; b) treatment with one health care provider; c) treatment with one component; d) other disease management programs (short duration of therapies);
intervention group, with regard to a) type of health care providers (i.e. general practitioner, lung specialist, physiotherapist, practice nurse); b) different components as listed by the EPOC classification (EPOC 2008); c) frequency and duration of intervention.

In case of substantial heterogeneity, we explored the data further, including subgroup analyses (see Subgroup analysis and investigation of heterogeneity) in an attempt to explain the heterogeneity.

Assessment of reporting biases

In order to determine whether reporting bias was present, we evaluated whether the protocol for the RCT was published before recruitment of patients of the study was started. For studies published after 1 July 2005, we screened the Clinical Trial Register at the International Clinical Trials Registry Platform of the World Health Organization (http://apps.who.int/trialsearch) (De Angelis 2004). For each study, we evaluated whether selective reporting of outcomes was present (outcome reporting bias). Furthermore, we made a funnel plot to assess the possibility of reporting bias.

Data synthesis

We pooled results of the studies using the random‐effects model. For continuous data, we recorded the mean change from baseline to endpoint and standard deviation (SD) for each group. For dichotomous data we recorded the number of participants with each outcome event and calculated the odds ratio (OR). We used results reported at three months, as our predetermined inclusion criteria postulated a program of at least three months duration (to ensure sufficient impact). If data at three months were unavailable, we analyzed the data measured most closely to this time point. We evaluated outcomes at short‐ (3 to 12 months) and long‐term (> 12 months) follow‐up.

We presented the main results of the review in a 'Summary of findings' table, which includes an overall grading of the evidence using the GRADE approach in accordance with the recommendations laid out in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). This involves making separate ratings for quality of evidence for each patient‐important outcome and identifies five factors that can lower the quality of evidence, including: study limitations; indirectness of evidence (also called clinical heterogeneity with regard to study population, intervention, control group and outcomes); unexplained heterogeneity or inconsistency of results (i.e. statistical heterogeneity); imprecision of results (i.e. due to small sample sizes and few events); and high probability of publication bias. However, other factors can increase the quality of evidence, such as large magnitude of effect; plausible confounding, which could reduce the demonstrated effect; and dose‐response gradient (GRADE Working Group 2004). We presented the short‐ and long‐term outcomes for our primary outcomes in the 'Summary of findings' table if possible.

Subgroup analysis and investigation of heterogeneity

In order to explain heterogeneity between the results of the included studies, we planned the following subgroup analyses a priori (where data were available) to determine if outcomes differed among:

patients with different severity of disease, according to GOLD stage (GOLD 2009) or MRC Dyspnea Scale (Bestall 1999) (e.g. patients with GOLD 1/2 versus GOLD 3/4, and/or patients with a MRC score 0 to 2 versus MRC 3 to 5);
the setting of the IDM intervention (e.g. primary, secondary or tertiary care);
design of the studies (individually randomised patients versus cluster‐randomized patients (with and without adjusting for design effect));
control group: a) no treatment; b) treatment with one health care provider; c) treatment with one component; d) other disease management interventions (short duration of therapies);
intervention group, with regard to a) type of health care provider (i.e. general practitioner, lung specialist, physiotherapist, practice nurse); b) different components as listed by the EPOC classification (EPOC 2008); c) frequency and duration of intervention.

Sensitivity analysis

We carried out sensitivity analyses for the primary outcome measurements, in order to explore effect size differences and the robustness of conclusions. We planned sensitivity analysis determined a priori based on:

studies without study limitations with regard to a) allocation concealment; b) blinding of participants and investigators; c) recruitment bias; d) baseline imbalance between groups; e) loss of follow‐up of clusters; f) adequate analysis;
method of analysis: a) results of studies using number of patients analyzed; b) studies using number of patients randomised.

We presented the main results of the review in a 'Summary of findings' table, which includes an overall grading of the evidence using the GRADE approach (GRADEpro; GRADE Working Group 2004) and a summary of the available data on the main outcomes, as described in Chapter 11 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Results

Description of studies

See Characteristics of included studies.

Results of the search

Our literature search identified 6700 titles and abstracts, resulting in 4776 references after de‐duplication. Two review authors (AK, NS) screened the title/abstracts of these studies based on the predetermined inclusion criteria. Studies that were excluded because of the IDM intervention were double‐checked by a third review author (NC). We retrieved the full‐text articles of these studies and they were discussed in a consensus meeting. Finally, we identified 49 potentially relevant articles about IDM in COPD patients. We obtained full‐text versions of these papers and data were extracted by one review author (AK) and double‐checked by a second review author (NC). Finally, a total of 26 (cluster) randomised controlled trials were included in the review. The PRISMA flow diagram is presented in Figure 1.

Figure 1

Study flow diagram.

Included studies

Characteristics of the included studies are described in Table 1, Table 2 and Characteristics of included studies.

Open in table viewer

Table 1. Characteristics of included studies

Study	Country	N (randomised)	N (completed)	Number of components intervention	Number of health care providers	Main component intervention	Setting	Control group
Aiken 2006	US	41	18	5	2	SF	PRIM	U
Bendstrup 1997	Denmark	42	32	4	7	E	SEC	U
Bourbeau 2003	Canada	191	165	4	4	SM	SEC	U
Boxall 2005	Australia	60	46	2	3	E	PRIM	U
Cambach 1997	Netherlands	43	23	2	2	E	PRIM	DRUG
Dheda 2004	UK	33	25	4	2	SF	SEC	U
Engstrom 1999	Sweden	55	50	4	5	E	SEC	U
Farrero 2001	Spain	122	94	2	2	SF	SEC	U
Fernandez 2009	Spain	50	41	2	2	E	PRIM	EDU
Gottlieb 2011	Denmark	61	26	4	Multidisciplinary team, not specified	E	PRIM	U
Güell 2000	Spain	60	47	3	3	E	SEC	U
Güell 2006	Spain	40	25	2	4	E	TERT	DRUG
Koff 2009	US	40	38	4	2	SM	PRIM	U
Littlejohns 1991	UK	152	133	4	3	SF	SEC	U
Mendes 2010	Brazil	117	85	2	2	E	PRIM/SEC	U
Rea 2004	New Zealand	135	117	5	4	SM/SF	PRIM/SEC	U
Rice 2010	US	743	743	3	2	SM	SEC	EDU
Smith 1999	Australia	96	36	8	3	SF	PRIM/SEC	U
Sridhar 2008	UK	122	104	4	3	E/SM	PRIM/SEC	U
Strijbos 1996	Netherlands	50	41	3	3	E	PRIM/SEC	U
Theander 2009	Sweden	30	26	4	4	E	SEC	U
Trappenburg	Netherlands	233	193	3	3	SM	SEC	U
Wakabayashi 2011	Japan	102	85	4	2	IT EDU	SEC	EDU
Wetering 2010	Netherlands	199	175	4	3	E	SEC	U
Wijkstra 1995	Netherlands	45	43	2	3	E	PRIM	U
Wood‐Baker 2006	Australia	135	112	3	2	SM	PRIM	EDU

Main component: SF: structural follow‐up; SM: self management; E: exercise; IT EDU: individually tailored education

Setting: PRIM: primary care; SEC: secondary care; TERT: tertiary care

Control group: U: usual care; DRUG: optimization of drug treatment; EDU: education

Open in table viewer

Table 2. Components of IDM in each included study

Author	Education	Self management	Exacerbation/action plan	Exercise	Psychosocial/occupational	Smoking	Optimal medication	Nutrition	Follow‐up	Case management	Multidisciplinary
Aiken 2006	x	x	x				x			x
Bendstrup 1997	x			x	x	x
Bourbeau 2003	x		x	x					x
Boxall 2005	x			x
Cambach 1997	x			x
Dheda 2004	x				x		x		x
Engstrom 1999	x			x	x			x
Farrero 2001									x	x
Fernandez	x			x
Gottlieb 2011	x			x		x		x
Güell 2000	x			x					x
Güell 2006	x			x
Koff 2009	x	x	x						x
Littlejohns 1991	x						x		x	x
Mendes 2010	x			x
Rea 2004	x		x	x					x		x
Rice 2010	x	x	x
Smith 1999	x	x	x	x		x	x			x	x
Sridhar 2008	x		x	x					x
Strijbos 1996	x			x					x
Theander 2009	x			x	x			x
Trappenburg 2011	x	x	x
Wakabayashi 2011	x	x	x			x
Wetering 2010	x			x		x		x
Wijkstra 1995	x			x
Wood‐Baker 2006	x	x	x

Twenty‐six RCTs met the eligibility criteria for the review, of which two were cluster‐randomized trials (Rea 2004; Wood‐Baker 2006). One trial was a cross‐over trial (Cambach 1997). The studies were published between 1994 and 2011. Five studies originated from the Netherlands (Wijkstra 1994; Strijbos 1996; Cambach 1997; van Wetering 2010; Trappenburg 2011), four studies from Spain (Güell 2000; Farrero 2001; Güell 2006; Fernandez 2009), three studies from Australia (Smith 1999; Boxall 2005; Wood‐Baker 2006), three from the United Kingdom (Littlejohns 1991; Dheda 2004; Sridhar 2008) and three from the United States (Aiken 2006; Koff 2009; Rice 2010). Two studies were conducted in Denmark (Bendstrup 1997; Gottlieb 2011), two originated from Sweden (Engstrom 1999; Theander 2009) and one each from Brazil (Mendes 2010), Canada (Bourbeau 2003), Japan (Wakabayashi 2011) and New Zealand (Rea 2004).

Participants

A total of 2997 COPD patients were randomised in the 26 studies, with a range of 30 to 713 patients per study. Of these, 2523 (84%) patients completed the studies (range 18 to 725). The mean age of the study population was 68 years (SD 3.7), with 68% being male. Patients had a mean FEV1 % predicted of 44.3% (range 28 to 66).

Interventions

Patients were treated in all types of healthcare settings: primary care (eight studies), secondary care (12 studies), tertiary care (one study) and a combination of primary and secondary health care (five studies). The number of health care providers involved in the IDM program ranged from two to seven, with a mean number of three. Furthermore, we calculated the number of components per program, which ranged from two to eight, with a mean number of four.

A priori, we planned to arrange the interventions in order to perform subgroup analysis based on type of intervention, according to type of health care providers, different components, and frequency and duration of intervention. However, it was not possible to determine the mean intensity, frequency or duration of all programs, due to lack of data. Furthermore, as the studies were too heterogeneous, it was not possible to arrange programs according to different combinations of components or combinations of health care providers. Therefore, we determined the dominant component of the IDM program in all studies. The main component of the intervention could directly be determined in nine studies (Littlejohns 1991; Smith 1999; Farrero 2001; Bourbeau 2003; Dheda 2004; Aiken 2006; Wood‐Baker 2006; Koff 2009; Trappenburg 2011) from the objective or title of the study. For example, in Aiken 2006: "The objective is to document outcomes of a randomised trial of the PhoenixCare demonstration program of palliative care and coordinated care/case management for seriously chronically ill individuals who simultaneously received active treatment from managed care organizations. Intensive home‐based case management provided by registered nurse case managers, in coordination with patients’ existing source of medical care, comprised the intervention".

