Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 338 - Semiparametric and Non-Parametric Methods in Survival Analysis
Type: Contributed
Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
Sponsor: Biometrics Section
Abstract #313521
Title: A Comparison of Modeling Techniques to Predict Time to Recurrence of Early Stage Oral Squamous Cell Carcinoma
Author(s): Nilesh Shah* and Yingci Liu
Companies: University of Pittsburgh and University of Pittsburgh
Keywords: random forest; survival analysis; proportional hazards; risk prediction; machine learning

A plethora of statistical methods and algorithms have become more widely accessible over recent years, sometimes making it challenging to determine which methods to utilize. Each method has advantages and disadvantages, and different methods may be better suited for different types of data. We investigated the difference in performance among various modeling techniques for survival analysis, including proportional hazards regression, conditional inference forests, and random forest for survival outcomes. We then applied these methods to a database of patients diagnosed with early stage oral squamous cell carcinoma and compared the performance of these methods using integrated Brier scores. The results of the analysis showed that all models similarly identified the driving factors associated with recurrence of oral cancer. Conditional inference forests performed better with regards to integrated Brier score, indicating better performance in predictive ability.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program