Abstract
Although participation in pulmonary rehabilitation (PR) improves the health outcomes in patients with Chronic Obstructive Pulmonary Disease (COPD), there are insufficient resources to provide PR to all patients with COPD. Thus, predicting which patients are at risk for drop-out and non-response to rehabilitation is necessary in order to optimize limited resources. This study examined which patient characteristics are predictive of PR drop-out and non-response. 814 patients with COPD took part in standard out-patient PR for 8 weeks. Demographic and standard clinical data were collected before the rehabilitation program had started. Data was analyzed retrospectively to determine if baseline patient characteristics could predict drop-out and non-response to rehabilitation. Drop-out was defined as participation in less than 50% of the rehabilitation sessions. Non-response was defined as improvement less than 4% on the St. George's Respiratory Questionnaire (SGRQ). A discriminant function analysis identified age, smoking history, and health status as predictors of patient drop-out, p < .0001, with younger, current smokers and patients with lower health status being at risk for drop-out. No variables measured significantly predicted who those at risk would be for non-response to rehabilitation, p > .05. Pulmonary function data did not predict drop-out or non-response to PR. These findings indicate that perceived impairment (i.e., health status) is more likely to influence completion of rehabilitation than actual pulmonary impairment and that demographic and standard clinical data do not adequately predict patient drop-out and non-response to rehabilitation.
Introduction
Pulmonary Rehabilitation (PR) has been established as the standard of care in the management of chronic obstructive pulmonary disease (COPD). Participation in PR results in improved health status, functional exercise capacity (Citation1, 2) and has been demonstrated to reduce acute exacerbations, hospitalizations, and health care costs (Citation2). Furthermore, research has shown that PR is a more effective therapeutic strategy than traditional pharmacological treatments at improving health outcomes in patients with COPD (Citation1).
Although there are important benefits to PR, recent work has shown that there are insufficient resources to provide PR to all patients with COPD (Citation3). Therefore, it is of interest to optimize the limited healthcare resources available to make the greatest possible improvements in health outcomes in individuals with COPD. Drop-out from PR is typically ∼ 30% (Citation4), and represents a failure to address patients’ needs, as well as a loss of healthcare resources, as patients who dropped out of PR would not experience the same benefits as those who complete PR. Unfortunately, there is limited information available that indicates which patient characteristics are predictive of rehabilitation drop-out. In addition, there is limited research pertaining to the patient characteristics that predict which patients will respond to treatment and which patients will not. This is particularly important as approximately 30% of patients who participated in rehabilitation do not improve their health status or exercise tolerance (Citation5). Thus, in order to improve health outcomes, it is paramount to identify those at risk for drop-out and not responding to treatment so that resources are allocated appropriately and additional support is provided to those who need it.
The prevalence of females diagnosed with COPD has increased as of late, however, previous literature has suggested that females are dramatically underrepresented in research populations (Citation6). Furthermore, there is interest as to whether rehabilitation programs are as effective in female compared to male patients with COPD (Citation7). It is therefore important to test if sex influences response to treatment in a sample that is representative of the female population with COPD. The purpose of this study was to (i) determine if clinical data can predict completion/drop-out during PR, (ii) identify factors that could predict responders/non-responders, and (iii) determine if sex can discriminate between responders and non-responders to treatment, in a sample that has approximately equal number of female and male participants with COPD.
Methods
Participants and Recruitment
Patients were referred by their physicians to attend out-patient PR at a large urban hospital, and enrolment took place between 2005 and 2008. All patients were screened by a Pulmonologist prior to enrolment in PR, and COPD was confirmed by a full pulmonary function test. Patients were excluded from the study if they had unstable cardiac conditions, dementia, interstitial lung disease, Talc Granulomatosis or had previously participated in the PR program. Current smokers were included in the study, as were patients with hypoxemia requiring oxygen therapy, and those with various co-morbidities, including stable chronic heart failure and stable coronary artery disease. In total, data were collected from 814 patients who met these stipulations. See for patient characteristics.