In the remaining 17 studies, the main component was not directly clear from the objective. In 15 studies (Wijkstra 1994; Strijbos 1996; Bendstrup 1997; Cambach 1997; Engstrom 1999; Güell 2000; Boxall 2005; Güell 2006; Fernandez 2009; Theander 2009; Mendes 2010; Rice 2010; van Wetering 2010; Gottlieb 2011; Wakabayashi 2011), we chose the main component of the intervention as the component on which most of the time of the intervention was spent. For example: Bendstrup 1997: "The intervention programme lasted 12 weeks. The programme consisted of the following components. Exercise training: the patients trained together at the hospital for 1h, three times a week for 12 weeks. Occupational therapy: two lessons each group. Education: 12 sessions. Smoking cessation: only for patients wishing to stop smoking."

In one study (Sridhar 2008) there were two components on which most of the time of the intervention was spent (exercise and self management action plan). In another study (Rea 2004) there were two main components: self management action plan and structured follow‐up. Therefore we arranged these two studies as separate categories.

We made the following categories:

IDM dominant component exercise (13 studies: Wijkstra 1994; Strijbos 1996; Bendstrup 1997; Cambach 1997; Engstrom 1999; Güell 2000; Boxall 2005; Güell 2006; Fernandez 2009; Theander 2009; Mendes 2010; van Wetering 2010; Gottlieb 2011).
IDM dominant component self management with an exacerbation action plan (five studies: Bourbeau 2003; Wood‐Baker 2006; Koff 2009; Rice 2010; Trappenburg 2011).
IDM structured follow‐up with nurses/GP (five studies: Littlejohns 1991; Smith 1999; Farrero 2001; Dheda 2004; Aiken 2006).
IDM exercise and self management action plan (one study: Sridhar 2008).
IDM self management action plan and structured follow‐up (one study: Rea 2004)
IDM program of educational sessions, follow by a phase of individually tailored education according to scores on the Lung Information Needs Questionnaire score (one study: Wakabayashi 2011).

In two studies, IDM was compared to another IDM intervention and a control group (Strijbos 1996; Mendes 2010). Both studies involved two intervention groups including an IDM program with a focus on exercise training and one control group. In both studies, we combined and pooled data from the two intervention arms as one group. One study had a cross‐over design with drug treatment after three months (Cambach 1997). Therefore, we used solely the data for the intervention and control group at baseline and at three months.

Control groups consisted of usual care in 20 studies, in two studies control patients received a mono‐disciplinary treatment including optimization of drug treatment (Cambach 1997; Güell 2006) and in four studies control patients received a treatment solely with education (Wood‐Baker 2006; Fernandez 2009; Rice 2010; Wakabayashi 2011). Usual care consisted in all studies of regular follow‐up visits to health care providers, which depended on the type of setting. There was access to health care providers on a 'need to' basis, without additional treatment or management programs. In all studies, no attempts were made to influence this usual care.

Outcomes

We recorded the number of studies reporting a specific outcome as follows:

Quality of life (22 studies)
Exercise capacity (18 studies)
Exacerbation‐related outcomes: measured by number of exacerbations; hospital admissions; hospitalisation days; emergency department (ED) visits; number of prednisolone or antibiotics courses (15 studies)
Lung function (14 studies)
Survival, mortality (five studies)
Depression (four studies)
Dyspnea, measured by MRC Dyspnea score (three studies) or Borg score (three studies)
Co‐ordination of care (three studies)

Details of the included studies are provided in Characteristics of included studies.

We requested additional data from the authors of 14 studies. Of these, 11 authors responded (79%) and six (43%) could provide us with additional data. Therefore, it was not necessary to impute missing data as described in our research protocol (see Dealing with missing data).

Excluded studies

After the first selection based on abstract and title, 49 potentially eligible studies were identified. Finally, after reading the full‐text papers, we excluded 23 studies for one of the following reasons:

not a RCT (n = 1);
no diagnosis of COPD or no obtainable results reported for COPD as a subgroup (n = 2);
intervention includes one component of care (n = 3);
intervention includes one health care provider of different disciplines (n = 4);
duration of intervention is less than three months (n = 4);
active treatment as a control group (n = 9).

The reasons for exclusion are further specified in Characteristics of excluded studies. For ongoing studies, refer to Characteristics of ongoing studies.

Risk of bias in included studies

For full details of 'Risk of bias' judgments see Characteristics of included studies and for an overview see Figure 2.

Figure 2

'Risk of bias' summary: review authors' judgments about each risk of bias item for each included study.

Allocation

Nineteen studies reported full details of adequate sequence generation and we judged them to be of low risk of bias. We judged the remaining seven studies as having unclear risk of bias as they were reported as randomised, but gave no description of the methods used to conceal the sequence. Fourteen studies reported adequate allocation concealment, while we judged four studies as high risk of bias. There were insufficient details for the remaining six studies for us to reach a firm conclusion so we judged them to be at unclear risk of bias. There were 13 studies in which both the sequence generation and concealment of allocation were adequately described, thus selection bias was minimized in these studies.

Blinding

The nature of the intervention precludes the possibility of blinding patients or health care providers. Therefore, we judged all the studies, except Trappenburg 2011, to be at high risk of performance bias. Trappenburg 2011 made a good attempt in using a modified informed consent procedure (postponed information), which meant that patients were unaware of the major aim of the study (education and an action plan), thereby enabling a single‐blind study design (Trappenburg 2011). Therefore, we scored this study as low risk of bias. While blinding of health care providers and patients is impossible with this type of intervention, outcome assessors could be blinded to participants' allocation. This was reported in nine trials indicating a low risk of bias. Outcome assessors were unblinded in seven studies (high risk) and 10 studies provided insufficient information (unclear risk).

Incomplete outcome data

We judged 19 out of the 26 studies as low risk of bias, as they had low drop‐out rates, drop‐out rates were balanced across groups or trial authors performed an intention‐to‐treat analysis. We rated seven studies as high risk of bias and they were likely to be subject to attrition bias. Three out of these seven studies (Dheda 2004; Mendes 2010; Gottlieb 2011) had unbalanced drop‐out rates, with higher rates in the intervention group compared to the control group. One study had a high drop‐out rate balanced in both groups (31%) and the authors performed no intention‐to treat‐analysis (Bendstrup 1997). Cambach 1997 excluded all patients who did not return for one or more of the assessments from the final analyses. In Farrero 2001, quality of life was only investigated in the first 40 consecutive patients, therefore inducing risk of bias. In Smith 1999, all control participants refused to fill in the quality of life questionnaire and expressed that the burden of participating in a study, including questionnaires, was greater than expected.

Selective reporting

We rated 21 studies as low risk of bias and five studies as high risk of bias. Three studies (Rice 2010; van Wetering 2010; Trappenburg 2011) published a study protocol, with which we could compare the results sections. In the other studies, we checked whether the outcomes reported in the methods section of the article were reported in the results section. Five studies (Littlejohns 1991; Smith 1999; Bourbeau 2003; Dheda 2004; Gottlieb 2011) selectively reported outcomes. In two studies (Bourbeau 2003; Dheda 2004) the authors reported no statistically significant difference in the outcome and therefore did not present data, indicating selection bias. In the other three studies (Littlejohns 1991; Smith 1999; Gottlieb 2011), it remained unclear why it was planned to measure an outcome but it was not ultimately published.

Other potential sources of bias

We included two cluster‐randomized trials (Rea 2004; Wood‐Baker 2006). Unfortunately, both studies introduced noteworthy biases related to cluster‐randomization in different ways. In one study (Wood‐Baker 2006) recruitment bias remained unclear, as the authors provided insufficient information regarding the cluster‐randomization process. In contrast, we judged Rea 2004 to have low risk of bias, as clusters were randomised before patients were recruited. Furthermore, we rated both studies as high risk of bias for baseline imbalance between groups, which could have been reduced when stratified or if pair‐matched randomisation of the clusters had been used instead (Higgins 2011). In the Rea 2004 study, there was loss to follow‐up of five clusters (four control and one intervention cluster), therefore this study was subject to bias. There was no follow‐up of clusters in Wood‐Baker 2006 (low risk of bias). Finally, both studies introduced bias as they analyzed data by incorrect statistical methods, not taking the clustering into account. This may account for the over‐precise results and can result in much more weight in a meta‐analysis (Higgins 2011). Therefore, in our meta‐analyses we adjusted for the design effect by reducing the size of the trial to its "effective sample size" (Rao 1992). Based on similar primary care cluster‐randomized trials, we used an intra‐class correlation coefficient (ICC) of 0.01 (Kerry 1998; Campbell 2001). For dichotomous data, we divided both the number of participants and the number experiencing the event by the design effect. For continuous data, we reduced the sample sizes; means and standard deviations remained unchanged (Higgins 2011).

Effects of interventions

See: Summary of findings for the main comparison Integrated disease management compared to control for patients with chronic obstructive pulmonary disease

In the majority of the outcomes, heterogeneity was not encountered. However, there was substantial heterogeneity present in SGRQ total score, six‐minute walk distance (6MWD), CRQ dyspnoea (long‐term), hospital admissions for all causes, hospital days and ED visits. If possible, we performed sensitivity and subgroup analysis on these outcomes to see if the heterogeneity could be explained. Our a priori determined subgroup analysis based on type of health care provider and the frequency and duration of intervention was impossible, as there was large heterogeneity among combinations of health care providers and the exact composition in terms of duration, frequency and intensity of programs was often not clearly reported. In addition, we were not able to perform subgroup analysis on GOLD stage or MRC Dyspnea score, as most studies did not report GOLD stages or MRC Dyspnea score. Furthermore, the definitions and classifications of GOLD stages have been changed over the years, resulting in large variation in severity within subgroups.

Instead, we performed subgroup analysis based on type of setting of the intervention (primary, secondary, tertiary care) and type of control group. Furthermore, we performed subgroup analysis with regard to the dominant component of the IDM program.

We used unadjusted data for meta‐analyses, as only unadjusted data were reported, with the exception of two studies (van Wetering 2010; Trappenburg 2011).

Primary outcomes

1. Quality of life

Of the 26 included studies, 23 measured HRQoL using six different instruments (see Characteristics of included studies):

St. George's Respiratory Questionnaire (SGRQ) (13 studies);
Chronic Respiratory Questionnaire (CRQ) (eight studies);
Short Form‐36 (SF‐36) (three studies);
Sickness Impact Profile (SIP) (two studies);
Dartmouth Primary Care Co‐operative Quality of Life questionnaire (COOP) (one study).

The SGRQ and CRQ are both disease‐specific quality of life questionnaires. However, a meta‐analysis combining CRQ and SGRQ score should not be used as Puhan 2006 has shown that the CRQ is more responsive than the SGRQ. Furthermore, the included generic quality of life questionnaires (SF‐36, SIP and COOP) measure other dimensions of generic health quality of life, and therefore combining data in a meta‐analysis across tools was not possible.

