Article type
Year
Abstract
Background: Diagnostic and prognostic literature is overwhelmed with studies reporting univariable associations. Currently, methods to incorporate such information in multivariable analysis of individual patient data (IPD) are underdeveloped. In addition, the few described methods are unfamiliar to many researchers, causing them to ignore existing evidence when performing prediction research with new data.
Methods: This article introduces a method to incorporate univariable associations in multivariable analysis of IPD that can be viewed as an extension of the adaptation method originally proposed by Greenland (1987) and Steyerberg (2000). Different variants of the method were tested in a simulation study, where performance was measured by comparing estimated associations with their true values according to the Mean Squared Error and coverage of the 95% confidence intervals.
Results: Results demonstrate that performance of estimated associations considerably improves for small datasets where external evidence is included. Although the error of estimated associations decreases with increasing amount of individual patient data, it does not disappear completely, even in very large datasets.
Conclusions: The authors conclude that the proposed method to incorporate previously published univariable predictor-outcome associations as evidence in new multivariable prediction analyses is superior to established approaches and is especially worthwhile when relatively limited individual patient data are available.
Methods: This article introduces a method to incorporate univariable associations in multivariable analysis of IPD that can be viewed as an extension of the adaptation method originally proposed by Greenland (1987) and Steyerberg (2000). Different variants of the method were tested in a simulation study, where performance was measured by comparing estimated associations with their true values according to the Mean Squared Error and coverage of the 95% confidence intervals.
Results: Results demonstrate that performance of estimated associations considerably improves for small datasets where external evidence is included. Although the error of estimated associations decreases with increasing amount of individual patient data, it does not disappear completely, even in very large datasets.
Conclusions: The authors conclude that the proposed method to incorporate previously published univariable predictor-outcome associations as evidence in new multivariable prediction analyses is superior to established approaches and is especially worthwhile when relatively limited individual patient data are available.