Table 1. Patient characteristics
Rehabilitation Program
The PR program is located within metro Edmonton, Alberta, Canada (population ∼ 1 million). Patients were enrolled in PR in groups of 14. Each participant was given a choice of attending class 3 days per week for 6 weeks, or 2 days per week for 8 weeks. As well, they chose either morning, afternoon or evening sessions. Each session included group exercise for 2 hours, and group education for 1 hour. Prior to enrollment, patient history was obtained by consultation with a pulmonary physician and subsequently performed a cardiopulmonary exercise stress test. At the beginning of PR each patient performed a minimum of two 12-min walk tests on 2 separate days, with the mean distance reported. Patients also completed the St. George's Respiratory Questionnaire (SGRQ), and the Short Form Health Survey (SF-36) to assess health status prior to PR. Exercise and health status data were also obtained immediately following PR. All data were stored in a central clinical database (Windows Access).
The exercise program was supervised by respiratory therapists and followed PR guidelines for exercise training (Citation2). Exercise programs included aerobic exercise consisting of hallway and treadmill walking, cycling and arm ergometer training. The exercise intensity was personalized based on patient symptoms and baseline exercise capacity. Patients performed light resistance exercise training that included hand weights and/or elastic bands/tubes. In addition, patients performed flexibility exercises and breathing retraining.
Educational modules directed towards self-management (Citation2) were offered for 1 hour each day by a multi-disciplinary team including respiratory therapists, physical therapists, pharmacists, dieticians, kinesiologists, and health psychologists. Topics included: basic pathophysiology of lung disease, pulmonary hygiene, exercise training, respiratory medications, inhaler devices and techniques, nutrition, relaxation/stress management, travel/home care and oxygen therapy.
Patient paid parking was available to those travelling to the PR program by car. No transportation to the program itself was provided, however transport services were available to patients at their own discretion. Transportation/parking expenses were the only direct fiscal cost to the patients attending this PR program. In addition, patients who smoked were provided smoking cessation support, which included information pamphlets and options to help them quit, including nicotine replacement, as well referral to specialized smoking cessation programs. Although patients were encouraged to quit smoking, the rehabilitation program did not exclude patients that smoked.
Procedures
Ethical approval was given by the University and Hospital Research Ethic boards. All demographic and clinical data (described later) were obtained prior to each patient starting the rehabilitation program. All data, including program adherence data, were stored in a central database, and were extracted from the program retrospectively.
Measures
Demographics
Age, sex, self-reported smoking history (pack-years), and months since quitting smoking data were collected by program staff. In addition, data on co-morbid diseases was gathered through pulmonologist consultation.
Exercise tolerance and peak exercise workload
The mean distance walked on a minimum of 2 self-paced 12-min walks was taken to assess exercise tolerance. At the time the data was collected, the 12-min walk distance was the standard measure of exercise tolerance at the rehabilitation centre. During the 12-min walk patients are taught to pace their walking speeds to control their breathlessness. In patients with respiratory disease, the 12-minute walk distance is highly correlated (r = 0.955) with the 6-min walk distance (Citation8). In addition, all patients performed a symptom-limited graded treadmill exercise test (Citation9). As many patients were on supplemental oxygen (i.e., 25%), which prevented measurement of peak oxygen consumption (VO2peak), VO2peak was calculated in ml/kg/min based on peak treadmill speed and grade obtained (Citation10).
The Short Form Health Survey (SF-36)(Citation11)
The SF-36 is a measure of general health status (Citation11). This survey comprises 36 items forming 8 subscales: physical functioning, role-physical, social functioning, bodily pain, general mental health, role-emotional, vitality, and general health perceptions, that in turn are summed into two general indicators: Physical Health and Mental Health. Reliability estimates for the physical and mental summary scores often exceed .90 (Citation12).
The St. George's Respiratory Questionnaire (SGRQ)(Citation13)
The SGRQ is a disease specific measure of health status for patients with COPD. (Citation13) This instrument consists of 50 items that are organized into three content areas: symptoms, activities, and impacts. This survey has been found to be a valid measure of health status in patients with COPD.(Citation14)
Lung function
Full lung function data (spirometry, lung volumes by plethysmography and diffusion capacity for carbon monoxide) were obtained in all subjects, and reported in% of predicted reference values.
