Feasibility, effectiveness and safety of self-management in pulmonary rehabilitation 155 Global02. A patient reported outcome (PRO) was chosen because it is increasingly important to capture the patients’ perspective of treatment effectiveness [30]. In addition, for sample size calculation, it is important to choose the most conservative outcome measure, as this requires the most participants [31]; in this case also health related quality of life. The effect is expressed as the absolute difference in mean health related quality of life within each patient over time. In the present design in which supervised PR is scaled down and replaced by self-management, it is especially important to be able to detect even very small declines in quality of life. Therefore, an effect size of d = 0.2 (difference is very small) is chosen. To calculate the sample size required to detect a very small, not necessarily clinically relevant, but undesirable effect, an a priori analysis is performed using G power. With an F test as test family, an ANOVA repeated measures within factors as statistical test, a given a, power and effect size (a = 0.05 and b = 0.95, effect size = 0.2), and number of measurements = 5 (T0-T4), a sample size of 48 patients is needed. Loss to follow-up must be taken into account; patients leaving the study early. In a previous study of patients with COPD who received PR [17], the drop-out rate was 6.1%, and a drop-out rate of 20% is still considered normal. Therefore, an average is taken and a drop-out rate of 13% is taken into account, meaning 54 patients have to be included. To make it also possible to perform a subgroup analysis between adherent and non-adherent patients, as defined at T0, the estimated sample size must be doubled; a total sample size of 108 patients is therefore needed. Sample size qualitative data Purposive sampling will be used to maximize variation, identifying adherent and non-adherent patients, and physiotherapists who want to continue the use of the PATCH tool and those who do not. Potential participants will be invited by the researcher. A purposive sampling approach will be used to ensure that not a sample of extremes will be gathered or that the perspectives of one group of individuals are overrepresented [42], for example, those who are non-adherent. We aim for 10 participants per group (n = 40), which is typically a sufficient number to reach thematic saturation, the point when new themes are no longer emerging from the data [42]. Statistical analysis Analysis quantitative data Data will be analyzed using R version 4.0.3. First, for missing data, the amount of missingness for each variable will be calculated; the difference between the sample size and the number of useable observations. Second, Fisher exact tests will be used to analyze differences in baseline characteristics between patients with missing and complete data. When data are missing at random, multiple imputation
RkJQdWJsaXNoZXIy MjY0ODMw