Design and analysis of studies in health sciences
One of the aims of CAPHRI is to improve the quality of research in health sciences. Quality improvement plays a role in the design stage of a study, in the measurement stage and in the final data analysis stage.
The focus of this research programme is on quality improvement in all three stages. In the programme, an international group of methodologists and statisticians join their expertise in a number of research projects.
The first focus is on improving the design of health studies. The aim is to find optimal and powerful designs, which are also efficient in terms of the costs of performing the study. An example of this research is the cluster randomised trial to investigate the effect of an intervention programme, EXBELT, on the reduction of use of belts in nursing homes, carried out by researchers from the CAHPRI programme, “Improvement and renewing care of the elderly”. Research on the optimality of cluster randomised and multicentre trials has increased knowledge about the optimal characteristics of such designs in terms of efficiency, power and costs.
Another example is centred around the design of cohort studies. The “Smile project” and the “Maastricht cohort study” are examples of such studies. They consist of large samples of patients in different cohorts which are measured repeatedly over a long period of time. This makes them very expensive. The focus of our research is on obtaining cohort designs which are highly efficient and have maximum power with minimum costs.
Health sciences studies often have a hierarchical or multilevel data structure, i.e. data are sampled from organizations (schools, hospitals, general practices) and subjects within these organisations (pupils, patients). Moreover, the data commonly have a longitudinal structure with repeated measurements per subject.
Another example is meta analysis, where units (patients, subjects) are nested within empirical studies. Meta analysis combines results from various empirical studies on the same research question and come with an overall conclusion. The analysis of the data from these studies requires complicated statistical modelling, which not only takes the hierarchical structure of the data into account, but also the correlations of the time-structured measurements.
The last aim of this programme is to provide CAPHRI researchers with adequate guidelines for the design and analysis of their studies, with emphasis on the practical
applicability of the results. Publications in medical and public health journals and implementation of software are important instruments for this last aim.