Purpose: Prediction models that assess a patient's risk of an event are used to inform treatment options and confirm screening tests. The concordance (c) statistic is one measure used to validate the accuracy of these models, but has many extensions when applied to censored data. The purpose of this research was to determine which c-statistic is most accurate for different levels of censoring.
Methods: A simulation study was conducted for n=1000, and censoring rates of 20%, 50%, and 80%. The bias of 5 different c-statistics, including one developed by the authors, was calculated. Mean square error (MSE) and coverage probability (CP) were also calculated, but were of secondary interest.
Results: The c-statistic developed by the authors yielded the smallest bias for larger rates of censoring. Similar results were found for MSE and CP.
Conclusion: The c-statistic developed by the authors appears to be most accurate in the presence of censored data. Thus, it is recommended to use this c-statistic to validate prediction models applied to censored data. This will improve the reliability and comparability across future time-to-event studies.
|