Analyses
The data were analyzed retrospectively using SPSS version 18.0. All 814 patients who met the study criteria were included in the first set of analyses that compared and differentiated program completers and non-completers. Similar to other studies, completion was defined as attendance to 50% or more of PR classes (Citation15). Factorial ANOVAs compared those who completed the program to those who dropped-out, to indicate on which baseline variables the program completers and drop-outs differed (dependent variables: age, sex, smoking history-pack years, smoking history-months since quit, FEV1% predicted, FEV1/FVC% predicted, 12-min walk distance, and VO2peak, SF-36 subscales: physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role emotional, mental health and SGRQ total score).
A discriminant function analysis was conducted to establish which baseline variables were relevant to program completion and program dropout at baseline (predictor variables: age, sex, smoking history-months since quitting, 12-min walk distance, FEV1% predicted, FEV1/FVC% predicted, SF-36 subscales: physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional, mental health, and VO2peak). Between 3.6% and 6.7% of scores were missing on some of the predictor variables. Due to this small number and the large data set, missing data were replaced by the series mean prior to analysis (Citation16). The final sample included 678 program completers and 136 non-completers (dropouts), indicating a 17% drop-out rate. Reasons for drop-out included respiratory illness (27%), non-respiratory illness/injury (28%), lost contact (20%), personal/family issue (4%), work commitments (4%), and transportation (2%). The drop-out rate was similar to other recent prospective data from the same centre (Citation17).
Repeated measures tests were conducted on the completer data to determine the influence of the PR program on patient outcomes. Repeated measures ANOVAs compared the 12-min walk distance and SGRQ total score over time. Last, a time (pre, post) x SF36 total component scores (Physical, Mental) repeated measures MANOVA was conducted to determine the change over the course of rehabilitation. Follow-up univariate ANOVAs were conducted for each SF-36 subscale over time, pending a statistically significant multivariate effect.
An additional analysis was conducted with the data from the participants who completed the program, in order to determine if baseline data could predict those who responded to treatment and those who did not respond to treatment. Program responders were defined as those who had a clinically significant improvement on the SGRQ subscales with rehabilitation (i.e. 4%).(Citation18) A discriminant function analysis was conducted on those who completed the program in order to identify the variables that were related to program responders and non-responders. The predictor variables for the discriminant function analysis were: age, sex, smoking history-pack years, smoking history-months since quit, FEV1% predicted, FEV1/FVC% predicted, 12-minute walk distance, and VO2peak, and SF-36 subscales.
Results
Program completion versus dropout
A multivariate ANOVA and follow-up univariate ANOVAs found statistically significant differences between program completers and drop-outs on the dependent variables: age, smoking history-months since quit, FEV1%, all SF-36 scales, and SGRQ total score, all ps < .05. Program completers tended to be older (mean age = 68.53 years old) than program dropouts (mean age = 63.13 years old), were less likely to be current smokers than program dropouts (26%, and 49%, respectively), had a higher FEV1 score than program dropouts (mean = 50.44, mean = 45.83 respectively), and better self-reported health status than program dropouts. See for the characteristics of program completers and drop-outs and corresponding F-tests. Sex did not influence completion rate.
Table 2. Characteristics of program completers and drop-outs
The discriminant function analysis yielded one significant function that differentiated program completers and dropouts, Eigenvalue = .092, Wilks’ Lambda = .916, Chi-square (Citation16) = 70.48, p < .0001. The strongest predictors and the pooled within-groups correlations between the discriminating variables and standardized canonical discriminant function are presented in . Based on these coefficients, age, bodily pain, social functioning, vitality, mental health, and smoking status, respectively, demonstrated the strongest relationship with the discriminant function, meaning they are the main variables associated with determination of participants as completers or dropouts. When predicting program completers and program dropouts, the discriminant function correctly classified 83.2% of the individuals in our sample.