1.1 Respiratory‐specific QoL

1.1.1.1 SGRQ total score ‐ short‐term

The SGRQ is a disease‐specific, validated questionnaire with a scale from 0 (good health) to 100 (worse health status). A negative sign on this questionnaire indicates improvement, and the minimal clinically important difference (MCID) is ‐4 points (Jones 1991). Thirteen studies with a total population of 1425 patients provided data on the SGRQ total score with a follow‐up of 3 to 12 months (Engstrom 1999; Bourbeau 2003; Dheda 2004; Boxall 2005; Wood‐Baker 2006; Koff 2009; Fernandez 2009; Theander 2009; Rice 2010; van Wetering 2010; Gottlieb 2011; Trappenburg 2011; Wakabayashi 2011). The pooled mean difference (MD) on the SGRQ total score was ‐3.71 in favor of IDM (95% confidence interval (CI) of ‐5.83 to ‐1.59; Analysis 1.1; Figure 3; summary of findings Table for the main comparison) which reached statistical significance (P < 0.001) and was close to, but did not reach, the MCID of ‐4 points. In other words, those treated with IDM had 3.71 out of 100 points better quality of life on this questionnaire. Pooling indicated a high degree of heterogeneity (I² = 56%, P = 0.01). Heterogeneity was due to differences in the quality of studies. We were able to reduce heterogeneity if we performed multiple sensitivity analyses based on studies with adequate allocation concealment, adequate blinding of outcome assessment, cluster‐randomization bias, or studies analyzing outcomes by intention‐to‐treat. Sensitivity analysis on studies with adequate allocation concealment (Bourbeau 2003; Boxall 2005; Koff 2009; Theander 2009; van Wetering 2010; Gottlieb 2011; Trappenburg 2011; Wakabayashi 2011) demonstrated that there was still a statistically significant effect in favor of the intervention group (MD ‐3.16; 95% CI ‐4.75 to ‐1.57, P < 0.001). In the same way, in trials (Engstrom 1999; Bourbeau 2003; van Wetering 2010; Rice 2010; Trappenburg 2011; Wakabayashi 2011) with adequate blinding of outcome assessment the effect did not change (MD ‐3.16; 95% CI ‐4.81 to ‐1.51, P < 0.001). A sensitivity analysis excluding the cluster‐randomized study of Wood‐Baker 2006, in which there was an unclear risk of recruitment bias and a high risk of bias on baseline imbalance, the effect changed to a clinically and statistically significant MD in favor of IDM (‐4.22; 95% CI ‐6.14 to ‐2.30, P < 0.001). Lastly, a sensitivity analysis on the studies that analyzed the data using the intention‐to‐treat principle (Bourbeau 2003; Rice 2010) showed a statistically significant and clinically relevant difference in favor of IDM (MD ‐4.65; 95% CI ‐6.69 to ‐2.62, P < 0.0001) compared to controls.

Figure 3

Forest plot of comparison: 1 Integrated disease management versus control, outcome: 1.1 SGRQ: short‐term (3 to 12 months).

Subgroup analysis based on type of setting

There were six studies conducted in primary care on 456 participants (Boxall 2005; Wood‐Baker 2006; Koff 2009; Fernandez 2009; van Wetering 2010; Gottlieb 2011) and seven studies in secondary care on 969 participants (Engstrom 1999; Bourbeau 2003; Dheda 2004; Theander 2009; Rice 2010; Trappenburg 2011; Wakabayashi 2011). No studies were performed in tertiary care. Subgroup analysis based on primary care studies showed a clinically relevant mean difference of ‐4.68 (95% CI ‐8.80 to ‐0.56) in favor of IDM. This result was statistically significant and clinically relevant. Subgroup analysis on secondary care studies showed a statistically significant difference of ‐3.41 (95% CI ‐5.97 to ‐0.85)(Analysis 1.3). This difference was not clinically relevant. The test for subgroup difference did not show a statistically significant difference in treatment effects in patients treated in different types of health care setting (Chi² = 0.27, df = 1 (P = 0.61)).

Subgroup analysis based on study design

We performed subgroup analysis based on study design and compared RCTs (n = 1304) versus cluster‐RCTs (n = 121). There was no difference in SGRQ total score between intervention and control in the cluster‐RCT of Wood‐Baker 2006 (MD 2.30; 95% CI ‐1.62 to 6.22; Analysis 1.4). Pooled meta‐analysis of RCTs showed a clinically relevant effect in favor of the IDM group of ‐4.22 (95% CI ‐6.14 to ‐2.30, P < 0.0001). The test for subgroup differences showed a statistically significant difference between the pooled analysis of the RCTs and the effect in the cluster‐RCT (Chi² = 8.57, df = 1 (P = 0.003)).

Subgroup analysis based on type of control group

In nine studies including 744 participants, control patients received usual care, and in four studies (n = 681) the control group received a mono‐disciplinary treatment of education. Meta‐analysis of the usual care studies showed a significant difference between groups of ‐4.09 (95% CI ‐6.35 to ‐1.84, P < 0.001) (Analysis 1.5). Subgroup analysis of studies in which the control group received education showed no significant difference in effect between groups (MD ‐2.98; 95% CI ‐7.69 to 1.74, P = 0.022), which was neither statistically nor clinically relevant. There was no statistically significant difference in the test for subgroup difference (Chi² = 0.17, df = 1 (P = 0.68)).

Subgroup analysis based on dominant component of the program

There were four studies including 942 patients (Bourbeau 2003; Wood‐Baker 2006; Koff 2009; Rice 2010) in which self management was the dominant component, and six studies including 373 patients in which exercise training was the dominant component (Engstrom 1999; Boxall 2005; Theander 2009; Fernandez 2009; van Wetering 2010; Gottlieb 2011). One study (Wakabayashi 2011) evaluated an individual tailored education program and one study (Dheda 2004) focused mainly on structured follow‐up with nurses and GPs. Subgroup analysis of the self management studies revealed neither a statistically nor a clinically relevant mean difference: MD ‐2.76 (95% CI ‐5.88 to 0.36, P = 0.08). Subgroup analysis of exercise studies showed a statistically and clinically relevant difference of ‐4.74 in favor of IDM (95% CI ‐7.05 to ‐2.43, P < 0.0001). There was no statistically significant difference between subgroups (Chi² = 1.00, df = 1 (P = 0.32)) (Analysis 1.6).

1.1.1.2. SGRQ ‐ long‐term

Two studies including 189 participants measured the long‐term effect on the SGRQ total score: at 18 (Gottlieb 2011) and 24 (van Wetering 2010) months follow‐up. There was no statistically significant difference between groups (MD ‐0.22; 95% CI ‐7.43 to 6.99, P = 0.95; I² = 54%, P = 0.14)(Analysis 1.2).

1.1.2.1 SGRQ domain scores ‐ short‐term

Eleven studies with a total population of 1377 patients reported scores on the SGRQ domains of symptoms, activity and impact. For all domains, there was no significant heterogeneity (I² between 35% and 28%) (Analysis 1.1). We found the following results:

Symptom domain: MD ‐2.39 (95% CI ‐5.31 to 0.53, P = 0.11)
Activity domain: MD ‐2.70 (95% CI ‐4.84 to ‐0.55, P = 0.01)
Impact domain: MD ‐4.04 (95% CI ‐5.96 to ‐2.11, P < 0.0001)

1.1.2.2. SGRQ domain scores ‐ long‐term

Two studies measured the long‐term effect on the SGRQ at 18 months (van Wetering 2010; Gottlieb 2011). Mean differences on all domains had wide confidence intervals and included zero (Analysis 1.2).

1.1.3.1. CRQ domain scores ‐ short‐term

The Chronic Respiratory Disease Questionnaire (CRQ), with a scale from 0 to 7 and a MCID of 0.5, was reported in eight trials (Wijkstra 1994; Bendstrup 1997; Cambach 1997; Güell 2000; Farrero 2001; Rea 2004; Güell 2006; Sridhar 2008). Three of these (Bendstrup 1997; Farrero 2001; Rea 2004) could not be used in a meta‐analysis. Bendstrup 1997 and Rea 2004 reported insufficient data and the authors could not provide us with additional data. In addition, Farrero 2001 administered the CRQ in the first 40 consecutive patients and therefore outcomes were not published.

The pooled results of four studies including 160 participants (Wijkstra 1994; Cambach 1997; Güell 2000; Güell 2006) measuring the CRQ until 12 months follow‐up are shown in Figure 4 and Analysis 1.7. For each of the CRQ domains, the MD was well above the MCID of 0.5 units and differences in scores were statistically significant: dyspnoea (MD 1.02; 95% CI 0.67 to 1.36, P < 0.0001), fatigue (MD 0.82; 95% CI 0.46 to 1.17, P < 0.0001), emotion (MD 0.61; 95% CI 0.26 to 0.95, P < 0.0005) and mastery (MD 0.75; 95% CI 0.38 to 1.12, P < 0.0001). The results showed homogeneity across studies.

Figure 4

Forest plot of comparison: 1 Integrated disease management versus control, outcome: 1.7 CRQ: short‐term (3 to 12 months).

1.1.3.2. CRQ domain scores ‐ long‐term

Two studies (n = 151) (Güell 2000; Sridhar 2008) measured the long‐term effectiveness on CRQ domain scores at 24 months follow‐up(Analysis 1.8). There was no difference between groups on the CRQ dyspnoea domain: MD 0.47 (95% CI ‐0.31 to 1.25, P = 0.24). Pooled data showed substantial heterogeneity (I² = 70%, P = 0.07), which was related to differences in the type of intervention (exercise in the Güell 2000 study versus structured follow‐up with a respiratory nurse and exacerbation plan in Sridhar 2008). Güell 2000 demonstrated a significant difference in favor of IDM (MD 0.92; 95% CI 0.19 to 1.65, P = 0.01). In contrast, there was no statistically significant difference between groups on the CRQ dyspnoea domain in Sridhar 2008 (MD 0.12; 95% CI ‐0.32 to 0.58, P = 0.61).

Pooled mean differences on the domains fatigue, emotion and mastery showed homogeneity across studies. On the CRQ fatigue domain, there was a statistically significant but not clinically relevant difference of 0.45 in favor of IDM (95% CI 0.05 to 0.85, P = 0.03). On the CRQ emotion and mastery domain, the statistically and clinically relevant effect was in favor of IDM: emotion MD 0.53 (95% CI 0.10 to 0.95, P = 0.02) and mastery MD 0.80 (95% CI 0.37 to 1.23, P < 0.01).

1.2 General health‐related QoL

General HRQoL was measured with the SF‐36 in three studies (Dheda 2004; Rea 2004; Aiken 2006). The authors of these studies could not provide us with sufficient data for pooling in a meta‐analysis. Neither study found a significant effect between groups. Two of these studies (Dheda 2004; Aiken 2006) suffer from small sample sizes varying from 15 to 10 patients per group per study, which makes it difficult to detect an effect (underpowered studies).

We pooled the data from two studies (Littlejohns 1991; Engstrom 1999) reporting data on the SIP (Analysis 1.9). No between‐group differences in any domain of the SIP were found. One other study used the York Quality of Life Questionnaire (Bendstrup 1997) and reported no significant difference. Smith 1999 used a modified version of the Dartmouth Primary COOP. In this study, the authors analyzed only the data from the intervention group (n = 30) due to lack of data in the control group. The authors concluded that the total COOP scores in the intervention group significantly improved HRQoL at 12 months.

2. Exercise capacity

Seventeen studies measured exercise capacity using either the 6MWD or the cycle ergometer test. The MCID on the 6MWD is estimated at 35 meters (Puhan 2008). There is no MCID reported in the current literature for the cycle ergometer test. Results are shown in Figure 5.

Figure 5

Forest plot of comparison: 1 Integrated disease management versus control, outcome: 1.10 Functional exercise capacity: 6MWD mean difference.

