Continuity of Care to Prevent Readmissions for Patients with Chronic Obstructive Pulmonary Disease: A Systematic Review and Meta-Analysis

ABSTRACT

Readmissions of patients with chronic obstructive pulmonary disease (COPD) to hospitals cast a heavy burden to health care systems. This meta-analysis was aimed to assess the efficacy of continuity of care as interventions, which reduced readmission and mortality rates of such patients. PubMed, Cochrane Library and Embase were searched for articles published before July 2015. A total of 31 reports with randomized controlled trials (RCTs) were finally included in this meta-analysis. The results showed that health education reduced all-cause readmission at 3 months. In addition, health education, comprehensive nursing intervention (CNI) and telemonitoring reduced all-cause readmissions over 6–12 months, and the effect of CNI was best because CNI also reduced COPD-specific readmissions. Home visits also reduced COPD-specific readmissions (the quality more than moderate), but it did not reduce the risk for all-cause readmissions (risk ratios (RRs), 0.92 [95% CI, 0.82–1.04]; moderate quality). There was no statistically significant difference in reducing mortality and quality of life (QOL) among various continued cares. In conclusion, CNI, telemonitoring, health education and home visits should receive more consideration than other interventions by caregivers seeking to implement continued care interventions for patients with COPD.

KEYWORDS:

• Chronic obstructive pulmonary disease
• continuity of care
• interventions
• readmissions

Introduction

COPD is a major global chronic progressive disease affecting millions of people worldwide (1). This disease causes considerable morbidity, hospital readmission and mortality worldwide (2–8). According to the World Health Organization, more than 67 million people had COPD and over 3 million people died of COPD in 2005, which represented 5% of all deaths globally in 2011, and total deaths from COPD are projected to increase by more than 30% in the following 10 years (9). By 2030, COPD is predicted to be the third leading cause of death worldwide (10). As COPD mainly affects old people, the associated disease burden is expected to increase in the coming decades due to the ageing global population. Potentially affecting one in four adults by the age of 80 years (11), COPD is already a leading cause of death in developed and developing countries (12,13).

COPD casts a heavy burden on healthcare systems worldwide (4,14) and on individuals as a result of impaired QOL (15,16). The majority of COPD healthcare costs are accounted for by hospital admissions (17). An audit of COPD in England found that nearly one-third of patients had a readmission within 90 days (18). One study reported that 18% of COPD patients were readmitted once and 14% twice within 1 year (19). Successful reduction of COPD rehospitalisation rates would yield important economic benefits. Therefore, COPD has been regarded as a major public health concern (20,21), requiring management strategies after discharges to prevent early rehospitalisations and reduce healthcare costs (22).

Continuity of care is recognised as a core prevention of readmissions and a potential improvement of health outcomes for patients with COPD (23,24). Continuity of care means that the care is provided after a patient is discharged from a hospital. Hereafter, it is interchangeably referred to as continued care in this paper. Although there are no clear compositions of continued care, they often include health education, medication reconciliation, self-management project and adjustment of the nursing measures according to the patient condition such as telemonitoring. The telemonitoring interventions used equipment installed in discharged patients' homes to remotely monitor the patients' biophysical data to assess their health condition (25). Unlike other chronic conditions (26,27), the efficacy of continued care to prevent COPD from rehospitalisation has been controversial (28–31). Two RCTs did not show positive effects (30,31), and a systematic review showed that insufficient evidence supported self-management interventions (32). A study that reported a continuum of self-management in a comprehensive patient education program could significantly reduce the utilization of health care services and improve health status (28), which contained weekly visits over a 2-month period and monthly telephone follow-ups. Rees (33) claimed that the self-management program (1-hour weekly home self-management program lasted 7–8 weeks, a customized action plan, an exercise program and telephone case manager) might be the major reason for readmission reductions and health improvement. Gallefoss argued that an education of mild-to-moderate COPD patients could improve their outcomes and reduce costs in a 1-year follow-up (29). Other researchers also reported that health education interventions could reduce hospital readmission (28,34–36). One comprehensive systematic review suggested that disease management programs for patients with COPD reduced hospitalisation and slightly improved QOF (37). However, other systematic reviews have reached different conclusions about the overall value of disease management for COPD, but all reviews have recognized the potential value of such interventions and the need for more detailed studies (38,39).

Most of these studies only evaluated the efficacy of a single intervention. In addition to inconsistent findings, few studies in the literature were comprehensive evaluations of the efficacy of multiple continuous care interventions. Therefore, the aims of this study were to update and synthesize findings on continued care interventions for patients with COPD and evaluate which intervention was the most effective in improving patients' outcomes.