Table 3. Predictor variables of discriminant function analysis for completers and drop-outs
Response to pulmonary rehabilitation
A total of 676 patients completed PR. In these patients, significant improvements in 12-minute walk distance, SGRQ total score, and SF-36 scores were observed with PR (see ).
Table 4. St. George's Respiratory Questionnaire (SGRQ), 12-minute walk distance, and Short Form Health Survey (SF-36) Pre- and Post-Pulmonary Rehabilitation
Program responders versus non-responders
The discriminant function analysis to determine if the clinical data we obtained could predict those who would respond to rehabilitation and those who would not respond to rehabilitation was non-significant, p > .05. These results indicate that although there were program responders, none of the dependent variables included in the discriminant function analysis significantly predicted their responder status.
Discussion
Predicting which patient characteristics are associated with drop-out and non-response to PR is critical in order to optimize economic and personnel resources, and to ultimately address the needs of patients with COPD in order to provide the greatest improvement in health outcomes in all patients. In this study, the variables associated with drop-out in COPD patients were age, smoking history, and health status. While smoking history and health status have previously been identified as predictors of drop-out in PR (Citation4), age as a predictor is a novel finding. Sex did not influence completion rate. In addition, our results indicated that standard baseline clinical data collected, including lung function and exercise tolerance, were unable to predict program responders.
Similar to previous research, our results indicate that having a history of smoking increases the risk of patient drop-out from rehabilitation. Young et al. (Citation19) suggest that the lack of adherence to PR may be carried over from the lack of adherence with other personal behavioural management strategies. The contention is that those who are unsuccessful at smoking cessation are also unsuccessful at attending exercise-based PR because of the common factor of poor behavioural management. In fact, smoking has been associated with the lack of adherence to other rehabilitation programs (i.e., cardiac rehabilitation(Citation20)) and Vagaggini et al. (Citation21) found that only ex-smokers in their sample completed PR, thus supporting this contention.
Baseline health status was also an important predictor of drop-out, with those who dropped out having poorer perceptions of health status. This finding has been demonstrated previously (Citation4) and suggests that a patient's initial perceived health status has an influence on subsequent attendance to PR. Like Garrod et al. (Citation4), our study included a respiratory-specific health status measure (i.e., SGRQ); however, unlike other studies assessing drop-out in patients with COPD, we also included a general measure of health status (i.e., SF-36). In the current study, the health status subscales that were most associated with drop-out were those from the SF-36. COPD is a highly co-morbid disease, and as such, the SF-36 may assess a greater variety of potential complications that patients with COPD may experience. Based on these findings, it would be important to include a measure of general health status when assessing outcomes in patients with COPD.
In terms of physiological parameters measured, such as lung function and exercise tolerance, the current study found that neither variable significantly differentiated those who completed rehabilitation from those who dropped-out. This finding is consistent with the results of studies by Garrod et al. (Citation4) and Young et al. (Citation19), and indicates that patients’ actual level of pulmonary/exercise impairment does not significantly influence PR drop-out. However, as indicated by the predictive value of health status measures, patients’ perceptions of their impairment, is much more likely to influence whether patients complete rehabilitation.
The current study indicated that the strongest predictor of drop-out was age, with younger patients at greater risk for drop-out. This is a novel finding that has not been previously identified in the literature. The average age of the patients in this study was similar to other samples, so it may be an additional variable, such as employment that is mediating the relationship between age and dropout. Patients who dropped out of our study were on average 63 years of age, which is lower than the typical retirement age of 65. It could be that those who dropped out were still employed, which could ultimately influence participation in PR. Although not statistically significant, Young et al. (Citation19) found that there was a lower retirement rate among those who did not adhere to rehabilitation (72%) compared to those who did adhere to rehabilitation (80%), suggesting a possible influence of employment on drop-out. As many studies have failed to include employment status as a predictor variable in the analysis, it may be useful for future research to include this variable and specifically examine the relationship between age, employment and drop-out in PR.
