Note: Intro -- Preface -- Why Read This Book? -- Intended Audience -- Other Sources -- Structure -- Acknowledgements -- Contents -- About the Author -- 1 Introduction -- 1.1 Diagnosis, Prognosis, and Therapy Choice in Medicine -- 1.1.1 Predictions for Personalized Evidence-Based Medicine -- 1.2 Statistical Modeling for Prediction -- 1.2.1 Model Assumptions -- 1.2.2 Reliability of Predictions: Aleatory and Epistemic Uncertainty -- 1.2.3 Sample Size -- 1.3 Structure of the Book -- 1.3.1 Part I: Prediction Models in Medicine -- 1.3.2 Part II: Developing Internally Valid Prediction Models -- 1.3.3 Part III: Generalizability of Prediction Models -- 1.3.4 Part IV: Applications -- Prediction Models in Medicine -- 2 Applications of Prediction Models -- 2.1 Applications: Medical Practice and Research -- 2.2 Prediction Models for Public Health -- 2.2.1 Targeting of Preventive Interventions -- 2.2.2 *Example: Prediction for Breast Cancer -- 2.3 Prediction Models for Clinical Practice -- 2.3.1 Decision Support on Test Ordering -- 2.3.2 *Example: Predicting Renal Artery Stenosis -- 2.3.3 Starting Treatment: The Treatment Threshold -- 2.3.4 *Example: Probability of Deep Venous Thrombosis -- 2.3.5 Intensity of Treatment -- 2.3.6 *Example: Defining a Poor Prognosis Subgroup in Cancer -- 2.3.7 Cost-Effectiveness of Treatment -- 2.3.8 Delaying Treatment -- 2.3.9 *Example: Spontaneous Pregnancy Chances -- 2.3.10 Surgical Decision-Making -- 2.3.11 *Example: Replacement of Risky Heart Valves -- 2.4 Prediction Models for Medical Research -- 2.4.1 Inclusion and Stratification in a RCT -- 2.4.2 *Example: Selection for TBI Trials -- 2.4.3 Covariate Adjustment in a RCT -- 2.4.4 Gain in Power by Covariate Adjustment -- 2.4.5 *Example: Analysis of the GUSTO-III Trial -- 2.4.6 Prediction Models and Observational Studies -- 2.4.7 Propensity Scores. , 2.4.8 *Example: Statin Treatment Effects -- 2.4.9 Provider Comparisons -- 2.4.10 *Example: Ranking Cardiac Outcome -- 2.5 Concluding Remarks -- 3 Study Design for Prediction Modeling -- 3.1 Studies for Prognosis -- 3.1.1 Retrospective Designs -- 3.1.2 *Example: Predicting Early Mortality in Esophageal Cancer -- 3.1.3 Prospective Designs -- 3.1.4 *Example: Predicting Long-term Mortality in Esophageal Cancer -- 3.1.5 Registry Data -- 3.1.6 *Example: Surgical Mortality in Esophageal Cancer -- 3.1.7 Nested Case-Control Studies -- 3.1.8 *Example: Perioperative Mortality in Major Vascular Surgery -- 3.2 Studies for Diagnosis -- 3.2.1 Cross-sectional Study Design and Multivariable Modeling -- 3.2.2 *Example: Diagnosing Renal Artery Stenosis -- 3.2.3 Case-Control Studies -- 3.2.4 *Example: Diagnosing Acute Appendicitis -- 3.3 Predictors and Outcome -- 3.3.1 Strength of Predictors -- 3.3.2 Categories of Predictors -- 3.3.3 Costs of Predictors -- 3.3.4 Determinants of Prognosis -- 3.3.5 Prognosis in Oncology -- 3.4 Reliability of Predictors -- 3.4.1 Observer Variability -- 3.4.2 *Example: Histology in Barrett's Esophagus -- 3.4.3 Biological Variability -- 3.4.4 Regression Dilution Bias -- 3.4.5 *Example: Simulation Study on Reliability of a Binary Predictor -- 3.4.6 Choice of Predictors -- 3.5 Outcome -- 3.5.1 Types of Outcome -- 3.5.2 Survival End Points -- 3.5.3 *Examples: 5-Year Relative Survival in Cancer Registries -- 3.5.4 Composite End Points -- 