2.1.1 Functional exercise capacity ‐ short‐term

We pooled data from 14 studies using the 6MWD including 871 participants. One study could not be pooled, as the authors reported no data because there was no significant difference between groups at 12 months follow‐up (Bourbeau 2003).

Patients treated with IDM improved their 6MWD by a statistically and clinically relevant 43.86 meters (95% CI 21.83 to 65.89)(Figure 5; Analysis 1.10). There was heterogeneity between the results of the studies (I² = 83%). This heterogeneity is explained by differences in the quality of studies. We performed sensitivity analysis on studies with adequate allocation concealment, which reduced heterogeneity (I² = 0%) and reduced the effect to a MD of 15.15 meters, which was still statistically significant (95% CI 6.37 to 23.93, P < 0.001), however no longer clinically relevant. Furthermore, we performed subgroup analysis based on type of setting, type of control group and dominant component of the intervention.

Subgroup analysis based on type of setting

There were seven studies with 427 participants (Wijkstra 1994; Cambach 1997; Boxall 2005; Fernandez 2009; van Wetering 2010; Mendes 2010; Gottlieb 2011) conducted in primary care, seven studies with 438 participants (Littlejohns 1991; Bendstrup 1997; Engstrom 1999; Güell 2000; Theander 2009; Mendes 2010; Wakabayashi 2011) in secondary care and one study in tertiary care with 35 participants (Güell 2006). Both subgroup analyses showed similar statistically and clinically relevant improvements: exercise training in primary care revealed a MD of 45.16 meters (95% CI 8.65 to 81.67, P = 0.02), whereas in the secondary care setting the MD was 49.18 meters (95% CI 14.28 to 84.08, P = 0.006). The tertiary care study showed a significant effect in favor of IDM of 85 meters (95% CI 30.43 to 139.57). Results are shown in Analysis 1.11 and Figure 6.

Figure 6

Forest plot of comparison: 1 Integrated disease management versus control, outcome: 1.11 Subgroup analysis 6MWD based on type of setting.

Subgroup analysis based on control group

We pooled four studies with 180 participants in which control patients received a treatment with optimal medication (Cambach 1997; Güell 2006) or an education session (Fernandez 2009; Wakabayashi 2011) in a subgroup analysis. In the same way, we pooled 10 studies (Littlejohns 1991; Wijkstra 1994; Bendstrup 1997; Engstrom 1999; Güell 2000; Boxall 2005; Theander 2009; Mendes 2010; van Wetering 2010; Gottlieb 2011) including 691 participants in which the control group consisted of usual care.

Subgroup analysis in which one component of treatment was used showed no difference between groups (MD 35.99; 95% CI ‐5.34 to 77.31, P = 0.09)(Analysis 1.12). In studies in which the control group consisted of usual care, the 6MWD improved clinically and statistically significantly by 46.59 meters in favor of IDM (95% CI 19.68 to 73.51, P = 0.0007). However, the test for subgroup differences did not show any difference between control groups (Chi² = 0.18, df = 1 (P = 0.67)).

Subgroup analysis based on dominant component of intervention

Twelve out of the 14 studies (n = 653) measuring exercise capacity incorporated some kind of exercise training in their IDM programs. We performed subgroup analysis, which showed that the 6MWD improved by 51.47 meters (95% CI 26.53 to 76.40). This effect was statistically and clinically relevant. In the remaining two studies (n = 218), exercise was not part of the IDM programs. In one study (Wakabayashi 2011), which consisted of individually tailored education sessions, there was no difference between groups (MD 0.40; 95% CI ‐39.64 to 40.44, P = 0.98). The other study (Littlejohns 1991), in which there was a focus on structured follow‐up with GP and nurses, revealed no effect (MD 3.50; 95% CI ‐28.31 to 35.31, P = 0.83). In conclusion, studies incorporating exercise training in their IDM programs demonstrated larger effect sizes; this was statistically significant using the test for subgroup difference (Chi² = 7.49, df = 2 (P = 0.02))(Analysis 1.13).

2.1.2 Functional exercise capacity ‐ long‐term

Two studies on 184 participants published long‐term results on the 6MWD (van Wetering 2010; Gottlieb 2011). Both studies showed that IDM statistically significantly improved exercise capacity measured on the 6MWD by 16.8 meters (MD 16.84; 95% CI 3.01 to 30.67) compared to the control group. However, this effect did not exceed the MCID. There was no heterogeneity present. Results are shown in Figure 5 and Analysis 1.10.

2.2. Maximal exercise capacity

Four studies on 298 participants assessed the maximal exercise capacity (in Watts) using the cycle ergometer test. Both studies showed that IDM statistically significantly improved the maximal exercise capacity by 7 Watts (MD 6.99; 95% CI 2.96 to 11.02, P < 0.0001)(Analysis 1.14).

3. Exacerbations

3.1.1 Number of patients experiencing at least one exacerbation ‐ short‐term

Two studies (Bourbeau 2003; Trappenburg 2011) including 407 patients reported on the number of patients experiencing at least one exacerbation during 12 months of follow‐up. Both studies used the same definition and defined an exacerbation as an increase in symptoms, with deterioration of dyspnoea or purulent sputum. Pooled meta‐analysis showed homogeneity and a pooled OR of 1.21 (95% CI 0.77 to 1.91) (Analysis 1.15), which showed no statistically or clinically relevant difference between groups. The trial authors of the Bourbeau 2003 study reported that although there were more patients experiencing at least one exacerbation in the intervention group (85 versus 81), the total number of exacerbations was higher in the control group (362) compared to the intervention group (299). This was of borderline significance (P = 0.06). Similarly, the number of patients experiencing three or more exacerbations during 12‐month follow‐up was higher in the control group (67.9%), compared to the action plan group (62.3%). Exacerbations in the intervention group were treated successfully at an early stage, which probably resulted in fewer patients with a hospital admission (17.2% versus 36.3%, P < 0.01). Trappenburg 2011 reported similar findings: although exacerbation rates did not differ between groups, exacerbations in the action plan group were perceived as substantially milder by patients, and they reported on average three days faster than those in the control group.

3.1.2. Number of patients experiencing at least one exacerbation ‐ long‐term

Two studies (Sridhar 2008; van Wetering 2010) including 301 patients assessed the number of patients experiencing at least one exacerbation at 24 months follow‐up. Both studies related the definition of an exacerbation to health care. Sridhar 2008 stated they defined an exacerbation as the "unscheduled need for healthcare, or need for steroid tablets, or antibiotics for worsening of their COPD". Similarly, van Wetering 2010 defined a moderate exacerbation as "a visit to the general practitioner or respiratory physician in combination with a prescription of antibiotics and/or prednisolone or a visit to the emergency department or day care of a hospital, which according to the patient, was related to a COPD exacerbation. A severe exacerbation was defined as a hospitalisation for a COPD exacerbation". Pooled meta‐analysis demonstrated no difference between groups (OR 1.53; 95% CI 0.90 to 2.60, P = 0.12) (Analysis 1.16). There was homogeneity between studies. Sridhar 2008 stated that patients in the intervention group were more likely to have exacerbations treated with oral steroids alone or oral steroids and antibiotics than the control group. The initiator of treatment was statistically more likely to be the patient themselves compared to the GP in the control group.

3.1.3 Mean exacerbation rate ‐ long‐term

Two studies (Güell 2000; van Wetering 2010) including 226 participants reported on the exacerbation rate in both groups at 24 months follow‐up. Data on exacerbations were skewed in the van Wetering study, therefore we decided not to pool both studies in a meta‐analysis. In Güell 2006, control group patients (n = 23) experienced 207 exacerbations, with an average of 6.9 (3.9) exacerbations per patients, ranging from 0 to 16 exacerbations during the 24 months. The IDM group experienced 111 exacerbations, with an average of 3.7 (2.2) exacerbations per patients, ranging from 0 to 9 exacerbations during the 24 months. This difference was statistically significant (P < 0.0001) favoring IDM. In van Wetering 2010, the exacerbation rate was 2.78 in the IDM group and 2.16 in the control group, resulting in a rate ratio of 1.29 (95% CI 0.89 to 1.87), which was not statistically significant (P = 0.113).

3.2.1 Hospital admissions, all causes ‐ short‐term

Two studies on 266 participants (Littlejohns 1991; Rea 2004) reported data on the number of patients who were admitted for all causes until 12 months follow‐up. There was no heterogeneity and there was no difference between groups (OR 0.62; 95% CI 0.36 to 1.07, P = 0.49) (Analysis 1.17).

3.2.2. Hospital admissions, all causes ‐ long‐term

Two studies including a total of 283 patients (Sridhar 2008; van Wetering 2010) assessed the number of patients admitted until 24 months follow‐up. Pooled results showed heterogeneity (I² = 53%), which could be explained as van Wetering 2010 showed a positive effect in favor of IDM and Sridhar 2008 showed no significant difference in effect between groups. Therefore, a pooled meta‐analysis showed no difference between groups (OR 0.78; 95% CI 0.38 to 1.57)(Analysis 1.18).

3.3.1. Respiratory‐related admissions ‐ short‐term

We pooled data from seven studies (Smith 1999; Bourbeau 2003; Rea 2004; Boxall 2005; Koff 2009; Rice 2010; Trappenburg 2011) measuring respiratory‐related admissions until 12 months follow‐up in a meta‐analysis. Studies were homogeneous. Pooled estimates showed a statistically significant difference in favor of IDM (OR 0.68; 95% CI 0.47 to 0.99, P = 0.04)(Analysis 1.19). In the control group 27 people out of 100 had a respiratory‐related hospital admission over 3 to 12 months, compared to 20 (95% CI 15 to 27) out of 100 in the integrated disease management group, as presented in Figure 7. Over the course of a year, the number needed to treat with IDM to prevent one hospital admission was NNT(B) 15 (95% CI 9 to 506).

Figure 7

In the control group 27 people out of 100 had a respiratory‐related hospital admission over 3 to 12 months, compared to 20 (95% CI 15 to 27) out of 100 for integrated disease management group.

3.3.2. Respiratory‐related admissions ‐ long‐term

Data from one trial (van Wetering 2010) presented data on the number of patients admitted until 24 months follow‐up. There was no difference between the control and IDM group on the number of respiratory‐related admissions (OR 0.59; 95% CI 0.28 to 1.22, P = 0.16)(Analysis 1.20).

3.4.1 Hospital days per patient ‐ short‐term

Six studies on 741 patients (Engstrom 1999; Farrero 2001; Bourbeau 2003; Rea 2004; Boxall 2005; Trappenburg 2011) reported the difference in mean hospitalisation days per patient per group (intervention versus control). Patients treated with IDM were on average discharged from the hospital nearly four days earlier compared to control patients, with a confidence interval from six to two days (MD ‐3.78; 95% CI ‐5.90 to ‐1.67, P < 0.001) (Analysis 1.21). There was heterogeneity in the results (I² = 55%). Inspection of the forest plot shows that this was the result of one outlying study (Engstrom 1999), which reported more days for intervention patients. The authors stated that the data on admission days in his study were skewed, as one patient accounted for 50% of the increase in the IDM group. Reanalysis with exclusion of this trial did not change the significance, direction or effect of the mean difference.

3.4.2. Hospital days per patient ‐ long‐term

One trial with 175 patients (van Wetering 2010) reported the difference in mean number of total hospital days per patient per group at 24 months follow‐up. There was no difference between groups (MD 0.60; 95% CI ‐3.01 to 4.21, P = 0.74) (Analysis 1.22).