A limitation of this study was that we failed to measure patients’ levels of depression. Depression has commonly been identified as a predictor of non-adherence to rehabilitation, with those scoring higher on depression being more likely to drop-out (Citation4, Citation22). In one study, depressed patients were almost twice as likely to drop-out compared to patients who were not depressed.(Citation4) Although clinical depression was not overtly addressed in the current study, the mental health component of the SF-36 was an important predictor of drop-out. Taken together, the evidence suggests that depression is likely a consistent predictor of PR drop-out, and that efforts to decrease the incidence of depression before rehabilitation commences are warranted.
Determining the definition of drop-out from rehabilitation is a somewhat arbitrary task. In the current study, patient drop-out was defined as attending less than 50% of the PR classes, in accordance with the PR program. This 50% cut-off was set within then clinic as it is generally observed that attendance of less than 50% of the classes does not lead to noticeable patient benefits. This definition of drop-out may have impacted the results of the discriminant function, in that some of the discriminating variables may have differed if another attendance cut off was used.
Several other studies have investigated the predictive value of baseline variables on response to PR (Citation4, Citation21, Citation23). In the current study, improvement in SGRQ total score was the outcome variable in the discriminant analysis and it was found that no baseline variables significantly predicted improvement in the SGRQ total score, including sex. This finding is similar to another study which found no sex differences in improvement in health status between men and women patients with COPD (Citation24). A strength of the current study was our large sample size, and equal representation of both sexes. Combined, these results indicate that both female and male patients with COPD are equally likely to benefit from PR.
Although other studies have also found no significant predictors of response in PR (Citation4, Citation21), Troosters et al. (Citation23) found that baseline health status predicted response to PR, such that those who had lower health status were more likely to show greater improvement. However, the authors noted that regression to the mean was the likely explanation of this finding. To date it seems there have been no variables identified that can consistently predict improvement in health status with rehabilitation.
Unlike other studies (Citation4, Citation21, Citation23), we did not examine the predictive ability of baseline variables on improvements in exercise tolerance. Troosters et al. (Citation23) found that lack of baseline strength was a significant predictor of improvement in exercise tolerance, and Vaggagini et al. (Citation21) found that BMI > 25 and baseline PaO2 < 60 mmHG significantly predicted an increase in exercise tolerance. It may be that there is more room for improvement among those who have greater initial deconditioning, and so they are able to benefit more from exercise training in PR. Importantly, Garrod et al. (Citation4) failed to identify any significant predictors of improvement in exercise tolerance, and no recent study has found pulmonary function to be a significant predictor of response (exercise or health status) to PR. While an increase in health status and physical function with exercise-based rehabilitation is well documented in patients with COPD (Citation15, Citation25), it still remains unclear which variables can predict individual PR success.
It is important to note that this study did not assess a number of variables that could potentially predict patient completion and response to PR. Demographic variables such as socioeconomic status, and education may influence patients’ willingness to attend and adhere to PR, and may also have impacted responses to PR. Other variables that may be of importance to the prediction of completion and response to PR, are choice of attending morning or afternoon classes, past history of exacerbations, as well as distance from home to the PR facility. In fact, a recent review by Keating et al. (Citation26) identified transportation to the PR facility to be a significant barrier to program uptake and completion in several studies (Citation22, Citation27–29).
However, transportation barriers have even been identified by participants when volunteer driving services have been offered (Citation30), suggesting that patient perceptions of barriers may need to be considered in addition to actual barriers. Past behaviour, and social-cognitive mechanisms may account for patient perceptions and therefore be useful variables to assess when predicting patient uptake and response to PR. In fact, Garrod et al. (Citation4) found that lower self-efficacy was related to greater improvement in 6-min walk distance, suggesting that social-cognitive variables play a role in the increase of patient physical function. Future research should further investigate the relationship between self-efficacy and patient response to treatment.
