Department of Biostatistics Seminar/Workshop Series
Statistical Considerations and Methods in the Development and Internal Validation of the Pediatric Acute Asthma Prediction Rule (APR) for Hospitalizations
Tebeb Gebretsadik, MPH, Senior Assoicate in Biostatistics, Vanderbilt University Medical Center
The seminar aims to serve as an overview of important analytical steps that are consequential to the performance and validation of a clinical prediction model. It is statistical practice oriented and developed with clinical researchers audience in mind. For biostatisticians, in light of growing interest in developing prediction models, the empirical data analysis illustrations could provide area of future statistical research in developing optimal analytical methods and strategies. In this talk,I will discuss:1) practices involved in clinical prediction rule model development that include the evaluation of outcome and the selection of predictor variables;2) data reduction techniques and methods that minimize over-fitting data in order to develop reliable multivariable regression models with robust associations;3) the role of penalized maximum likelihood estimation, adjustment for optimism in measures of predictive ability using the bootstrapping approach to report the prediction model results and its performance;4) the step by step development of an asthma prediction rule and the methods applied in response to reviewers’ concerns as clinical statistical practice example.