3.5 Emergency Department (ED) visits

Six trials (Smith 1999; Farrero 2001; Bourbeau 2003; Rea 2004; Rice 2010; Wakabayashi 2011) assessed in various ways the number of ED visits. We were able to pool the data from four studies with 1161 patients (Smith 1999; Bourbeau 2003; Rea 2004; Rice 2010), which revealed no difference between groups with high heterogeneity (OR 0.64; 95% CI 0.33 to 1.25; I² = 71%)(Analysis 1.23). Sensitivity analysis on two studies which analyzed by intention‐to‐treat and which blinded outcome assessors revealed a mean difference of 0.49 in favor of the control group (MD 0.49; 95% CI 0.36 to 0.67, P < 0.0001, I² = 0%). Three studies could not be pooled, due to lack of required data. Of these excluded studies, Trappenburg 2011 and Wakabayashi 2011 reported the mean ED visits per patient at baseline and follow‐up. Both studies concluded no statistically significant difference between groups compared to baseline. On the other hand, Farrero 2001 reported a significant decrease in ED visits per patient in favor of the IDM group (0.45 ± 0.83 for intervention group, 1.58 ± 1.96 for control group; P = 0.0001). There were no data presented on the number of ED visits at long‐term follow‐up.

3.6 Patients using at least one course of oral steroids

We pooled data from three studies including 348 patients (Littlejohns 1991; Farrero 2001; Rea 2004) on the number of patients using at least one course of oral steroids until 12 months follow‐up. Results were homogeneous and there was no difference between groups (OR 1.13; 95% CI 0.64 to 2.01, P = 0.66) (Analysis 1.24).

3.7. Patients using at least one course of antibiotics

There were two studies with 236 participants (Littlejohns 1991; Rea 2004) reporting on the number of patients using at least one course of antibiotics. The studies presented conflicting results and heterogeneity was large, as Rea 2004 was a primary care, cluster‐randomized trial and Littlejohns 1991 was a RCT in the secondary care setting. The number of patients using at least one course of antibiotics was not different between groups, and the OR had a wide confidence interval (OR 1.43; 95% CI 0.24 to 8.48, P = 0.69)(Analysis 1.25).

Secondary outcomes

4. Dyspnea

Four studies reported the MRC Dyspnea Scale as an outcome (Mendes 2010; van Wetering 2010; Gottlieb 2011; Wakabayashi 2011), however Gottlieb failed to publish any results. We pooled data from the remaining three studies, including 345 patients. Dyspnea was improved in the IDM group by ‐0.30 points (MD ‐0.30; 95% CI ‐0.48 to ‐0.11, I² = 0%, P < 0.001)(Analysis 1.26).

Furthermore, three studies on 145 patients used the Borg score to detect changes in perceived dyspnoea (Güell 2000; Boxall 2005; Gottlieb 2011). These data were pooled and revealed no change in dyspnoea (MD 0.14; 95% CI ‐0.70 to 0.98, P = 0.74, I² = 39%)(Analysis 1.27).

5. Mortality

Five trials assessing 1207 patients explicitly recorded mortality as an outcome. Of these, four trials assessed mortality at 12 months (Littlejohns 1991; Smith 1999; Farrero 2001; Rice 2010) and one study at 24 months (Sridhar 2008). There was no statistically significant difference between groups at short‐ (OR 0.96; 95% CI 0.52 to 1.74, P = 0.33; I² = 59%) and long‐term follow‐up (OR 0.45; 95% CI 0.16 to 1.28, P = 0.13) (Analysis 1.28). Heterogeneity in the short‐term studies is due to different dominant components of the interventions.

6. Lung function

Lung function was measured in three different ways in 10 trials (Littlejohns 1991; Wijkstra 1994; Güell 2000; Farrero 2001;Bourbeau 2003; Wood‐Baker 2006; Sridhar 2008; Fernandez 2009; van Wetering 2010; Wakabayashi 2011). Therefore, we created three different subgroups, which we pooled in two different meta‐analyses: forced expiratory volume in one second (FEV1) in liters and FEV1 as per cent predicted for age, gender and height (FEV1% predicted), as well as the mean difference in FEV1% predicted from baseline. All pooled data on short‐ as well as on long‐term outcome revealed no significant difference in lung function between groups(Analysis 1.29; Analysis 1.30).

7. Anxiety and depression

Four studies assessed depression as an outcome (Engstrom 1999; Littlejohns 1991; Güell 2006; Trappenburg 2011). Two studies (Littlejohns 1991; Trappenburg 2011) used the HADS, one study (Engstrom 1999) used the Mood Adjective Check List (MACL) and one study (Güell 2006) used a Revised Symptom Checklist. We pooled results on the HADS in a meta‐analysis including 316 patients, which revealed no statistically significant difference between groups for anxiety (MD 0.22; 95% CI ‐0.41 to 0.85, I² = 0%) or depression (MD 0.21, 95% CI ‐0.39 to 0.81, I² = 0%)(Analysis 1.31). Engstrom 1999 used the MACL, a shortened 38‐item version covering three basic dimensions of mood: pleasantness/unpleasantness, activation/deactivation and calmness/tension. No significant differences were found between groups. The aim of Güell 2006 was specifically to evaluate the effect of a pulmonary rehabilitation program on psychosocial morbidity (without including any specific psychological intervention), as well as effort capacity and HRQoL. Therefore, the authors used a Revised Symptom Checklist, containing 90 items, which included depression and anxiety. Following a per protocol analysis, the intervention group showed a significant improvement in depression (P ≤ 0.01) and anxiety (P ≤ 0.05).

8. Co‐ordination of care

Three studies (Littlejohns 1991; Bendstrup 1997; Koff 2009) reported in some way on the co‐ordination of care. However, these studies had different intervention programs and reported on co‐ordination of care in different ways. Therefore, interpretation of outcomes is difficult. Bendstrup 1997 reported an attendance rate of 78% of patients following a 12‐week IDM program (consisting of education, exercise training, smoking cessation and occupational therapy).

Patient satisfaction with regard to the provided health care was measured in two studies. In Koff 2009, satisfaction with a self management/action plan program was assessed on a scale from 1 to 10 in the intervention group, with 1 being strongly dissatisfied and 10 completely satisfied. Patients expressed high satisfaction with all of the equipment used, except for the pedometer. Littlejohns 1991 designed a satisfaction questionnaire for his study, which included questions on satisfaction with level of care, the information given to patients and their knowledge of medication. The questionnaire was used in both study groups. At 12 months follow‐up, there was little difference in the level of satisfaction with the service provided between groups.

Discussion

Summary of main results

We reviewed the results of 26 randomised controlled trials evaluating the effect of an integrated disease management (IDM) program in patients with COPD. All included studies contained a program provided by caregivers from at least two different disciplines, with two different components (for example exercise, education, self management etc) and with a duration of at least three months. Firstly, pooled data showed statistically and clinically relevant improvements in disease‐specific quality of life on the CRQ in the IDM group: dyspnoea (MD 1.02; 95% CI 0.67 to 1.36); fatigue (0.82; 95% CI 0.46 to 1.17); emotional (0.61; 95% CI 0.26 to 0.95) and mastery (0.75; 95% CI 0.38 to 1.12). All domains (dyspnoea, fatigue, emotional and mastery) exceeded the minimum clinically relevant difference until 12 months follow‐up. Only two studies measured long‐term results on the CRQ, which showed that the positive effect was maintained for the fatigue, emotion and mastery domains at 24 months follow‐up. Furthermore, disease‐specific quality of life was also measured with the SGRQ. There was considerable heterogeneity in the score on the SGRQ. After multiple sensitivity analyses, we concluded that there was a difference in the SGRQ total score in favor of patients treated with IDM, which lies around the minimal clinically relevant difference of four units. The effect was greatest for the impact domain. We could not find a difference in the SGRQ total score at long‐term follow‐up. Remarkably, only two studies could provide data.

Second, the pooled data showed statistically significant improvements in maximal and functional exercise capacity, with an improvement of 7 Watts and 44 meters in favor of the IDM group, respectively. Sensitivity analysis of the 6MWD lowered the effect to 15 meters, indicating the likelihood of an overestimated effect in the lower quality studies.

Thirdly, the total number of patients with at least one respiratory‐related hospital admission decreased from 27 per 100 to 20 per 100 patients in favor of the intervention group, with a number needed to treat of 15 patients to prevent one being admitted to hospital over three to 12 months. Mean hospitalisation days decreased on average by three days in the IDM group. The effects on the aforementioned primary outcomes are summarized in the summary of findings Table for the main comparison. There was no evidence of an effect on generic quality of life, the number of patients with at least one exacerbation, the number of hospital admissions for all causes, emergency department visits, courses of antibiotics/prednisolone, dyspnoea, lung function parameters or depression scores.

Overall completeness and applicability of evidence

We found sufficient studies to address the objective of this review. All studies reported at least one primary outcome, and all studies were included in at least one pooled analysis. The COPD population in the included studies ranged from mild to very severe COPD and trials were conducted across all types of healthcare settings in a range of different countries. Although the results of this review appear therefore to be applicable to all COPD patients worldwide, one should bear in mind that applicability may depend on the context of available healthcare resources. The IDM programs included in this review differed in the type of health care providers involved, type of components and duration of intervention, reflecting the diversity of daily practice. Overall, programs containing at least two health care providers and two different elements, showed improvements in quality of life and exercise capacity, and reduced the number of hospital admissions and days spent in the hospital. We found no differences in quality of life and exercise tolerance between patients treated in primary or secondary care. Although the mean differences between groups were lower in studies using a mono‐disciplinary treatment as a control group compared to usual care, the subgroup difference did not reach statistical significance. Furthermore, subgroup analysis on studies focusing mainly on exercise programs showed a statistically significant greater improvement in exercise capacity. Further research is required to define the optimal combination, intensity and duration of components in IDM programs.

Quality of the evidence

We included RCTs only and found 26 trials assessing almost 3000 participants. A priori, we intended to perform meta‐analyses on some outcomes when feasible. However, with this amount of data we were able to perform pooled data analysis for all outcomes. As a result of the complex intervention, there was a certain amount of clinical and statistical heterogeneity among studies. We have incorporated heterogeneity into the estimated effects by using random‐effects analyses, where possible. Using the GRADE approach, we specified the levels of quality of the evidence (high, moderate, low and very low) in our 'Summary of findings' table. According to this approach, we checked if the included trials had limitations in terms of design, indirectness of the evidence, unexplained heterogeneity or inconsistency of the results, imprecision of the results or high probability of publication bias. If one of these factors was present, we downgraded the evidence. On the SGRQ, there was considerable variation in risk of bias between studies. Risk of bias tended to be lower in the more recently published trials compared to older trials. Sensitivity analyses based on studies with low quality did not change the direction, significance or magnitude of the effect. Therefore we concluded that the quality of the evidence was 'high'. For the CRQ, there were four studies which were all of moderate quality and presented with some form of bias, therefore we did downgrade the evidence to 'moderate' quality. We downgraded the evidence on functional exercise capacity for inconsistency, as substantial heterogeneity (I² = 84%) was present. After performing sensitivity analysis, the mean difference substantially decreased to 15 meters. We did not downgrade for respiratory‐related admissions or hospitalisation days, as we feel the studies presented consistent, homogeneous results. We expect that additional trials with proper description of their methods and data collection could upgrade the quality of evidence and further our findings.