3.5.5 *Example: Composite End Points in Cardiology -- 3.5.6 Choice of Prognostic Outcome -- 3.5.7 Diagnostic End Points -- 3.5.8 *Example: PET Scans in Esophageal Cancer -- 3.6 Phases of Biomarker Development -- 3.7 Statistical Power and Reliable Estimation -- 3.7.1 Sample Size to Identify Predictor Effects -- 3.7.2 Sample Size for Reliable Modeling -- 3.7.3 Sample Size for Reliable Validation. , 3.8 Concluding Remarks -- 4 Statistical Models for Prediction -- 4.1 Continuous Outcomes -- 4.1.1 *Examples of Linear Regression -- 4.1.2 Economic Outcomes -- 4.1.3 *Example: Prediction of Costs -- 4.1.4 Transforming the Outcome -- 4.1.5 Performance: Explained Variation -- 4.1.6 More Flexible Approaches -- 4.2 Binary Outcomes -- 4.2.1 R2 in Logistic Regression Analysis -- 4.2.2 Calculation of R2 on the Log-Likelihood Scale -- 4.2.3 Models Related to Logistic Regression -- 4.2.4 Bayes Rule -- 4.2.5 Prediction with Naïve Bayes -- 4.2.6 Calibration and Naïve Bayes -- 4.2.7 *Logistic Regression and Bayes -- 4.2.8 Machine Learning: More Flexible Approaches -- 4.2.9 Classification and Regression Trees -- 4.2.10 *Example: Mortality in Acute MI Patients -- 4.2.11 Advantages and Disadvantages of Tree Models -- 4.2.12 Trees Versus Logistic Regression Modeling -- 4.2.13 *Other Methods for Binary Outcomes -- 4.2.14 Summary of Binary Outcomes -- 4.3 Categorical Outcomes -- 4.3.1 Polytomous Logistic Regression -- 4.3.2 Example: Histology of Residual Masses -- 4.3.3 *Alternative Models -- 4.3.4 *Comparison of Modeling Approaches -- 4.4 Ordinal Outcomes -- 4.4.1 Proportional Odds Logistic Regression -- 4.4.2 *Relevance of the Proportional Odds Assumption in RCTs -- 4.5 Survival Outcomes -- 4.5.1 Cox Proportional Hazards Regression -- 4.5.2 Prediction with Cox Models -- 4.5.3 Proportionality Assumption -- 4.5.4 Kaplan-Meier Analysis -- 4.5.5 *Example: Impairment After Treatment of Leprosy -- 4.5.6 Parametric Survival -- 4.5.7 *Example: Replacement of Risky Heart Valves -- 4.5.8 Summary of Survival Outcomes -- 4.6 Competing Risks -- 4.6.1 Actuarial and Actual Risks -- 4.6.2 Absolute Risk and the Fine & -- Gray Model -- 4.6.3 Example: Prediction of Coronary Heart Disease Incidence -- 4.6.4 Multistate Modeling -- 4.7 Dynamic Predictions. , 4.7.1 Multistate Models and Landmarking -- 4.7.2 Joint Models -- 4.8 Concluding Remarks -- 5 Overfitting and Optimism in Prediction Models -- 5.1 Overfitting and Optimism -- 5.1.1 Example: Surgical Mortality in Esophagectomy -- 5.1.2 Variability Within One Center -- 5.1.3 Variability Between Centers: Noise Versus True Heterogeneity -- 5.1.4 Predicting Mortality by Center: Shrinkage -- 5.2 Overfitting in Regression Models -- 5.2.1 Model Uncertainty and Testimation Bias -- 5.2.2 Other Modeling Biases -- 5.2.3 Overfitting by Parameter Uncertainty -- 5.2.4 Optimism in Model Performance -- 5.2.5 Optimism-Corrected Performance -- 5.3 Bootstrap Resampling -- 5.3.1 Applications of the Bootstrap -- 5.3.2 Bootstrapping for Regression Coefficients -- 5.3.3 Bootstrapping for Prediction: Optimism Correction -- 5.3.4 Calculation of Optimism-Corrected Performance -- 5.3.5 *Example: Stepwise Selection in 429 Patients -- 5.4 Cost of Data Analysis -- 5.4.1 *Degrees of Freedom of a Model -- 5.4.2 Practical Implications -- 5.5 Concluding Remarks -- 6 Choosing Between