Methods

Data sources and searches

Reports published before July 2015 that involved only humans were searched in the Embase, Cochrane Library and PubMed (including MEDLINE). References of the selected studies were also manually checked to identify additional relevant studies.

Study selection

Inclusion of each study was determined independently by two reviewers, and inclusion criteria were evaluated. Each study had to meet four criteria. First, it had to be based on an RCT design and reported in full text with a title and abstract. Second, it had to include adults with a clinical diagnosis of COPD and compare a continued care with other eligible interventions or usual care, (namely, routine or standard care, as defined by the primary studies). Third, it had to have the interventions with one or more of the following components: health education before or after discharge, discharge action plans, planned or scheduled home visits, frequent telephone contact, self-management, telemonitoring, CNI (health education, self-management, action plan and home visits/follow-up telephone), home-based rehabilitation (HBR), or interventions to increase provider continuity. Fourth, the primary outcome for patients had to include a readmission or mortality rate, or a composite outcome (all-cause readmission or mortality). Secondary outcomes included QOF, hospital days of subsequent readmissions, subjective health status and caregiver or self-care cost. Studies that examined only health care service expenditures and costs were excluded.^{Figure 1}

Data extraction and quality assessment

Each article that met these inclusion criteria was independently scrutinized by two reviewers to extract a description of objectives, design, participants, interventions and follow-up time. Any disagreement was resolved by discussion among the reviewers, and a final decision was made by the third reviewer.

The Cochrane Group's predesigned table (40) was used to ensure standardized scoring. Methodological quality was assessed in terms of selection, confounding, performance, attrition and detection biases. The two reviewers independently assessed risks of biases for each study. Disagreements were resolved by consensus from discussions. Studies were given a score of one point for each fulfilled criterion. Based on the scores, the quality of the studies was divided into three levels: low (≤3 points), moderate (4–6 points) and high (7–9 points). All individual patients' data were checked for consistency internally and with published reports.

Data synthesis and analysis

A meta-analysis was conducted with DerSimonian–Laird random effects models. The outcomes reported by similar multiple studies were combined for the analysis. Also the meta-analyzed RCTs were reported the number of deaths or number of persons readmitted in each group. If the number of persons readmitted could not be obtained, the articles were excluded.^{Table 1Table 2Figure 2Figure 3Figure 4Figure 5}

Table 1. Quality assessments of the included studies.

Study and year	Randomization	Assignment concealment	Similarity at baseline	Inclusion/exclusion criteria	Researchers/participants blinded	Blinding of assessors	Attrition rate reported	Participants lost to follow-up described	Intention-to-Treat analysis	Power analysis	Total
Wood-Baker et al. (63)	Yes	No	Yes	Yes	No	No	Yes	Yes	Yes	Yes	7
Weinberger et al. (65)	Yes	No	Yes	Yes	No	No	Yes	Yes	Yes	Yes	7
Martin et al. (64)	Yes	No	Yes	Yes	No	No	Yes	Yes	No	No	5
McGeoch et al. (60)	Yes	No	Yes	Yes	No	No	Yes	Yes	No	Yes	6
Wei et al. (61)	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	No	8
Tommelein et al. (62)	Yes	No	Yes	Yes	Yes	No	Yes	Yes	No	Yes	7
Kenealy et al. (43)	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	9
Paré et al. (25)	Yes	Yes	Yes	Yes	No	No	Yes	Yes	No	No	6
Pedone et al. (54)	Yes	No	Yes	Yes	No	No	Yes	No	Yes	Yes	6
Pinnock et al. (55)	Yes	Yes	Yes	Yes	Yes	No	Yes	No	No	Yes	7
Bentley et al. (44)	Yes	No	Yes	Yes	No	No	Yes	Yes	Yes	No	6
de Toledo et al. (46)	Yes	No	Yes	Yes	No	No	No	No	No	No	3
Dinesen et al. (56)	Yes	Yes	Yes	Yes	Yes	No	Yes	No	Yes	Yes	8
Koff et al. (69)	Yes	Yes	Yes	Yes	No	No	Yes	Yes	No	No	6
Khdour et al. (36)	Yes	No	Yes	Yes	No	No	Yes	Yes	No	Yes	6
Fan et al. (58)	Yes	No	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	8
Kruis et al. (57)	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	9
Casas et al. (47)	Yes	No	Yes	Yes	No	No	Yes	Yes	Yes	Yes	7
Rice et al. (59)	Yes	No	Yes	Yes	No	No	No	No	No	Yes	4
Smith et al. (50)	Yes	No	Yes	Yes	No	No	No	Yes	Yes	Yes	6
Hermiz et al. (30)	Yes	No	Yes	Yes	No	No	Yes	Yes	No	Yes	6
Farrero et al. (51)	Yes	No	Yes	Yes	No	No	No	No	No	No	3
Cotton et al. (48)	Yes	No	Yes	Yes	No	No	Yes	Yes	Yes	No	6
Utens et al. (52)	Yes	Yes	Yes	Yes	No	No	Yes	No	Yes	Yes	7
Skwarska et al. (53)	Yes	No	Yes	Yes	No	No	No	Yes	No	No	4
Bourbeau et al. (28)	Yes	No	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	8
Cockcroft et al. (49)	Yes	No	Yes	Yes	Yes	No	No	No	No	No	4
Lainscak et al. (45)	Yes	No	Yes	Yes	No	No	No	No	Yes	Yes	5
Behnke et al. (68)	Yes	No	Yes	Yes	No	No	No	No	No	No	3
Eaton et al. (67)	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	9
Seymour et al. (66)	Yes	No	Yes	Yes	No	No	No	No	Yes	Yes	5