Potential biases in the review process

Several methodological strengths minimized the risk of bias in this review. As definitions of IDM are still under debate, we a priori strictly determined the inclusion criteria for an IDM program, which was published in our protocol. Our definition was derived from the definitions published in the literature (Peytremann‐Bridevaux 2009; Schrijvers 2009). Overall, they reported on "multiple interventions, designed to manage chronic conditions, with a focus on a multidisciplinary approach". Furthermore, these definitions suggest that IDM interventions should "focus on maximum clinical outcome, regardless of treatment setting(s) or typical reimbursement patterns". As a result, we chose to include all interventions, independent of treatment setting, and to keep our definition as simple as possible, in order to be easily understandable for readers and easy to use for us as authors when checking on all relevant literature. Therefore, we restricted the inclusion of trials to multi‐component, multidisciplinary programs of at least three months duration. Furthermore, we performed comprehensive searches to identify possible studies, leading to almost 4800 potentially relevant abstracts being identified. Subsequently, three different assessors assessed the abstracts. All studies that were excluded by two authors because of the type of intervention were triple‐checked by a third review author to make sure all studies describing an IDM program were included. We reached consensus on all included studies. Although we followed the inclusion criteria for IDM as stated in our protocol, final decisions on the inclusion of studies are open to interpretation or criticism.

Limitations of this review include possible bias from inconsistent reporting of data from included studies. We requested additional data from 14 authors and received an answer from 11. Six of them could provide us with additional data, which could potentially have biased the results. Furthermore, only three out of 26 studies published a study protocol with which we could compare the results sections. In the other studies, we examined whether the outcomes reported in the methods section of the paper were reported in the results section. It is possible that this could have introduced bias if the authors blanked out outcomes from their methods section.

Lastly, there was heterogeneity present in the control group as we used a broad a priori definition of controls, varying from no treatment to treatment including one component of COPD care. We acknowledge the fact that controls and usual care differ between countries and between healthcare settings. Therefore, we performed subgroup analysis to investigate to what extent a difference between the control groups possibly influenced the results. From these analyses we concluded that the effect between intervention and control groups is less strong if patients in control groups receive one component of IDM compared to patients receiving no treatment or usual care.

Agreements and disagreements with other studies or reviews

This review adds to the results of four earlier systematic reviews analyzing IDM for COPD patients (Adams 2007; Niesink 2007; Peytremann‐Bridevaux 2008; Lemmens 2009). The current review brings together new trials that were not included in any of these reviews. Some of these earlier reviews analyzed some of our primary outcomes. Adams 2007 examined the effectiveness of programs for COPD patients including chronic care model components and pooled six trials including at least two components. Pooled results did not demonstrate statistically significant differences on the SGRQ. Patients with COPD who received interventions with two or more chronic care model components had lower rates of hospitalisation and a shorter length of stay compared with control groups, comparable to our results. Lemmens 2009 examined the effectiveness of multiple interventions in asthma and COPD patients. The authors pooled data on the SGRQ from three studies in which two components of IDM were compared to usual care and three studies in which three components of IDM were compared to usual care. The effect on the SGRQ was larger if three components of IDM were used (MD ‐4.69; 95% CI ‐8.34 to ‐0.83 versus MD ‐0.95; 95% CI ‐4.23 to 2.34). Pooled data from five studies showed a decrease in the number of respiratory‐related hospitalizations, with a pooled OR of 0.58, which is comparable to the OR of 0.67 found in this review. Niesink 2007 evaluated quality of life in COPD patients, but did not perform a meta‐analysis; reasons for this were not clearly described. Five out of 10 studies showed a clinically relevant improvement in quality of life. Peytremann‐Bridevaux 2008 examined the effectiveness of IDM in COPD patients on exercise tolerance, quality of life, hospital admissions and mortality. Only data on hospital admissions and exercise tolerance were pooled. Positive effects on exercise capacity are in line with this review. The authors demonstrated a mean improvement of 32 meters on the 6MWD in five studies, which is comparable to our results. Furthermore, a pooled odds ratio of 0.85 (95% CI 0.54 to 1.36) for mortality is comparable to our review. Differences between this review and these other reviews are related to differences in the inclusion criteria for patients and the focus of programs. All reviews used different definitions of IDM; however there was some overlap with this review. Lemmens 2009 et al also based their definition on the EPOC list (EPOC 2008), whereas Adams 2007 and Steuten 2009 based their definition of IDM on the chronic care model as reported by Wagner 1996. The definition used by Peytremann‐Bridevaux 2008 was similar to our definition, with the only difference being a duration of the intervention of at least 12 months instead of three months. Finally, all the aforementioned systematic reviews included study designs other than RCTs.

Our findings from the St. George's Respiratory Questionnaire (SGRQ) showed improvements of a similar magnitude to those reported in two recent Cochrane reviews evaluating two other supposedly important pharmaceutical cornerstones of COPD treatment, tiotropium (Karner 2012a) and inhaled corticosteroids (Yang 2012). IDM resulted in a higher MD on the SGRQ of ‐3.71 compared to the MD of tiotropium (‐2.89); however, the confidence interval for IDM is wider (95% CI ‐5.83 to ‐1.59) compared to the confidence interval (95% CI ‐3.35 to ‐2.44) for tiotropium.

Eight studies in this review are also evaluated in a Cochrane review assessing the effectiveness of pulmonary rehabilitation (Lacasse 2006) and four studies included in this review are also evaluated in a Cochrane review assessing the effect of self management programs (Effing 2007). In line with the review of Effing 2007 (OR 0.64; 95% CI 0.47 to 0.89) we found a decrease in respiratory‐related hospital admissions (OR 0.64; 95% CI 0.47 to 0.89). Furthermore, both reviews demonstrated improvements in disease‐specific quality of life, although the effects tended to be higher and clinically relevant in the pulmonary rehabilitation review (Lacasse 2006), whereas in the self management review the improvement was too small to be of clinical relevance (Effing 2007). A priori we determined subgroup analyses on the type of dominant intervention in the program. Subgroup analysis of studies containing some form of exercise training showed greater improvement in quality of life, which exceeded the clinically relevant threshold on almost all domains. These results are in line with the Lacasse review. However, a subgroup analysis performed on studies that mainly focused on self management did not exceed the minimum clinically important difference, in line with the Effing review.

Furthermore, Effing 2007 and Lacasse 2006 reported pooled estimates for functional exercise capacity. Not surprisingly, as the focus in most included pulmonary rehabilitation studies lies on exercise training, the 6MWD improved significantly by 48 meters in the Lacasse review. This effect size is comparable to our overall estimate of 44 meters and our subgroup analyses on studies including an exercise program in which we found a mean difference of 50 meters. In contrast to these results, Effing did not find any significant differences in exercise capacity at all (weighted mean difference ‐6.25; 95% CI ‐24.05 to 11.05).

We did not find a difference between groups in the number of patients with at least one exacerbation. However, we concluded that there was a reduction in the number of patients admitted and the mean number of hospital days related to exacerbations. Self management education including the use of action plans might lead to more and better self treatment of exacerbations. As a result, hospital admissions will decrease (Effing 2007). In our included studies, a self management program caused patients to respond three days sooner on complaints (Trappenburg 2011). Furthermore, patients more often initiated treatment by themselves, which could then be successfully treated with oral steroids at an early stage (Sridhar 2008). As a result, perceived exacerbations were rated as substantially milder (Trappenburg 2011) and were less likely to result in an admission (Bourbeau 2003).

In the past few years, several systematic reviews evaluating IDM for various other chronic conditions have been published (Norris 2002; Badamgarav 2003; Gonseth 2004; Neumeyer‐Gromen 2004; Knight 2005; Roccaforte 2005; Pimouguet 2010). Overall, quality of care improved with these programs, however some of the differences were in fact clinically modest (Peytremann‐Bridevaux 2008). We found that the results of this review were most comparable to a systematic review evaluating patients with heart failure, which demonstrated that all‐cause and heart failure‐related hospitalisation rates were significantly reduced: OR 0.76 (CI 0.69 to 0.94, P < 0.0001) and OR 0.58 (CI 0.50 to 0.67, P < 0.0001), respectively (Roccaforte 2005). In studies evaluating depression and diabetes, differences in health care use and quality of care were less clear (Neumeyer‐Gromen 2004; Knight 2005).

Figure 1

Study flow diagram.

Navigate to figure in ReviewOpen in new tab

Figure 2

'Risk of bias' summary: review authors' judgments about each risk of bias item for each included study.

Navigate to figure in ReviewOpen in new tab

Figure 3

Forest plot of comparison: 1 Integrated disease management versus control, outcome: 1.1 SGRQ: short‐term (3 to 12 months).

Navigate to figure in ReviewOpen in new tab

Figure 4

Forest plot of comparison: 1 Integrated disease management versus control, outcome: 1.7 CRQ: short‐term (3 to 12 months).

Navigate to figure in ReviewOpen in new tab

Figure 5

Forest plot of comparison: 1 Integrated disease management versus control, outcome: 1.10 Functional exercise capacity: 6MWD mean difference.

Navigate to figure in ReviewOpen in new tab

Figure 6

Forest plot of comparison: 1 Integrated disease management versus control, outcome: 1.11 Subgroup analysis 6MWD based on type of setting.

Navigate to figure in ReviewOpen in new tab

Figure 7

In the control group 27 people out of 100 had a respiratory‐related hospital admission over 3 to 12 months, compared to 20 (95% CI 15 to 27) out of 100 for integrated disease management group.

Navigate to figure in ReviewOpen in new tab

Analysis 1.1

Comparison 1 Integrated disease management versus control, Outcome 1 SGRQ: short‐term (3 to 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.2

Comparison 1 Integrated disease management versus control, Outcome 2 SGRQ: long‐term (> 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.3

Comparison 1 Integrated disease management versus control, Outcome 3 Subgroup analysis SGRQ (total score) based on type of setting.

Navigate to figure in ReviewOpen in new tab

Analysis 1.4

Comparison 1 Integrated disease management versus control, Outcome 4 Subgroup analysis SGRQ (total score) based on type of study design.

Navigate to figure in ReviewOpen in new tab

Analysis 1.5

Comparison 1 Integrated disease management versus control, Outcome 5 Subgroup analysis SGRQ (total score) based on type of control group.

Navigate to figure in ReviewOpen in new tab

Analysis 1.6

Comparison 1 Integrated disease management versus control, Outcome 6 Subgroup analysis SGRQ (total score) based on dominant component of intervention.

Navigate to figure in ReviewOpen in new tab

Analysis 1.7

Comparison 1 Integrated disease management versus control, Outcome 7 CRQ: short‐term (3 to 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.8

Comparison 1 Integrated disease management versus control, Outcome 8 CRQ: Long‐term (> 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.9

Comparison 1 Integrated disease management versus control, Outcome 9 General health QoL: SIP mean difference.

Navigate to figure in ReviewOpen in new tab

Analysis 1.10

Comparison 1 Integrated disease management versus control, Outcome 10 Functional exercise capacity: 6MWD mean difference.

Navigate to figure in ReviewOpen in new tab

Analysis 1.11

Comparison 1 Integrated disease management versus control, Outcome 11 Subgroup analysis 6MWD based on type of setting.