Alternative Models -- 6.1 Prediction with Statistical Models -- 6.1.1 Testing of Model Assumptions and Prediction -- 6.1.2 Choosing a Type of Model -- 6.2 Modeling Age-Outcome Relations -- 6.2.1 *Age and Mortality After Acute MI -- 6.2.2 *Age and Operative Mortality -- 6.2.3 *Age-Outcome Relations in Other Diseases -- 6.3 Head-to-Head Comparisons -- 6.3.1 StatLog Results -- 6.3.2 *Cardiovascular Disease Prediction Comparisons -- 6.3.3 *Traumatic Brain Injury Modeling Results -- 6.4 Concluding Remarks -- Developing Valid Prediction Models -- 7 Missing Values -- 7.1 Missing Values and Prediction Research -- 7.1.1 Inefficiency of Complete Case Analysis -- 7.1.2 Interpretation of CC Analyses -- 7.1.3 Missing Data Mechanisms -- 7.1.4 Missing Outcome Data -- 7.1.5 Summary Points. , 7.2 Prediction Under MCAR, MAR and MNAR Mechanisms -- 7.2.1 Missingness Patterns -- 7.2.2 Missingness and Estimated Regression Coefficients -- 7.2.3 Missingness and Estimated Performance -- 7.3 Dealing with Missing Values in Regression Analysis -- 7.3.1 Imputation Principle -- 7.3.2 Simple and More Advanced Single Imputation Methods -- 7.3.3 Multiple Imputation -- 7.4 Defining the Imputation Model -- 7.4.1 Types of Variables in the Imputation Model -- 7.4.2 *Transformations of Variables -- 7.4.3 Imputation Models for SI and MI: X and y -- 7.4.4 Summary Points -- 7.5 Success of Imputation Under MCAR, MAR and MNAR -- 7.5.1 Imputation in a Simple Model -- 7.5.2 Other Simulation Results -- 7.5.3 *Multiple Predictors -- 7.6 Guidance to Dealing with Missing Values in Prediction Research -- 7.6.1 Patterns of Missingness -- 7.6.2 Simple Approaches -- 7.6.3 More Advanced Approaches -- 7.6.4 Maximum Fraction of Missing Values Before Omitting a Predictor -- 7.6.5 Single or Multiple Imputation for Predictor Effects? -- 7.6.6 Single or Multiple Imputation for Deriving Predictions? -- 7.6.7 Missings and Predictions for New Patients -- 7.6.8 *Performance Across Multiple Imputed Data Sets -- 7.6.9 Reporting of Missing Values in Prediction Research -- 7.7 Concluding Remarks -- 7.7.1 Summary Statements -- 7.7.2 *Available Software and Challenges -- 8 Case Study on Dealing with Missing Values -- 8.1 Introduction -- 8.1.1 Aim of the IMPACT Study -- 8.1.2 Patient Selection -- 8.1.3 Potential Predictors -- 8.1.4 Coding and Time Dependency of Predictors -- 8.2 Missing Values in the IMPACT Study -- 8.2.1 Missing Values in Outcome -- 8.2.2 Quantification of Missingness of Predictors -- 8.2.3 Patterns of Missingness -- 8.3 Imputation of Missing Predictor Values -- 8.3.1 Correlations Between Predictors -- 8.3.2 Imputation Model -- 8.3.3 Distributions of Imputed Values. , 8.3.4 *Multilevel Imputation.

Note: Steyerberg_FM.pdf -- Steyerberg_Ch01.pdf -- Steyerberg_Ch02.pdf -- Steyerberg_Ch03.pdf -- Steyerberg_Ch04.pdf -- Steyerberg_Ch05.pdf -- Steyerberg_Ch06.pdf -- Steyerberg_Ch07.pdf -- Steyerberg_Ch08.pdf -- Steyerberg_Ch09.pdf -- Steyerberg_Ch10.pdf -- Steyerberg_Ch11.pdf -- Steyerberg_Ch12.pdf -- Steyerberg_Ch13.pdf -- Steyerberg_Ch14.pdf -- Steyerberg_Ch15.pdf -- Steyerberg_Ch16.pdf -- Steyerberg_Ch17.pdf -- Steyerberg_Ch18.pdf -- Steyerberg_Ch19.pdf -- Steyerberg_Ch20.pdf -- Steyerberg_Ch21.pdf -- Steyerberg_Ch22.pdf -- Steyerberg_Ch23.pdf -- Steyerberg_Ch24.pdf -- Steyerberg_References.pdf -- Steyerberg_Index.pdf.