My areas of interest are predictive modeling and model validation, missing data, diagnostic and prognostic research, health services research, cohort studies, clinical trials, cardiovascular disease, reproducible research, and Bayesian methods.
As reflected in his almost 300 peer-reviewed publications (5 with more than 1000 citations), Dr Harrell has devoted his career to the study of patient outcomes in general and specifically to the development of accurate prognostic and diagnostic models and models for many other patient responses. Much of Dr Harrell's work has been applied to health services and outcomes research, technology evaluation, observational databases, and clinical trials. His primary methodologic research relates to development of reliable statistical models, quantifying predictive accuracy, modeling strategies utilizing data reduction methods, estimating covariable transformations, model validation methods, penalized estimation (shrinkage), missing data imputation, clinical trials, flexible Bayesian clinical trial design, pharmaceutical safety, statistical graphics, and statistical reporting. He has researched methods to estimate how continuous predictors relate to outcomes without assuming linearity, showing the advantages of piecewise cubic polynomials or spline functions. All of this work has taken into account that a risk model's likely performance on a new subject sample should be the touchstone. He has extended Efron's bootstrap estimator of the "optimism" in a model's predictive accuracy to validate more complex survival and risk models. His book Regression Modeling Strategies with Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis (2nd Edition 2015, Springer-Verlag) contains theory, examples, and detailed case studies demonstrating the use of many modern statistical modeling tools.