Risk ratios (RRs) were calculated for readmission and mortality rates. Statistical heterogeneity was measured with the chi-square and I² statistics (41). Begg's plot was used to test the publication bias. The meta-analyses were carried out using Stata (version 12.0). Subgroup analyses were planned on the type of intervention and the time of follow-up. Health-related QOF was assessed using a standardized self-completed questionnaire, namely, the St. George's Respiratory Questionnaire (SGRQ) (42). Weighted mean differences were estimated using a random-effects model.

Results

The literature flow diagram (Figure 1) summarizes the search and selection of articles. From the 1,451 potentially relevant citations initially identified, 1,129 were excluded. The remaining 322 retrieved reports were selected for in-depth screening and 287 studies for detailed evaluation. A total of 31 RCTs were finalized as eligible for the analysis. Seven RCTs from four countries (New Zealand, Spain, England, Slovenia) both reported the primary outcome of COPD readmission and mortality (43–49). All studies reported readmission, but some did other outcomes. Begg's plot was used to examine publication bias, and the results showed that there was no evident publication bias (p = 1.000).

Figure 1. Summary of evidence search and publication selection.

Methodological quality

Overall methodological quality of the studies (Table 1) was relatively high, with scores ranging from 3 to 9. Three articles scored 3, 15 did less than or equal to 6 and 13 did no more than 9. All the studies were randomized, with inclusion/exclusion criteria and similarity at baselines. The most common reason for lower scores was the absence of double-blind procedure, which was impossible due to the nature of the intervention. The assessors were not blinded to outcome in 26 studies (83.9%), and researchers/participants were not blinded in 23 studies (74.2%). Eleven studies (35.5%) did not report the characteristics of participants lost to follow-ups. Only eight studies (25.8%) involved an intervention group of more than 100 participants and concealed allocation.

Basic characteristics of included studies

RCT characteristics are shown in Table 2. Most RCTs compared a continued care with a usual care. Twenty studies were single-centre RCTs, and eleven studies were conducted in multicentres. Twenty-two RCTs were conducted in European and American countries, nine were in Australia and one was in China. The sample sizes of the included studies ranged from 14 to 554 participants for the intervention groups and from 12 to 532 patients for the control groups. The mean age of study participants varied from 63 to 74 and was under 70 years in 20 studies and 70 years or above in eleven studies. Only seven studies reported the patients' smoking history. Follow-up was conducted for a range of 3–12 months and a mean of 8 months. The presence of lung dysfunction was indicated by a mean FEV¹ /FVC or FEV¹ (%Predicted) of less than 70% in 20 RCTs and non-reported in 11 RCTs.

Table 2. Description of the included trails.