In the current study, a 4-unit change of the SGRQ was used to classify responders to PR. There is ongoing debate as to the magnitude of change in SGRQ required to show clinical significance (Citation31). Further, this definition is useful for classifying individuals into groups; it may not be as useful for determining individual success to rehabilitation. Assessing changes in exercise capacity, breathlessness, and perhaps other measures may capture meaningful responses to PR that may not be fully captured by the SGRQ. Using a variety of outcome measures to assess response to PR may provide the greatest indication of patient success to PR.
Conclusions
Due to the inability to provide PR to all patients with COPD, it is important to optimize the limited resources available. The results of this study, which reviewed a large sample of patients with COPD enrolled in PR, indicate that age, smoking status and health status are key predictors of program drop-out. In contrast, no clinical data predicted patients who failed to show a clinically significant improvement with rehabilitation. Of particular note, sex was not found to be a predictor of response to treatment, suggesting that both female and male participants are equally as likely to benefit from rehabilitation. It may be beneficial to provide smoking cessation and self-management programs to patients with COPD prior to enrollment in PR. Likewise, additional resources for younger patients with COPD aimed at increasing accessibility of programs and decreasing the burden of time away from competing priorities may result in better attendance to PR and greater improvements in health outcomes. More work is needed to determine the role of social-cognitive and motivational variables in patient drop-out and success in PR, which may be instrumental in patient attendance and response to treatment
Declaration of Interest
The authors report no conflicts of interest; funding sources have been specified earlier in this article. The authors alone are responsible for the content and writing of the paper. Funding for this study was provided by the Canadian Institutes of Health Research. Dr. Stickland was supported by a Canadian Institutes of Health Research New Investigator Award and Heart and Stroke Foundation of Canada New Investigator Award. Dr. Stickland has received speaking honoraria from GlaxoSmithKline. The remaining others have no confl icts of interest to disclose.
References
- Lacasse Y, Goldstein R, Lasserson TJ, Martin S. Pulmonary rehabilitation for chronic obstructive pulmonary disease. Cochrane Database Syst Rev. 2006(4): CD003793.
- Ries AL, Bauldoff GS, Carlin BW, Casaburi R, Emery CF, Mahler DA, Pulmonary rehabilitation: Joint ACCP/AACVPR Evidence-based clinical practice guidelines. Chest 2007; 131(5 Suppl):4S–42S.
- Brooks D, Lacasse Y, Goldstein RS. Pulmonary rehabilitation programs in Canada: national survey. Can Respir J 1999; 6(1):55–63.
- Garrod R, Marshall J, Barley E, Jones PW. Predictors of success and failure in pulmonary rehabilitation. Eur Respir J. 2006; 27(4):788–794.
- Garrod R, Ford K, Daly C, Hoareau C, Howard M, Simmonds C. Pulmonary rehabilitation: analysis of a clinical service. Physiother Res Int. 2004; 9(3):111–120.
- Lacasse Y, Wong E, Guyatt GH, King D, Cook DJ, Goldstein RS. Meta-analysis of respiratory rehabilitation in chronic obstructive pulmonary disease. Lancet 1996;348(9035): 1115–1159.
- Marciniuk DD, Brooks D, Butcher S, Debigare R, Dechman G, Ford G, Optimizing pulmonary rehabilitation in chronic obstructive pulmonary disease–practical issues: a Canadian Thoracic Society Clinical Practice Guideline. Can Respir J 2010; 17(4):159–168.
- Butland RJ, Pang J, Gross ER, Woodcock AA, Geddes DM. Two-, six-, and 12-minute walking tests in respiratory disease. Br Med J (Clin Res Ed) 1982; 284(6329):1607–1608.
- ATS/ACCP. Statement on cardiopulmonary exercise testing. Am J Respir Crit Care Med 2003;167:211–277.
- ACSM. ACSM's Guidelines for Exercise Testing and Prescription. 6th ed. New York: Lippincott Williams & Wilkins; 2000.
- Ware JE Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care 1992; 30(6):473–483.
- Ware JE, Kosinski M, Keller SD. SF-36 Physical and Mental Health Summary Scales: A User's Manual. 1994. The Health Institute, Boston, MA.
