Modern Epidemiology

Citation

Lash, T. L., VanderWeele, T. J., Haneuse, S., & Rothman, K. J. (2021). Modern Epidemiology (4th ed.). Wolters Kluwer.

Chapter Summary

The book has 43 chapters in four parts, running to roughly 1,200 pages. The four named authors wrote most of the foundation, design and analysis chapters; contributors from many sub-disciplines wrote the more specialised ones, and the Stata Bookstore listing counts 35 further experts. The first three parts (foundations, design, analysis) form the methodological core, and Part IV consists of topic chapters that apply it.

Part I, Chapters 1–6: Foundations

The scope of epidemiology (1); causal inference and scientific reasoning (2); formal causal models (3); measures of occurrence (4); measures of effect and measures of association (5); and epidemiologic study design with validity and efficiency considerations (6). Causation is placed right after the opening chapter, ahead of measures and design. In a review in the European Journal of Epidemiology, Anders Ahlbom notes that the measures-of-occurrence chapter alone runs to about 25 pages, because real populations make incidence and prevalence harder to calculate than textbook life-table models suggest.

Part II, Chapters 7–15: Study Design and Interpretation

Cohort studies (7) and case-control studies (8) are followed by public health surveillance (9), field methods (10) and studies relying on secondary data (11). The rest of the part is about validity: confounding and confounders (12), measurement and measurement error (13), selection bias and generalizability (14), and precision and study size (15). Ahlbom observes that this part also includes data acquisition and principles for quantifying bias, which goes beyond the study-design coverage of most textbooks.

Part III, Chapters 16–31: Data Analysis

The analysis part starts with fundamentals of epidemiologic data analysis (16), introduction to categorical statistics (17), stratification and standardization (18) and categorical analysis of polytomous exposures and outcomes (19). Regression follows in two chapters: model specification (20), and model fitting and assessment (21). Then come time-to-event analysis (22), an introduction to Bayesian statistics (23), longitudinal and cluster-correlated data analysis (24), causal inference with time-varying exposures (25), analysis of interaction (26), mediation analysis (27), instrumental variables and quasi-experimental approaches (28), bias analysis (29), ecologic studies and analysis (30) and agent-based modeling (31). The publisher lists marginal structural models, instrumental variable analysis and structural nested models among the new or updated material.

Part IV, Chapters 32–43: Special Topics

Twelve contributed chapters apply the methods to a field: infectious disease epidemiology (32), reproductive epidemiology (33), psychiatric epidemiology (34), clinical epidemiology (35), molecular epidemiology (36), genetic epidemiology (37), injury and violence epidemiology (38), social epidemiology (39), environmental epidemiology (40), occupational epidemiology (41), nutritional epidemiology (42) and pharmacoepidemiology (43).

Key Concepts

Part I: Foundations

• Causal inference and causal models: Chapters 2 and 3 treat reasoning about causes and then formal models of causation, including causal diagrams.
• Measures of occurrence: incidence and prevalence, treated in detail in Chapter 4.
• Measures of effect and association: the quantities by which exposure and outcome are compared (Chapter 5).

Part II: Design and Validity

• Cohort and case-control designs: the two main observational designs, each with its own chapter.
• Confounding, measurement error and selection bias: the three validity threats that have a chapter each.
• Precision and study size: the planning counterpart to validity.

Part III: Analysis

• Stratification and standardization: the categorical tools that precede regression.
• Regression analysis: specification, fitting and assessment of models.
• Bayesian statistics and bias analysis: probabilistic approaches to uncertainty beyond random error.
• Mediation, interaction and time-varying exposures: causal questions that need dedicated methods.
• Instrumental variables and quasi-experimental approaches: designs for settings where confounding cannot be fully measured.

Part IV: Special Topics

• Field-specific epidemiology: infectious disease, reproductive, psychiatric, clinical, molecular, genetic, injury and violence, social, environmental, occupational, nutritional and pharmacoepidemiology.

Critical Analysis

Strengths

• Causation is built into the structure. Two chapters on causal reasoning and causal models come before measures and designs, which, in Ahlbom's reading, reflects the central place causal inference now has in the discipline. The same review notes that the authors do not claim to settle open questions, such as whether a cause must be manipulable.
• Validity gets its own methods. Beyond naming biases, the book gives separate chapters to confounding, measurement error, selection bias and a dedicated bias-analysis chapter, so quantification of bias is part of the standard toolkit rather than an afterthought.
• Analysis written for epidemiology. Ahlbom explains that the analysis part exists because epidemiologic data and designs need methods that general statistics books do not cover, and that the book keeps its long-standing advice against simple null-hypothesis testing in favour of the P-value function.
• Two ways to read it. The review expects some readers to read the entire book but more to use it as a reference and encyclopedia; the publisher similarly describes it as a comprehensive treatment of principles and methods.

Limitations

• Not an introduction. The book assumes an introductory epidemiology course and some biostatistics, and Part III moves quickly into regression, Bayesian methods and causal estimation.
• No meta-analysis chapter. Ahlbom expected a discussion of meta-analysis and pooling of studies and did not find one. The contents list has no such chapter, so readers who need to pool studies should use a separate source.
• Size and uneven authorship. At more than 1,100 pages with dozens of contributors, the specialised Part IV chapters are surveys of a field, written by different people, and are not a systematic course in each subfield.
• Methods change. The review points to artificial intelligence and big data as areas the book does not yet treat, and the fourth edition of 2021 is the latest.

Where It Fits in the Course

Modern Epidemiology is the graduate-level reference of this course list. Gordis Epidemiology is an introductory course text for readers without quantitative training, and Essential Epidemiology is a fuller course textbook for public health and the health professions; both should come first. Epidemiology: An Introduction develops causation, occurrence, design, bias and confounding from first principles with an emphasis on ideas over formulas, and it is the closest conceptual preparation for the way Modern Epidemiology is written.

Epidemiology: Study Design and Data Analysis (Woodward) is the closest companion on the statistical side, working through real data sets with methods of analysis. Modern Epidemiology's Part III covers some of the same territory more broadly and at a more advanced level, with the added focus on causal methods and bias analysis. Use Woodward when you need worked computations, and Modern Epidemiology when you need the conceptual justification, the limits of a method, or a topic that Woodward does not reach.

How to Use This Book

The book is aimed at graduate students in epidemiology and public health, doctoral researchers and practising epidemiologists who have finished an introductory epidemiology course and a biostatistics course that includes regression.

A practical plan: read Part I in order, because the later parts use its vocabulary of causation, occurrence and effect measures. Read Chapters 7, 8 and 12 to 15 next, which form the core of design and validity, and come back to them whenever you plan or appraise a study. Take Part III as a reference: read the fundamentals (16 to 19) and regression chapters (20 and 21) in sequence, and pick the others (time-to-event, Bayesian, interaction, mediation, instrumental variables, bias analysis) when a project requires them. Part IV is for the field you work in; the other chapters are useful as worked contexts for the same methods. Because the book does not cover meta-analysis, pair it with a separate source if you need to pool studies.