			Participants(n)
Source	Country	Setting	Intervention group	Control group	Smoking history (pack-years)	Mean age (years)	FEV^1 /FVC%, mean (SD)	Duration of follow-up (months)	Intervention
Wood-Baker et al. (63)	Australia	Single-center	67	72	>10	70	54(14)	12	Action plan
Weinberger et al. (65)	England USA	Multicenter	295	288	NR	63	NR	6	Action plan
Martin et al. (64)	New Zealand	Single-center	44	49	NR	70	35(2)^*	12	Action plan
McGeoch et al. (60)	New Zealand	Single-center	84	70	>10	71	51(11)	12	Health education
Wei et al. (61)	China	Single-center	42	45	NR	64	53(10)	12	Health education
Tommelein et al. (62)	Belgium	Single-center	371	363	≧10	68	NR	3	Health education
Kenealy et al. (43)	New Zealand	Multicenter	24	24	NR	67	NR	6	Telemonitoring
Paré et al. (25)	Canada	Multicenter	60	60	NR	68	NR	12	Telemonitoring
Pedone et al. (54)	Italy	Single-center	50	49	NR	74	69(9)	9	Telemonitoring
Pinnock et al. (55)	England	Single-center	128	128	NR	69	42(18)^*	12	Telemonitoring
Bentley et al. (44)	England	Single-center	23	25	NR	67	NR	6	Telemonitoring
de Toledo et al. (46)	Spain	Single-center	67	90	NR	71	NR	12	Telemonitoring
Dinesen et al. (56)	Denmark	Multicenter	57	48	NR	68	NR	10	Telemonitoring
Koff et al. (69)	USA	Single-center	19	19	NR	65	32(10)^*	3	Telemonitoring
Khdour et al. (36)	England	Single-center	71	72	NR	67	57(10)	12	CNI
Fan et al. (58)	USA	Multicenter	209	217	>10	66	47(12)	12	CNI
Kruis et al. (57)	Netherlands	Multicenter	554	532	NR	68	68(20)^*	12	CNI
Casas et al. (47)	Spain	Multicenter	65	90	NR	71	48(18)	12	CNI
Rice et al. (59)	USA	Multicenter	372	371	NR	70	37(14)^*	12	CNI
Smith et al. (50)	Australia	Single-center	48	48	NR	70	NR	12	Home visits
Hermiz et al. (30)	Australia	Multicenter	84	93	NR	67	NR	3	Home visits
Farrero et al. (51)	Spain	Single-center	60	62	NR	69	69(10)	12	Home visits
Cotton et al. (48)	England	Single-center	41	40	NR	67	46(2)	2	Home visits
Utens et al. (52)	New Zealand	Multicenter	70	69	≧10	68	48(2)^*	3	Home visits
Skwarska et al. (53)	England	Single-center	122	62	NR	69	NR	2	Home visits
Bourbeau et al. (28)	Canada	Multicenter	96	95	≧10	69	46	12	Home visits
Cockcroft et al. (49)	England	Single-center	42	33	NR	70	NR	12	Home visits
Lainscak et al. (45)	Slovenia	Single-center	118	135	NR	71	36(16)^*	6	Home visits
Behnke et al. (68)	Germany	Single-center	14	12	NR	67	36(7)	18	HBR
Eaton et al. (67)	New Zealand	Single-center	47	50	≧10	70	36(16)^*	3	HBR
Seymour et al. (66)	England	Single-center	30	30	NR	66	52(21)^*	3	HBR

NR, not reported; CNI, comprehensive nursing intervention: health education and self-management and action plan and home visits/follow-up telephone; HBR, home-based rehabilitation;

^*FEV¹ (%Predicted) (SD).

Interventions characteristics of included studies

Five RCTs compared home visits with the usual care (28,45,49,50). Four studies compared an early discharge home visiting program with the conventional hospital care (48,51–53). Most interventions involved home visits and health education; three RCT interventions added regulated monthly telephone calls (28,45,51). In most RCTs, community nurses conducted the home visits, most of which began within 7 days after discharge. Two RCTs included visits within 1–3 months of discharge (49,51), and only one RCT specified that visits were done within 14 days of discharge (50).

Seven RCTs compared home telemonitoring with usual care (25,43,46,53–56), and one RCT compared home telemonitoring with home visits (44). Five evaluated remote clinical data monitoring using equipment installed in patients' homes after discharge, which transmitted data to a central site (43,44,53,55,56). Most telemonitoring systems needed the patients to input the clinical data to assess their health condition. One RCT used a wristband that could get the data such as heart rate and couple with cellular telephone via a Bluetooth, and then, the telephone sent the data to the monitoring system (54). One study used cell phone call connected to the chronic care telemedicine system (46).

Five RCTs evaluated CNI (37,47,57–59). Most of the measures contained health education, disease self-management, action plan and home visits or telephone follow-up. One assessed enhanced access to individual practice plan on a network platform, such as smoking cessation, drug adherence and physiotherapeutic reactivation (57). RCTs that described disease management interventions emphasized more patients' participation and also had a multidisciplinary care team (nurses, physiotherapists and pharmacists).

