Closed schuemie closed 1 year ago
Another challenge is that currently there's no notion of a 'main' effect we're estimating (unlike CohortMethod, where there is a single exposure that we're estimating the effect for). The main effect (effect of interest) is the one where for example want to apply empirical calibration). One solution would be to require the user to identify the effect(s) of interest when calling createEraCovariateSettings ()
.
Based on the files generated by the
runSccsAnalyses()
function, this function should generate CSV files that are human-reviewable, and contain no patient-level data. This should be enforced using aminCellCount
argument like in other HADES packages.As a first step, the data model for these CSV files should be designed. Especially challenging will be how to communicate the various plots that can show patient-level data, such as the plot generated by the
plotAgeSpans ()
function. Perhaps some binning could be applied?The resulting set of CSV files are intended for sharing between OHDSI data sites, and should be sufficient to fully understand the study diagnostics and results.
See the SkeletonComparativeEffectStudy results data model for ideas. (note that this will soon be incorporated into CohortMethod itself)