Psychiatric epidemiology: Looking to the future

In this final chapter to the second edition of Practical Psychiatric Epidemiology, developments in psychiatric epidemiology since the first edition are summarized and the editors offer a view on where the future may lie. The themes summarized in this chapter include those related to large-scale datasets or ‘big data’, new technologies and science communication (including data generated through GPS tracking systems and the impact of social media), expanding biological data and biobanks, as well as the impact of globalization, migration, and culture on understanding psychiatric epidemiological principles. The last part of this chapter raises the important issue of open science initiatives. The chapter concludes with a brief discussion on the constancy and ongoing evolution of psychiatric epidemiology.

Bio-informatics and psychiatric epidemiology

This chapter highlights the methodologies which are increasingly being applied to large datasets or ‘big data’, with an emphasis on bio-informatics. The first stage of any analysis is to collect data from a well-designed study. The chapter begins by looking at the raw data that arises from epidemiological studies and highlighting the first stages in creating clean data that can be used to draw informative conclusions through analysis. The remainder of the chapter covers data formats, data exploration, data cleaning, missing data (i.e. the lack of data for a variable in an observation), reproducibility, classification versus regression, feature identification and selection, method selection (e.g. supervised versus unsupervised machine learning), training a classifier, and drawing conclusions from modelling.

Genetic epidemiology: Overview

This chapter demonstrates the principles behind some of the major genetic study designs used in psychiatry research. The first part focuses on behavioural genetic designs, while the second part describes designs for ‘gene mapping’. Behavioural genetics examines the genetic basis of behavioural phenotypes, including both disorders and ‘normal’ dimensional traits. The theoretical basis is derived from population genetics, including properties such as segregation ratios, random mating, genetic variance, and genetic correlation between relatives. The second part of the chapter deals with gene mapping designs, in which specific genetic variants or genomic regions associated with a disorder or trait are identified. A brief outline of the most popular current approaches to the analysis of the genetics of complex human disorders is also provided.

Inference 2: Causation

Most epidemiological research, beyond the simple descriptive study, is attempting to elucidate a causal relationship. This chapter continues the consideration of causal inference in a broader context, covering the principles of inductivism and refutationism that emerged in seventeenth-century Western philosophy and which have had profound influences on modern science. However, life sciences cannot rely on perfectly controlled experimental conditions and consequently a number of other principles have had to be developed to allow knowledge to accumulate despite uncertainties in hypothesis testing. For epidemiology, these include the principle of consensus (repeated experiments contributing to a ‘verdict of causality’) and the causal criteria outlined by Bradford Hill. Finally, the potential combinations of variables under investigation (causal, confounding, mediation, effect modification) are considered in relation to analysis designs.

Surveillance, case registers, and big data

Health and social care organizations routinely collect huge amounts of data, which are a potentially useful resource for researchers who wish to study clinical issues ‘in vivo’. This chapter describes the types of research that typically use these data in mental healthcare, with a focus on surveillance, case register, and ‘big data’ approaches. To start, the chapter illustrates the importance of surveillance in the study of mental health, with a focus on using clinical surveillance to study patterns of rare psychiatric disorders and events. It then summarizes the contributions of psychiatric case registers to descriptive, analytical, and trial study designs. Finally, it reviews how digitized health and social information are being linked and analysed using big data techniques. Throughout, the chapter outlines the strengths and weaknesses of these epidemiological approaches, and provides practical guidance on how researchers may address the methodological and governance challenges that these clinical data sources present.

Randomized controlled trials

This chapter on randomized controlled trials (RCTs) considers some of the key factors in the design, conduct, analysis, and interpretation of RCTs. The chapter provides an overview of what constitutes an RCT and why they are needed. The chapter also provides an overview of the major practical elements of the design and conduct of RCTs, including undertaking a background review of literature, the need for formulation of a clear primary hypothesis and objective, selection and definition of the study population, collecting outcomes at baseline and follow-up, and appropriate methods of statistical analysis and inference. The chapter concludes with a consideration of the need for clinical trial units, complex interventions, and alternative RCT designs.

Introduction: A ‘fourth age’ of psychiatric epidemiology?

This is the introduction to the second edition of ‘Practical Psychiatric Epidemiology’ published by Oxford University Press. In this introduction the Editors reflect on developments since the first edition. Themes touched upon include the ongoing need for high quality descriptive data, the contribution of wearable devices and technologies to generating data for psychiatric epidemiological studies, developments relating to the availability of large-scale data resources or so-called ‘big data’ in psychiatric epidemiology, ongoing issues relating to accurate measurement in psychiatric epidemiology and the contribution of complex interventions to effective healthcare service delivery, and in particular, the way in which these are effectively implemented. The chapter concludes with a reflection on the continued importance of psychiatric epidemiology to the field of mental health.

Statistical techniques in psychiatric epidemiology

Epidemiology and medical statistics have been partner disciplines since the nineteenth century, despite disagreement between their founding fathers. This chapter begins with a summary of the discrete uses of statistical techniques in epidemiological research, followed by some guidance on constructing regression models—a common task, but one which it is important to think through carefully. This topic is developed further through a more detailed consideration of mediating and causation, which are particularly important in psychiatric epidemiology because of the long and complex pathways of causation, and lifelong interrelationships between exposure and outcome states characteristic of mental disorders and their risk factors. Finally, four key emerging themes are considered: the use of propensity scores, dealing with missing data, multilevel modelling, and latent class analyses.

Critical appraisal

This chapter provides an introduction to the principles of critical appraisal and explains why critical appraisal skills are important in practice and research. Guidance is provided on how to approach the critical appraisal of different types of study including cross-sectional studies, case–control studies, cohort studies, clinical trials, systematic reviews, and qualitative studies. A worked example is provided at the end of the chapter to illustrate the process. Developing skills in critical appraisal will help readers to assess the credibility, relevance, and value of the results of research and is an essential component of practising evidence-based medicine.

Inference 1: Chance, bias, and confounding

Inference describes the process of deriving conclusions from observations to generalizations and is a key activity in all research. This chapter commences with considering how the findings from a research sample can be applied to the population from which that sample was drawn—that is, the role of chance in accounting for observed findings, and the possibility that they might have arisen because of errors in the design of the study (bias)—whether relating to the people in the sample (selection bias) or the measurements applied (information bias). The chapter then begins to consider the extent to which a causal relationship can be inferred from an observed association, by considering the role of confounding factors as alternative explanations and ways in which these are addressed in statistical analyses.