- Jones PW, Quirk FH, Baveystock CM. The St George's Respiratory Questionnaire. Respir Med 1991; 85 Suppl B:25–31; discussion 3–7.
- Jones PW, Quirk FH, Baveystock CM, Littlejohns P. A self-complete measure of health status for chronic airflow limitation. The St. George's Respiratory Questionnaire. Am Rev Respir Dis 1992; 145(6):1321–1327.
- Wedzicha JA, Bestall JC, Garrod R, Garnham R, Paul EA, Jones PW. Randomized controlled trial of pulmonary rehabilitation in severe chronic obstructive pulmonary disease patients, stratified with the MRC dyspnoea scale. Eur Respir J 1998; 12(2):363–369.
- Tabachnick BG, Fidell, LS. Using Multivariate Statistics. 5th ed. Boston MA: Allyn and Bacon, 2007.
- Stickland M, Jourdain T, Wong EY, Rodgers WM, Jendzjowsky NG, Macdonald GF. Using Telehealth technology to deliver pulmonary rehabilitation in chronic obstructive pulmonary disease patients. Can Respir J: J Can Thorac Soc 2011; 18(4):216–220.
- Jones PW. Health status measurement in chronic obstructive pulmonary disease. Thorax. 2001; 56(11):880–887.
- Young P, Dewse M, Fergusson W, Kolbe J. Respiratory rehabilitation in chronic obstructive pulmonary disease: predictors of nonadherence. Eur Respir J 1999;13(4):855–8559.
- Oldridge NB, Streiner DL. The health belief model: predicting compliance and dropout in cardiac rehabilitation. Med Sci Sports Exerc 1990; 22(5):678–683.
- Vagaggini B, Costa F, Antonelli S, De Simone C, De Cusatis G, Martino F, Clinical predictors of the efficacy of a pulmonary rehabilitation programme in patients with COPD. Respir Med 2009;103(8):1224–1230.
- Fan VS, Giardino ND, Blough DK, Kaplan RM, Ramsey SD. Costs of pulmonary rehabilitation and predictors of adherence in the National Emphysema Treatment Trial. COPD 2008; 5(2):105–116.
- Troosters T, Gosselink R, Decramer M. Exercise training in COPD: how to distinguish responders from nonresponders. J Cardiopulm Rehabil 2001; 21(1):10–17.
- Haave E, Skumlien S, Hyland ME. Gender considerations in pulmonary rehabilitation. J Cardiopul Rehabil Prevent 2008; 28(3):215–219.
- Berry MJ, Rejeski WJ, Adair NE, Zaccaro D. Exercise rehabilitation and chronic obstructive pulmonary disease stage. Am J Respir Crit Care Med 1999; 160(4):1248–1253.
- Keating A, Lee A, Holland AE. What prevents people with chronic obstructive pulmonary disease from attending pulmonary rehabilitation? A systematic review. Chron Respir Dis 2011; 8(2):89–99.
- Fischer MJ, Scharloo M, Abbink JJ, Thijs-Van A, Rudolphus A, Snoei L, Participation and drop-out in pulmonary rehabilitation: a qualitative analysis of the patient's perspective. Clin Rehabil 2007; 21(3):212–221.
- Sabit R, Griffiths TL, Watkins AJ, Evans W, Bolton CE, Shale DJ, Predictors of poor attendance at an outpatient pulmonary rehabilitation programme. Respir Med 2008; 102(6):819–824.
- O'Shea SD, Taylor NF, Paratz JD, But watch out for the weather: factors affecting adherence to progressive resistance exercise for persons with COPD. J Cardiopul Rehabil Prevent 2007; 27(3):166–174; quiz 75–76.
- Keating A, Lee AL, Holland AE. Lack of perceived benefit and inadequate transport influence uptake and completion of pulmonary rehabilitation in people with chronic obstructive pulmonary disease: a qualitative study. J Physiother 2011; 57(3):183–190.
- Jones PW. St. George's Respiratory Questionnaire: MCID. COPD 2005; 2(1):75–79.