Navigate to figure in ReviewOpen in new tab

Analysis 1.12

Comparison 1 Integrated disease management versus control, Outcome 12 Subgroup analysis 6MWD based on type of control group.

Navigate to figure in ReviewOpen in new tab

Analysis 1.13

Comparison 1 Integrated disease management versus control, Outcome 13 Subgroup analysis 6MWD based on dominant component of intervention.

Navigate to figure in ReviewOpen in new tab

Analysis 1.14

Comparison 1 Integrated disease management versus control, Outcome 14 Maximal exercise capacity: cycle test (W‐max).

Navigate to figure in ReviewOpen in new tab

Analysis 1.15

Comparison 1 Integrated disease management versus control, Outcome 15 Number of patients experiencing at least one exacerbation: short‐term (3‐12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.16

Comparison 1 Integrated disease management versus control, Outcome 16 Number of patients experiencing at least one exacerbation: long‐term (> 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.17

Comparison 1 Integrated disease management versus control, Outcome 17 All hospital admissions: short‐term (3 to 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.18

Comparison 1 Integrated disease management versus control, Outcome 18 All hospital admissions: long‐term (> 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.19

Comparison 1 Integrated disease management versus control, Outcome 19 Respiratory‐related hospital admissions: short‐term (3 to 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.20

Comparison 1 Integrated disease management versus control, Outcome 20 Respiratory‐related hospital admissions: long‐term (> 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.21

Comparison 1 Integrated disease management versus control, Outcome 21 Hospital days per patient (all causes): short‐term (3 to 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.22

Comparison 1 Integrated disease management versus control, Outcome 22 Hospital days per patient: long‐term (> 12 months).

Navigate to figure in ReviewOpen in new tab

Analysis 1.23

Comparison 1 Integrated disease management versus control, Outcome 23 ED visits.

Navigate to figure in ReviewOpen in new tab

Analysis 1.24

Comparison 1 Integrated disease management versus control, Outcome 24 Number of patients using at least one course of oral steroids.

Navigate to figure in ReviewOpen in new tab

Analysis 1.25

Comparison 1 Integrated disease management versus control, Outcome 25 Number of patients using at least one course of antibiotics.

Navigate to figure in ReviewOpen in new tab

Analysis 1.26

Comparison 1 Integrated disease management versus control, Outcome 26 MRC dyspnea score.

Navigate to figure in ReviewOpen in new tab

Analysis 1.27

Comparison 1 Integrated disease management versus control, Outcome 27 Borg score.

Navigate to figure in ReviewOpen in new tab

Analysis 1.28

Comparison 1 Integrated disease management versus control, Outcome 28 Mortality.

Navigate to figure in ReviewOpen in new tab

Analysis 1.29

Comparison 1 Integrated disease management versus control, Outcome 29 FEV1 (liter).

Navigate to figure in ReviewOpen in new tab

Analysis 1.30

Comparison 1 Integrated disease management versus control, Outcome 30 FEV1 (% predicted).

Navigate to figure in ReviewOpen in new tab

Analysis 1.31

Comparison 1 Integrated disease management versus control, Outcome 31 Anxiety and depression (HADS).

Navigate to figure in ReviewOpen in new tab

Summary of findings for the main comparison. Integrated disease management compared to control for patients with chronic obstructive pulmonary disease

Integrated disease management compared to control for patients with chronic obstructive pulmonary disease
Patient or population: patients with chronic obstructive pulmonary disease Settings: 8 studies in primary care, 12 studies in secondary care, 1 study in tertiary care, 5 studies in both primary and secondary care Intervention: integrated disease management Comparison: control
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Disease management
Quality of life measured on the SGRQ (St George's Respiratory Questionnaire) total score. Scale from: 0 to 100. Lower score indicates improvement Follow‐up: 3 to 12 months	The mean change in the SGRQ (total score) ranged from 3.4 lower to 6.24 higher	The mean SGRQ (total score) in the intervention groups was 3.71 lower (5.83 to 1.59 lower)	MD ‐3.71 (‐5.83 to ‐1.59)	1425 (13 studies)	⊕⊕⊕⊕ high²	MCID = ‐4 points, lower score means improvement
Quality of life measured on the CRQ dyspnoea domain Scale from: 0 to 7. Higher score indicates improvement Follow‐up: 3 to 12 months	The mean change in the CRQ (dyspnoea domain) ranged from 0 to 0.2 lower	The mean CRQ dyspnoea domain in the intervention groups was 1.02 higher (0.67 to 1.36 higher)	MD 1.02 (0.67 to 1.36)	160 (4 studies)	⊕⊕⊕⊝ moderate¹	MCID = 0.5 points Results on the other domains of the CRQ (fatigue, emotion, mastery) were also all statistically and clinically relevant
Functional exercise capacity 6‐minute walking distance (6MWD) Follow‐up: 3 to 12 months	The mean change in the 6MWD ranged from 38 lower to 36 higher	The mean functional exercise capacity in the intervention groups was 43.86 higher (21.83 to 65.89 higher)	MD 43.86 (21.83 to 65.89)	838 (14 studies)	⊕⊕⊕⊝ moderate³	MCID = 35 meters. Sensitivity analysis did show there was inconsistency in the effect. After removing low‐quality studies, the MD was 15.15 meters (95% CI 6.37 to 23.93, P < 0.001)
Respiratory‐related hospital admissions Follow‐up: 3 to 12 months	27 per 100	20 per 100 (15 to 27)	OR 0.68 (0.47 to 0.99)	1470 (7 studies)	⊕⊕⊕⊕ high
Hospital days per patient (all causes) Follow‐up: 3 to 12 months	The mean change in hospital days ranged from 1.6 to 11.9 higher	The mean number of hospital days per patient (all causes) in the intervention groups was 3.78 lower (5.9 to 1.67 lower)	MD ‐3.78 (‐5.9 to ‐1.67)	741 (6 studies)	⊕⊕⊕⊕ high
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; IDM: integrated disease management; MCID: minimal clinically important difference; MD: mean difference; OR: odds ratio
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹We downgraded one as there was considerable risk of bias in two studies on allocation concealment and two studies did not blind the outcome assessor. ²We did not downgrade due to risk of bias, as studies contributing more than 2.7% to the meta‐analysis had a low risk of bias. Sensitivity analysis on high‐risk studies did not change the effect or significance of the effect. ³We downgraded one as all included studies were of moderate to low quality. If we removed studies which had high or unclear risk of bias on allocation concealment, the effect decreased to 15 meters.

Summary of findings for the main comparison. Integrated disease management compared to control for patients with chronic obstructive pulmonary disease

Navigate to table in Review

Table 1. Characteristics of included studies

Study	Country	N (randomised)	N (completed)	Number of components intervention	Number of health care providers	Main component intervention	Setting	Control group
Aiken 2006	US	41	18	5	2	SF	PRIM	U
Bendstrup 1997	Denmark	42	32	4	7	E	SEC	U
Bourbeau 2003	Canada	191	165	4	4	SM	SEC	U
Boxall 2005	Australia	60	46	2	3	E	PRIM	U
Cambach 1997	Netherlands	43	23	2	2	E	PRIM	DRUG
Dheda 2004	UK	33	25	4	2	SF	SEC	U
Engstrom 1999	Sweden	55	50	4	5	E	SEC	U
Farrero 2001	Spain	122	94	2	2	SF	SEC	U
Fernandez 2009	Spain	50	41	2	2	E	PRIM	EDU
Gottlieb 2011	Denmark	61	26	4	Multidisciplinary team, not specified	E	PRIM	U
Güell 2000	Spain	60	47	3	3	E	SEC	U
Güell 2006	Spain	40	25	2	4	E	TERT	DRUG
Koff 2009	US	40	38	4	2	SM	PRIM	U
Littlejohns 1991	UK	152	133	4	3	SF	SEC	U
Mendes 2010	Brazil	117	85	2	2	E	PRIM/SEC	U
Rea 2004	New Zealand	135	117	5	4	SM/SF	PRIM/SEC	U
Rice 2010	US	743	743	3	2	SM	SEC	EDU
Smith 1999	Australia	96	36	8	3	SF	PRIM/SEC	U
Sridhar 2008	UK	122	104	4	3	E/SM	PRIM/SEC	U
Strijbos 1996	Netherlands	50	41	3	3	E	PRIM/SEC	U
Theander 2009	Sweden	30	26	4	4	E	SEC	U
Trappenburg	Netherlands	233	193	3	3	SM	SEC	U
Wakabayashi 2011	Japan	102	85	4	2	IT EDU	SEC	EDU
Wetering 2010	Netherlands	199	175	4	3	E	SEC	U
Wijkstra 1995	Netherlands	45	43	2	3	E	PRIM	U
Wood‐Baker 2006	Australia	135	112	3	2	SM	PRIM	EDU
Main component: SF: structural follow‐up; SM: self management; E: exercise; IT EDU: individually tailored education Setting: PRIM: primary care; SEC: secondary care; TERT: tertiary care Control group: U: usual care; DRUG: optimization of drug treatment; EDU: education

Table 1. Characteristics of included studies

Navigate to table in Review

Table 2. Components of IDM in each included study

Author	Education	Self management	Exacerbation/action plan	Exercise	Psychosocial/occupational	Smoking	Optimal medication	Nutrition	Follow‐up	Case management	Multidisciplinary
Aiken 2006	x	x	x				x			x
Bendstrup 1997	x			x	x	x
Bourbeau 2003	x		x	x					x
Boxall 2005	x			x
Cambach 1997	x			x
Dheda 2004	x				x		x		x
Engstrom 1999	x			x	x			x
Farrero 2001									x	x
Fernandez	x			x
Gottlieb 2011	x			x		x		x
Güell 2000	x			x					x
Güell 2006	x			x
Koff 2009	x	x	x						x
Littlejohns 1991	x						x		x	x
Mendes 2010	x			x
Rea 2004	x		x	x					x		x
Rice 2010	x	x	x
Smith 1999	x	x	x	x		x	x			x	x
Sridhar 2008	x		x	x					x
Strijbos 1996	x			x					x
Theander 2009	x			x	x			x
Trappenburg 2011	x	x	x
Wakabayashi 2011	x	x	x			x
Wetering 2010	x			x		x		x
Wijkstra 1995	x			x
Wood‐Baker 2006	x	x	x

Table 2. Components of IDM in each included study

Navigate to table in Review

Comparison 1. Integrated disease management versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 SGRQ: short‐term (3 to 12 months) Show forest plot	13		Mean Difference (IV, Random, 95% CI)	Subtotals only

1.1 SGRQ: Total	13	1425	Mean Difference (IV, Random, 95% CI)	‐3.71 [‐5.83, ‐1.59]
1.2 SGRQ: Symptoms	11	1377	Mean Difference (IV, Random, 95% CI)	‐2.39 [‐5.31, 0.53]
1.3 SGRQ: Activity	11	1352	Mean Difference (IV, Random, 95% CI)	‐2.70 [‐4.84, ‐0.55]
1.4 SGRQ: Impact	11	1355	Mean Difference (IV, Random, 95% CI)	‐4.04 [‐5.96, ‐2.11]
2 SGRQ: long‐term (> 12 months) Show forest plot	2		Mean Difference (IV, Random, 95% CI)	Subtotals only

