We undertook a retrospective cohort study to analyze the long-term mortality of individuals who enrolled in outpatient PR between January 2000 and December 2002 at a tertiary hospital in Melbourne, Australia. We established a control group, matched by type of chronic lung disease, from individuals who underwent a lung function test at the Respiratory Laboratory at the same tertiary hospital during the same timeframe.
Data were obtained from a prospectively maintained PR database that included all patients referred to the outpatient PR program, completed an initial assessment, and attended at least one PR session. Patients were usually referred by a respiratory specialist diagnosed with a chronic respiratory disease. They were enrolled in the program when clinically stable (at least four weeks after an exacerbation or hospital admission) and remained symptomatic (Medical Research Council score ≥ 1) despite optimal medical therapy. The PR program consisted of baseline assessment followed by twice weekly attendance at the hospital for six weeks with content provided according to consensus guidelines2. Each session included components of supervised exercise and education, lasting for two hours. The exercise component consisted of upper and lower limb endurance training, thoracic mobilization exercises, warm up and cool down stretches and was supplemented by a home exercise program that included daily walking. The education component of the PR program was provided by a multidisciplinary team and involved interactive sessions covering topics of lung disease, medications, oxygen therapy, symptom management, airway clearance, nutritional advice, energy conservation, relaxation and anxiety management, coping with chronic illness and social supports.
Patients were identified as “cases” and included for analyses if they 1) had a respiratory disease diagnosis and 2) had a valid spirometry result. For those who had multiple enrolments in PR, only the last one in the specified period was considered and, where multiple spirometry results were available, only the result closest to the initial PR appointment was used. We selected cases from 2000 to 2002 to ensure sufficient outcomes of interest (mortality) in the comparisons.
Patients attending the Austin Health Respiratory Laboratory for spirometry between January 2000 and December 2002 were included in the control pool (usual care) if they 1) had a respiratory disease diagnosis, and 2) did not participate in PR at the Austin between 1998–2007. If multiple spirometry results were available for controls, only the last one from the specified period was used.
Cases were categorized for primary lung disease, and a frequency match performed to find controls matched for cases in each category. Patients in the control pool were also categorized for primary lung disease and then arranged according to their hospital unique record number (UR). To identify matches for our cases, an equivalent number of patients were selected for each category from the control pool, starting from the uppermost section of the UR-sorted list.
Age was computed in years based on the baseline date (January 1, 2000) for each individual. The severity of lung disease was represented by the recorded FEV 1 percent predicted, which was extracted from the Austin Health Respiratory Laboratory database. In cases where spirometry results were absent from the respiratory laboratory database, data from the PR database were used.
Other variables such as sex, BMI, and pack-year smoking and current smoking status were extracted from the same lung function report. The Index of Relative Socio-economic Advantage and Disadvantage (IRSAD) for each individual was derived from their postal code, utilizing data from Socio-Economic Indexes for Areas 2001 (SEIFA 2001), accessible on the Australian Bureau of Statistics website. This index was used to determine social and economic wellbeing for each individual. Low values indicate areas of disadvantage and high values indicate areas of advantage.
To determine the date of death (DOD) for both groups, cross-referencing was performed using the Austin Health medical record and an online open-source record (mytribute.com.au) in June 2022. When there was no evidence of death, the date of the last hospital encounter was recorded as the censored date.
Baseline differences were examined between the case group and the matched control group. A two-sample t-test was applied to assess disparities in FEV 1 percent predicted, age, BMI, IRSAD, and pack-year smoking. Additionally, a chi-squared test was utilized to evaluate discrepancies in sex distribution and current smoking status at baseline.
A Kaplan-Meier plot was generated to contrast the survival probabilities of cases and controls as a function of time. Subsequently, a log-rank test was utilized to assess the significance of the differences observed between the survival curves.
Univariable and multivariable Cox regression analyses were conducted to explore the individual associations of PR attendance, age, sex, FEV 1 percent predicted, BMI, pack-year smoking, current smoking status and IRSAD with mortality, both independently and after accounting for other variables. The concordance index was computed to evaluate the precision of the Cox model. This analysis was repeated for a subset of only COPD patients. A significant level of p-value less than 0.05 was adopted, and all analyses were carried out using the Survival package in R version 4.3.1.