Patient-reported outcomes (PROs) play a central role in individualizing treatments in psychology and psychosomatic medicine. Clinicans often use cutpoints or decision-thresholds to decide e.g. whether or not a patient with a depression score of "20" needs additional treatment for depression. Accordingly, many studies aim at determining optimal cutpoints for PROs. Planning, reporting, and harmonizing such studies can be improved by quantifying the variability of optimal cutpoints. At present no commercial or open-source tool supports methods to quantify this variability.
The main aim of the project is the development and dissemination of an R-package and interactive websites to quantify the variability of optimal cutpoints. The R-package will enable biometricians to report the variability of cutpoints and calculate the required sample-size for a specific precision of the cutpoint estimate. The package will make use of multicore functionality and will be made available via the Comprehensive R Archive Network. For applied researchers the Shiny framework will be used to develop interactive and user-friendly websites that provide a graphical user interface to these functions.
A secondary aim of the project is the validation of the summary based approaches to quantify the variability of cutpoints. While this approach would enable post-hoc meta-analysis of published studies, it assumes a normal distribution of test-results. A simulation study will be performed to test whether the summary-based method still gives accurate and unbiased results if this assumption is not met.