Rationale for sample sizes (including sample size calculations, budgetary constraints, practical considerations, data saturation, as appropriate)
| Example Sample size |
| Assuming an alpha of 0.05 and a beta of 0.90, an improvement in perceived daily functioning (defined as a score less than or equal to 4 on the Daily Functioning Thermometer, our primary outcome) at T12 occurring in 20% of the patients in the intervention group versus 5% of those in the control group requires at least a net number of 116 patients per arm (N=232; 5 patients per practice nurse). It will be necessary to take account of a possible dependence between observations on patients of the same practice nurse (PN). The intra-class correlation coefficient (ICC) is assumed to be 0.04, a median value for cluster-RCTs in the primary care setting.33 Assuming a 30% loss to follow-up we need to recruit at least 331 patients (8 per PN). Since participation in the screening procedure will not necessarily mean that patients also give informed consent for the effect evaluation, 10 consecutive patients for each PN will be invited to participate in the effect evaluation (N=460).24 |
| Explanation It is important to recruit sufficient participants to be able to address the study's implementation objectives; the rationale for the number of sites and/or people recruited to the study needs to be justified. In a trial (eg, a cluster RCT), this will be based on a sample size calculation using the primary implementation outcome. If health outcomes are also being assessed, consideration may need to be given to the sample size for the primary health outcome. Design-specific advice on reporting sample size calculations can be found in relevant reporting standards.27–29 31 75 In studies using qualitative methodology, data saturation may inform the final sample size. Budgetary constraints and other pragmatic considerations may also be relevant (such as evaluating an initiative in which size is already determined; in the second example, the sample was ‘all active asthmatics’ in the practice18). |