Three RCTs evaluated a health education intervention. One compared the effects of 1-hour duration standardized self-management plan education from a practice nurse or respiratory educator in association with their general practitioner (60). One RCT investigated individualised education (20–30 minutes per session, 5–6 sessions) in a structured fashion step by step during the clinic visit and a series of 10-minute telephone counselling (61). One RCT featured education on inhalation technique and adherence to maintenance therapy at the start of the RCT by pharmacists (62).

Three RCTs evaluated action plan interventions (63–65). Two studies assessed patients' discharge needs and developed discharge action plans. One of them updated the action plan when it is necessary after discharge (65); the other one visited by the nurse at the study start and thereafter at 3, 6 and 12 months to provide routine support (64). One trail investigated a written self-management plan, which was developed in consultation with their treating general practitioner and patients (63).

Three RCTs evaluated HBR (66–68). Two RCTs used structured hospital-based supervised training such as walking and home-based training after discharged (67,68). One RCT investigated outpatients participated in standard pulmonary rehabilitation classes consisting of twice-weekly exercise, which was a mixture of limb strengthening and aerobic activities tailored to individual baseline function and exercised after discharge (66).

Readmission rates

Figures 2 and 3 presented our meta-analyses and RR calculations of RCTs reported all-cause readmission and COPD readmission, respectively. Results in both figures are stratified by types of post-discharge intervention and follow-up time.

Figure 2. All-cause readmission for transitional care compared with usual care. Weights are from the random-effects analysis.

Figure 3. COPD readmission for transitional care compared with usual care. Weights are from the random-effects analysis.

Readmission rates within 3 months after discharge

One RCT evaluating health education found a lower risk for readmission among patients receiving health education than that of the usual care group (RR, 0.25 [95% CI, 0.12–0.52]) (62). The intervention placed emphasis on educating inhalation technique, self-management and adherence to maintenance therapy at the start of the RCT by pharmacists, which had a high methodological quality.

Four RCTs about home visits did not reduce all-cause readmission rates, with quality scores ranging from 4 to 7. One of them found no difference in the risk for COPD-specific readmissions between persons receiving home visits at intervals of 2–3 days and those receiving inpatient care (53). The other RCTs found that patients with acute exacerbations of COPD discharged early, and supported home treatment by nurse did not reduce all-cause readmission rates (30,48,52).

Three other RCTs across different intervention categories reported no more than 3-month COPD-specific readmission: 1 telemonitoring RCT (69), 2 HBR RCTs (66,67). None of these interventions found a statistically significant reduction in COPD readmissions.

Readmission rates in 6–12 months after discharge

The interventions of health education (RR, 0.40 [95% CI, 0.27–0.59]), CNI (RR, 0.70 [95% CI, 0.63–0.48]) and telemonitoring (RR, 0.78 [95% CI, 0.58–0.88]) reduced all-cause readmissions over 6–12 months after discharge. Both CNI (RR, 0.42 [95% CI, 0.32–0.55]) and home visits (RR, 0.63 [95% CI, 0.47–0.85]) interventions reduced COPD readmissions in 6–12 months after discharge, with methodogical quality above moderate. The effects of CNI and home visit interventions were better than usual care. However, home visits did not reduce the risk for all-cause readmissions (RR, 0.92 [95% CI, 0.82–1.04]; moderate quality).

Readmission rates more than 12 months after discharge

One RCT found that patients with COPD who had an individually defined, 18-month home-based, long-term walking program initiated by a short hospital-based training, had a lower risk for all-cause readmission than those in the control group (RR, 0.24 [95% CI, 0.10–0.61]; low quality) (68). However, because of limited evidence from a single RCT and small sample (n = 26), we graded the evidence on increasing access to HBR as insufficient. Evidence was insufficient to determine whether HBR interventions were effective in reducing all-cause readmission.

Mortality rates

Figure 4 presents the mortality rates stratified by intervention category and follow-up time. There was no statistical significance in reducing mortality of various interventions and usual care. Telemonitoring, action plan and home visit interventions did not reduce mortality (high quality). Evidence for a reduction in mortality was insufficient for CNI interventions. Two studies demonstrated that there was no statistical significance in reducing mortality, but one study showed that the mortality of the patients in the intervention group was even higher than that of the usual care group (RR, 2.91 [95% CI, 1.54–5.83]; high quality).

Figure 4. All-cause mortality rate for transitional care compared with usual care. Weights are from the random-effects analysis.

