Background:
Diagnostic and prognostic literature is overwhelmed with studies reporting univariablepredictor-outcome associations. Currently, methods to incorporate such information in theconstruction of a prediction model are underdeveloped and unfamiliar to manyresearchers.
Methods:
This article aims to improve upon an adaptation method originally proposed by Greenland(1987) and Steyerberg (2000) to incorporate previously published univariable associationsin the construction of a novel prediction model. The proposed method improves upon thevariance estimation component by reconfiguring the adaptation process in establishedtheory and making it more robust. Different variants of the proposed method were testedin a simulation study, where performance was measured by comparing estimatedassociations with their predefined values according to the Mean Squared Error andcoverage of the 90% confidence intervals.
Results:
Results demonstrate that performance of estimated multivariable associationsconsiderably improves for small datasets where external evidence is included. Althoughthe error of estimated associations decreases with increasing amount of individualparticipant data, it does not disappear completely, even in very large datasets.
Conclusions:
The proposed method to aggregate previously published univariable associations withindividual participant data in the construction of a novel prediction models outperformsestablished approaches and is especially worthwhile when relatively limited individualparticipant data are available.