From the reviews:
"This book covers an important topic, because these prediction models are essential for individualizing diagnostic and treatment decision making. The topic is of increased importance as evidence-based medicine is increasingly implemented and as scientific and technological advances reveal new potential predictors of outcome. This book presents an approach for developing, validating, and updating prediction models.… [I]t provides ways to optimally utilize regression techniques to predict an outcome.… This book is written in a clear and accessible style.… [A]valuable resource for anyone interested in developing or applying a prediction model." (Todd A. Alonzo, American Journal of Epidemiology, 2009; Vol. 170, No. 4)
"Overall I think this is a well-written book, which will have a wide appeal. The idea of defining a strategy to deal with clinical prediction problems might be somewhat controversial, but considering the variable quality of statisticalanalyses that appear in the medical literature, I believe such an approach is desirable. The book appears to have struck a good balance between practical examples and descriptions of statistical techniques.... It is refreshing to see a practical book applying many modern regression techniques to real problems." (David Ohlssen, Journal of Biopharmaceutical Statistics, Issue 6, 2009)
"Dr Steyerberg … aims to provide an insight and also a practical illustration on how modern statistical concepts and regression methods can be applied in medical prediction outcomes. The book…will be of interest to those who work in medical cybernetics and indeed all cybernetics and systems researchers who are studying such medical problems and wish to apply statistical approaches and methodologies. It is worth examining the detailed contents list … and individual chapters may be of particular value to potential readers." (C. J. H. Mann, Kybernetes, Vol. 38 (6), 2009)
"The book … will be of interest to those who work in medical cybernetics and indeed all cybernetics and systems researchers who are studying such medical problems and wish to apply statistical approaches and methodologies." (C. J. H. Mann, Kybernetes, Vol. 38, No. 6, 2009)
“…and excellent practical guide for developing, assessing and updating clinical models both for disease prognosis and diagnosis. The book’s clinical focus in this era of evidence-based medicine is refreshing and serves as a much-needed addition to statistical modeling of clinical data. The book assumes a basic familiarity with modeling using generalized linear models, focusing instead on the real challenges facing applied biostatisticians and epidemiologists wanting to create useful models: dealing with a plethora of model choices, small sample sizes, many candidate predictors and missing data. This is an example-based book illuminating the vagaries of clinical data and offering sound practical advice on data exploration, model selection and data presentation. …The author uses simple simulations using a few reproducible R commands to motivate the use of imputation methods and shrinkage. These simple but illuminating illustrations are one of the highlights of the book and serve as excellent pedagogical tools for motivating good statistical thinking. …” (International Statistical Review 2009, 77, 2)
“This is an excellent text that should be read by anyone performing prediction modeling. … the text has three audiences epidemiologists and applied biostatisticians who want to develop or apply a prediction model health care professionals who want to judge a study that presents a prediction model and theoretical researchers … . I found the book very useful and I believe clinicians and policy makers will be similarly well served. … All are excellent summaries for readers and provide links to resources for further investigation.” (Chris Andrews, Technometrics, Vol. 53 (1), February, 2011)
Reviewer: Jarrod E. Dalton, Ph.D.(Cleveland Clinic)
Description: This book covers domain-specific issues in development, validation, and application of prediction algorithms for decision-making in medicine and provides several useful updates that coincide with the expansion of data availability in the biomedical sciences since the first edition was published in 2009. It discusses unique aspects of clinical prediction methodology under a range of epidemiologic study designs (randomized controlled trials, cohort studies, diagnostic studies, prognostic studies, etc.), as well as models and validation approaches (including graphics and appropriate statistical measures of model accuracy) for all standard outcome types (e.g., continuous, binary, and survival). It covers practical issues in clinical prediction science, including addressing missing data, selection of predictor variables, modeling nonlinear relationships, and implementing the model for clinical decisions is demonstrated through a series of examples.
Purpose: The purpose is to fill a gap for many trainees in the medical and statistical sciences who, for various reasons, may lack adequate skills in the art of modeling. The lack of skills has been reflected in the growing number of peer-reviewed publications using suboptimal or flawed prediction modeling methods. Detailed, outlined, and thoroughly referenced, this book presents readers with a path to excellence in clinical prediction science and enables them to be more strategic in developing, validating, and deploying their models.
Audience: This is a thorough and up-to-date book on the state of the art in clinical prediction science, written by a person who has significantly impacted the community since it was only a niche field. Students will benefit from this book as it goes beyond the basic understanding of how different algorithms work and how to code them. Medically trained scientists will benefit from its organization and structure, making it easy for them to navigate the topic without falling down the rabbit holes. Even skilled prediction scientists, who have all benefitted from the author's contributions over several decades, will find the book valuable as a reference and guide to important papers in the field.
Features: This book covers all aspects of clinical prediction science, including issues in study design, sample size considerations, suitable model classes, and validation methods for different types of outcome variables (e.g., continuous, binary, categorical, ordinal, time-to-event, and mixed outcome types). It describes strategies for reducing the risk of bias associated with common problems that arise in model development, including imputation of missing data; model selection in the presence of large numbers of predictor variables (relative to the effective sample size) and/or collinear predictor variables; reducing the influence of outlier observations; and nonlinear relationships. Measures and graphical techniques for assessing internal and external validity of clinical prediction models are provided for each type of outcome variable. The presentation of model predictions in the clinical setting, which is a complex topic that incorporates human elements not routinely discussed in statistical training, is discussed. Finally, these techniques are put into practice through a series of examples.
Assessment: This book excels in its strategic (rather than technical) orientation and is appropriate for anyone interested in clinical prediction science. Many high-quality books on the science of prediction modeling, such as The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd edition, Hastie et al. (Springer, 2009), can be inaccessible to those newer to the field because of the high level of mathematical rigor. This book doesn't duplicate those efforts, but rather serves as a scaffolding for readers to navigate the field. In particular, the book uniquely addresses issues in translating models to clinical practice, and for that reason alone it should be required reading for those who attempt to do so.