2.1 SGRQ: Total	2	189	Mean Difference (IV, Random, 95% CI)	‐0.22 [‐7.43, 6.99]
2.2 SGRQ: Symptoms	2	191	Mean Difference (IV, Random, 95% CI)	5.65 [‐8.88, 20.18]
2.3 SGRQ: Activity	2	191	Mean Difference (IV, Random, 95% CI)	‐2.13 [‐7.89, 3.63]
2.4 SGRQ: Impact	2	182	Mean Difference (IV, Random, 95% CI)	‐1.62 [‐5.50, 2.25]
3 Subgroup analysis SGRQ (total score) based on type of setting Show forest plot	13	1425	Mean Difference (IV, Random, 95% CI)	‐3.77 [‐5.90, ‐1.64]

3.1 Primary care	6	456	Mean Difference (IV, Random, 95% CI)	‐4.68 [‐8.80, ‐0.56]
3.2 Secondary care	7	969	Mean Difference (IV, Random, 95% CI)	‐3.41 [‐5.97, ‐0.85]
4 Subgroup analysis SGRQ (total score) based on type of study design Show forest plot	13	1425	Mean Difference (IV, Random, 95% CI)	‐3.71 [‐5.83, ‐1.59]

4.1 RCTs	12	1304	Mean Difference (IV, Random, 95% CI)	‐4.22 [‐6.14, ‐2.30]
4.2 CRCTs	1	121	Mean Difference (IV, Random, 95% CI)	2.3 [‐1.62, 6.22]
5 Subgroup analysis SGRQ (total score) based on type of control group Show forest plot	13	1425	Mean Difference (IV, Random, 95% CI)	‐3.71 [‐5.83, ‐1.59]

5.1 Control group: usual care	9	744	Mean Difference (IV, Random, 95% CI)	‐4.09 [‐6.35, ‐1.84]
5.2 Control group: mono‐disciplinary treatment	4	681	Mean Difference (IV, Random, 95% CI)	‐2.98 [‐7.69, 1.74]
6 Subgroup analysis SGRQ (total score) based on dominant component of intervention Show forest plot	11	1315	Mean Difference (IV, Random, 95% CI)	‐3.61 [‐5.67, ‐1.55]

6.1 Dominant component self management	5	942	Mean Difference (IV, Random, 95% CI)	‐2.76 [‐5.88, 0.36]
6.2 Dominant component exercise	6	373	Mean Difference (IV, Random, 95% CI)	‐4.74 [‐7.05, ‐2.43]
7 CRQ: short‐term (3 to 12 months) Show forest plot	4		Mean Difference (IV, Random, 95% CI)	Subtotals only

7.1 CRQ: Dyspnea	4	160	Mean Difference (IV, Random, 95% CI)	1.02 [0.67, 1.36]
7.2 CRQ: Fatigue	4	161	Mean Difference (IV, Random, 95% CI)	0.82 [0.46, 1.17]
7.3 CRQ: Emotion	4	161	Mean Difference (IV, Random, 95% CI)	0.61 [0.26, 0.95]
7.4 CRQ: Mastery	4	161	Mean Difference (IV, Random, 95% CI)	0.75 [0.38, 1.12]
8 CRQ: Long‐term (> 12 months) Show forest plot	2		Mean Difference (IV, Random, 95% CI)	Subtotals only

8.1 CRQ: Dyspnea	2	151	Mean Difference (IV, Random, 95% CI)	0.47 [‐0.31, 1.25]
8.2 CRQ: Fatigue	2	151	Mean Difference (IV, Random, 95% CI)	0.45 [0.05, 0.85]
8.3 CRQ: Emotion	2	151	Mean Difference (IV, Random, 95% CI)	0.53 [0.10, 0.95]
8.4 CRQ: Mastery	2	151	Mean Difference (IV, Random, 95% CI)	0.80 [0.37, 1.23]
9 General health QoL: SIP mean difference Show forest plot	2		Mean Difference (IV, Random, 95% CI)	Subtotals only

9.1 SIP total	2	183	Mean Difference (IV, Random, 95% CI)	‐1.06 [‐3.00, 0.89]
9.2 SIP: physical	2	183	Mean Difference (IV, Random, 95% CI)	‐2.63 [‐5.55, 0.30]
9.3 SIP: psychosocial	2	183	Mean Difference (IV, Random, 95% CI)	‐0.86 [‐3.17, 1.44]
10 Functional exercise capacity: 6MWD mean difference Show forest plot	14		Mean Difference (IV, Random, 95% CI)	Subtotals only

10.1 6MWD: short‐term (3 to 12 months)	14	871	Mean Difference (IV, Random, 95% CI)	43.86 [21.83, 65.89]
10.2 6MWD: long‐term (> 12 months)	2	184	Mean Difference (IV, Random, 95% CI)	16.84 [3.01, 30.67]
11 Subgroup analysis 6MWD based on type of setting Show forest plot	14		Mean Difference (IV, Random, 95% CI)	Subtotals only

11.1 Primary care	7	427	Mean Difference (IV, Random, 95% CI)	45.16 [8.65, 81.67]
11.2 Secondary care	7	438	Mean Difference (IV, Random, 95% CI)	49.18 [14.28, 84.08]
11.3 Tertiary care	1	35	Mean Difference (IV, Random, 95% CI)	85.0 [30.43, 139.57]
12 Subgroup analysis 6MWD based on type of control group Show forest plot	14	871	Mean Difference (IV, Random, 95% CI)	43.86 [21.83, 65.89]

12.1 Control group: mono disciplinary treatment	4	180	Mean Difference (IV, Random, 95% CI)	35.99 [‐5.34, 77.31]
12.2 Control group: usual care	10	691	Mean Difference (IV, Random, 95% CI)	46.59 [19.68, 73.51]
13 Subgroup analysis 6MWD based on dominant component of intervention Show forest plot	14	871	Mean Difference (IV, Random, 95% CI)	43.86 [21.83, 65.89]

13.1 Dominant component: exercise	12	653	Mean Difference (IV, Random, 95% CI)	51.47 [26.53, 76.40]
13.2 Dominant component: structured follow‐up	1	133	Mean Difference (IV, Random, 95% CI)	3.50 [‐28.31, 35.31]
13.3 Dominant component: individually tailored education program	1	85	Mean Difference (IV, Random, 95% CI)	0.4 [‐39.64, 40.44]
14 Maximal exercise capacity: cycle test (W‐max) Show forest plot	4	298	Mean Difference (IV, Random, 95% CI)	6.99 [2.96, 11.02]

15 Number of patients experiencing at least one exacerbation: short‐term (3‐12 months) Show forest plot	2	407	Odds Ratio (M‐H, Random, 95% CI)	1.21 [0.77, 1.91]

16 Number of patients experiencing at least one exacerbation: long‐term (> 12 months) Show forest plot	2	301	Odds Ratio (M‐H, Fixed, 95% CI)	1.53 [0.90, 2.60]

17 All hospital admissions: short‐term (3 to 12 months) Show forest plot	2	266	Odds Ratio (M‐H, Random, 95% CI)	0.62 [0.36, 1.07]

18 All hospital admissions: long‐term (> 12 months) Show forest plot	2	283	Odds Ratio (M‐H, Random, 95% CI)	0.78 [0.38, 1.57]

19 Respiratory‐related hospital admissions: short‐term (3 to 12 months) Show forest plot	7	1470	Odds Ratio (M‐H, Random, 95% CI)	0.68 [0.47, 0.99]

20 Respiratory‐related hospital admissions: long‐term (> 12 months) Show forest plot	1	179	Odds Ratio (M‐H, Random, 95% CI)	0.59 [0.28, 1.22]

21 Hospital days per patient (all causes): short‐term (3 to 12 months) Show forest plot	6	741	Mean Difference (IV, Random, 95% CI)	‐3.78 [‐5.90, ‐1.67]

22 Hospital days per patient: long‐term (> 12 months) Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

23 ED visits Show forest plot	4	1161	Odds Ratio (M‐H, Random, 95% CI)	0.64 [0.33, 1.25]

24 Number of patients using at least one course of oral steroids Show forest plot	3	348	Odds Ratio (M‐H, Random, 95% CI)	1.13 [0.64, 2.01]

25 Number of patients using at least one course of antibiotics Show forest plot	2	236	Odds Ratio (M‐H, Random, 95% CI)	1.43 [0.24, 8.48]

26 MRC dyspnea score Show forest plot	3	345	Mean Difference (IV, Random, 95% CI)	‐0.30 [‐0.48, ‐0.11]

27 Borg score Show forest plot	3	145	Mean Difference (IV, Random, 95% CI)	0.14 [‐0.70, 0.98]

28 Mortality Show forest plot	5	1235	Odds Ratio (M‐H, Random, 95% CI)	0.85 [0.49, 1.46]

28.1 Short‐term (3 to 12 months)	4	1113	Odds Ratio (M‐H, Random, 95% CI)	0.96 [0.52, 1.74]
28.2 Long‐term (> 12 months)	1	122	Odds Ratio (M‐H, Random, 95% CI)	0.45 [0.16, 1.28]
29 FEV1 (liter) Show forest plot	3		Mean Difference (IV, Random, 95% CI)	Subtotals only

29.1 FEV1 (liter): short‐term	2	234	Mean Difference (IV, Random, 95% CI)	0.00 [‐0.14, 0.14]
29.2 FEV1 (liter): long‐term	1	104	Mean Difference (IV, Random, 95% CI)	‐0.11 [‐0.28, 0.06]
30 FEV1 (% predicted) Show forest plot	9		Mean Difference (IV, Random, 95% CI)	Subtotals only

30.1 FEV1 (% predicted): short‐term	4	280	Mean Difference (IV, Random, 95% CI)	‐0.57 [‐3.54, 2.39]
30.2 FEV1 (% predicted; mean change): short‐term	4	514	Mean Difference (IV, Random, 95% CI)	2.15 [0.38, 3.91]
30.3 FEV1 (% predicted): long‐term	1	104	Mean Difference (IV, Random, 95% CI)	‐4.60 [‐11.26, 2.06]
31 Anxiety and depression (HADS) Show forest plot	2		Mean Difference (IV, Random, 95% CI)	Subtotals only

31.1 HADS: depression	2	316	Mean Difference (IV, Random, 95% CI)	0.21 [‐0.39, 0.81]
31.2 HADS: anxiety	2	316	Mean Difference (IV, Random, 95% CI)	0.22 [‐0.41, 0.85]

Comparison 1. Integrated disease management versus control

Navigate to table in Review

Cochrane Review language

Website language

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

PICOs

PICOs

Population

Intervention

Comparison

Outcome

Plain language summary

Integrated disease management for chronic obstructive pulmonary disease

Visual summary

Authors' conclusions

Implications for practice

Implications for research

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Clinical outcomes

Process‐related outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Results

Description of studies

Results of the search

Included studies

Participants

Interventions

Outcomes

Excluded studies

Risk of bias in included studies

Allocation

Blinding

Incomplete outcome data

Selective reporting

Other potential sources of bias

Effects of interventions

Primary outcomes

1. Quality of life

1.1 Respiratory‐specific QoL

1.1.1.1 SGRQ total score ‐ short‐term

1.1.1.2. SGRQ ‐ long‐term

1.1.2.1 SGRQ domain scores ‐ short‐term

1.1.2.2. SGRQ domain scores ‐ long‐term

1.1.3.1. CRQ domain scores ‐ short‐term

1.1.3.2. CRQ domain scores ‐ long‐term

1.2 General health‐related QoL

2. Exercise capacity