QOF

As shown in Figure 5, it was possible that the interventions of action plan and home-based rehabilitation could improve patients' QOF. CNI and telemonitoring interventions found no statistically significant reduction in quality of life (moderate-to-high quality). Evidence was insufficient to determine whether health education could improve patients' QOF.

Figure 5. Quality of life (QOL) scores with transitional care compared with usual care. Weights are from the random-effects analysis.

Sensitivity analysis

Sensitivity analysis comparing fixed and random-effects statistical models yielded relatively same results (RR = 0.82; p = 0.000). The effect of sequentially omitting a low-quality study (68) and recalculating the pooled estimates for the remaining studies did not significantly alter the effect on all-cause readmission (RR, 0.82 vs. 0.80; p = 0.000).

Discussion

Our meta-analysis showed that CNI and telemonitoring interventions currently have the best evidence for reducing all-cause readmissions up to 6 months for patients with COPD. The effect of CNI was better than that of telemonitoring because CNI also reduced COPD-specific readmissions during 1-year follow-up. Sensitivity analysis showed that excluding small studies (n < 100) altered the overall estimates for readmission insignificantly. Despite multiple components of continued care and study heterogeneity, the beneficial effect of continued care in reducing readmission rates remained stable across the range of patients, post-discharge interventions, follow-up periods and countries.

Such positive effects could have stemmed from the patient characteristics and additive and/or interaction effects of multiple components of continued care. Most of the patients in CNI group were outpatients with a confirmed diagnosis of COPD without serious complications, and they received a plethora of measures, which mainly included self-management of the disease, individual health education, action plan for self-treatment of exacerbations, scheduled telephone calls and a 12-month follow-up. This was similar with two systematic reviews, which showed that simple self-management education without support was not effective (70), but supported self-management interventions could be effective in reducing respiratory admissions among stable COPD patients (71). This explanation was also supported by our result that telemonitoring was effective for only stable COPD patients. There was no statistical significance among various continued cares in reducing mortality of various continue nursing interventions, which was consistent with the reports of Bergner et al. and Cockcroft et al. (49,72). One RCT of CNI (58) was even stopped because the mortality of intervention group became worse than that of the control group, and the author could not explain this phenomenon. We attributed such failures to the severity and complications of the patients' COPD and the simplicity of continued cares. For instance, the study (58) was based on 416 COPD patients recruited from 20 hospital-based outpatient clinics of Veterans Affairs. The intervention included COPD education during four 90-minute weekly individual and one group sessions, an action plan for identification and treatment of exacerbations and scheduled proactive telephone calls for case management. The number of reinforcement telephone calls was once per month for 3 months and every 3 months thereafter, as was fewer than other studies (28,59). In contrast, our meta-analysis selected clearly defined COPD patients with an average age of 69 or older and excluded patients with complex diseases. Thus, our result showed no statistical differences among various continued cares in reducing mortality rates.

Several limitations of this review should be mentioned here. First, most RCTs compared an intervention with “usual care” the details of which were not reported. Second, we only selected studies that reported the proportion of readmitted patients but ignored information about secondary outcomes such as costs. Third, we were unable to obtain individual patient-level data. Information on specific components of the intervention was limited. Some RCTs did not clearly describe the methods used for assessing readmissions. Methods for handling missing data varied across studies.

Future interventions may unify standard specification requirements and compare them, respectively, to see which measure is the most effective. Future studies may also evaluate whether interventions that reduce readmission rates over 6–12 months also reduce readmission rates within 3 months, and could directly compare one intervention with another (for example, home visits vs. health education). The included studies in this review did not evaluate the cost-effectiveness of interventions. It is critical to study how the costs may be distributed to the care providers and patients to sustain any continued care. Finally, it should be considered to determine whether increased active management of psychological morbidity may improve outcomes and reduces the financial burden to achieve the optimal cost-effectiveness.

In conclusion, the best intervention was CNI, which reduced all-cause readmission and COPD-specific readmission in 6–12 months after hospital discharge; health education and telemonitoring reduced all-cause readmissions; home visits reduced COPD-specific readmission but not all-cause readmission. These interventions may be adopted differentially by systems or providers seeking to implement transitional care interventions for patients with COPD.

Declaration of interest

The authors are responsible for the content of the paper and have no conflicts of interest to disclose.

Funding

The study was funded by the Hubei Province Health and Family Planning Scientific Research Project (Grant No. WJ2015Z058) and the Hubei University of Chinese Medicine Scientific Research Project.