Annals of Epidemiology 25 (2015) 63e64
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Book review Porta Miquel, editor. A Dictionary of Epidemiology. Sixth Edition. New York: Oxford University Press; 2014. xxxii, 343 pages, $29.26, paperback.
Introduction This dictionary is not merely a list of explanations of epidemiologic terms but also a distillation and mirror of contemporary controversies. A dictionary may attempt to subdue or at least tame controversy through the provision of a standard or even consensual view on key concepts and methods, but the effect is unpredictable. As a reviewer of the new edition of A Dictionary of Epidemiology, I have been exceptionally fortunate for three reasons. First, I am passionate about the meanings of words and a compiler of explanations (with modest contributions to the fourth edition and comprehensive glossaries in my books). Second, I have known both the editors who have shaped this dictionary (Last and Porta) and have insights on their approach. Third, I attended the workshop launching the dictionary at the 20th World Congress of Epidemiology (WCE) in Alaska, 2014. At the workshop, Miquel Porta emphasized the revolutions underway in epidemiology and challenged us to spot the changes (e.g., his changes to risk factor) and his jokes. Alex Broadbent considered the dictionary’s approach to measures of causal strength and concluded it was not just mapping the terrain but shaping it. Neil Pearce warned us that the science of epidemiology is not its tools but its subject matter, and gave a reminder that a dictionary should reﬂect the language people use, not what they should use. Given this background, what did I think of the sixth edition? My method of examination of the dictionary First, I read the preface and then “surfed” the contents. In surﬁng mode, the dictionary was a source of fascinating insights in the subjects I know and a challenging and tantalizing introduction to subjects I do not know. Second, I chose, in advance, the topics in the box. Third, I examined a few terms that came up in a day’s work. Finally, after the WCE workshop, I had to see what Porta did with the term “risk factor” and to look for his jokes. Findings I have all six editions of the dictionary. The ﬁrst edition (published 1983) came into my possession in September 1985, a few months after I entered academic public health, and as I started teaching. The ﬁrst edition was 114 pages with two pages of bibliography and no references. The latest dictionary is a bumper editiond299 pages on the meanings of terms, 800 references and six pages of bibliography of key books and readings. http://dx.doi.org/10.1016/j.annepidem.2014.09.011
A dictionary of a multidisciplinary ﬁeld like epidemiology needs to be a ready source of sound deﬁnitions for both epidemiologists and allied professionals and their students. For it also to meet the needs of the public would be asking for the impossible, and the impossible is not achieved here for much of the dictionary is technical. Even for the professional epidemiologist, the dictionary is often a hard read (so keep other dictionaries at hand). The dictionary is on epidemiology not epidemiologists or epidemiologic history. The dictionary is a remarkable collaborative work that is an essential companion in the serious study of, and engagement with, epidemiology and its codependent disciplines. Let me illustrate some observations on speciﬁc terms. Epidemiology Twenty-seven subtypes are listed, and I was surprised to learn that it includes the study of plants (new entry). Epidemiology is the study of the health of populations, so why not, indeed, populations of all living things including plants? Cause There is no entry but there are about three pages on causality in its various guises, including causal diagrams, which are not, however, explained. The entries give glimpses into the increasingly inﬂuential world of counterfactual approaches to causality. Prevalence and incidence We might expect the treatment of these two fundamental and interrelated measures to be similar but prevalence is described as a measure of disease occurrence, whereas incidence is not. The distributed efforts of many contributors, each with their own style and preferences, are exempliﬁed by these distinct entries for these core measures. As in the ﬁfth edition, the dictionary states prevalence is a proportion not a rate, but it does not use the phrase prevalence proportion. Pearce’s comment at the workshop is worth reﬂecting on i.e. a dictionary reﬂects the language people use, not what they should use. Although many people (especially in disciplines allied to epidemiology but also in epidemiology) still use “prevalence rate,” the more aware simply evade the issue of terminology by saying/writing the loose abbreviated form “prevalence,” with the correct phrase “prevalence proportion” remaining extremely rare (the same applies to incidence proportion). So we have moved from the slightly inaccurate “prevalence rate” to a mixture of the loose “prevalence” and the continuing use of rate (the same applies to incidence proportion) rather than the unfamiliar but correct “prevalence proportion.”
Book review / Annals of Epidemiology 25 (2015) 63e64
Conﬁdence interval Box. Terms examined for this book review Terms I chose in advance to sample the dictionary Broad definitions: epidemiology, screening, social epidemiology Core concepts: causality, confounding, interaction, ethnicity Study design: ecological study, Mendelian randomization, case control study Techniques: person-time denominator, structural equation modeling, and directed acyclic graphs Measures: prevalence, incidence, numbers needed to treat History and people: Bradford Hill, Framingham Study, Golberger Terms I examined in the course of a day’s work Poisson regression, robust variance, confidence interval, Mantel Haenszel, Peto method, inverse variance Terms I examined as a result of the WCE workshop Risk factor, risk markers, risk indicator
The technical deﬁnition provided sent me scurrying to my statistics dictionary where the explanation was much simpler. Relative risk The editor’s bold changes did not work for me. From the ﬁrst edition through to the ﬁfth, this phrase is fairly consistently explained as having three meanings but in essence, it refers to an associated factor where causality remains contested. In the sixth edition, it is said to be “A factor that is causally related to a change in the risk of a relevant health process, outcome, or condition.” If so, the phrase risk factor is redundant because causal factor could and should be used. By contrast, a noncausal association is said to be a risk marker. Many examples of risk factor are given, for example, socioeconomic characteristics, but such factors are not agreed to be causal. The trouble is that we are able to demonstrate associations but categorizing them as causal or noncausal is problematic, an area where much work is required. Indeed, at the WCE, I led a workshop on a proposal for a World Council on Epidemiology and Causality to confront this challenge (report on workshop under preparation). Conclusions
Ecological If, as stated, this is a study where the units of analysis are populations rather than individuals, does that not mean it is all epidemiology? I think the non-epidemiologist could conclude so.
Mendelian randomization This is a very helpful entry but the word canalization (part of the entry) is not in the dictionary or in the other dictionaries on my shelf. Contributors and editors should not assume knowledge of these kinds of words.
Case-control study Given the emphasis on modernizing epidemiologic terminology, it is surprising to ﬁnd this described as a study of cases compared with people without the disease. It is even more surprising to ﬁnd in the next paragraph that this is incorrect! What happened here?
Poisson regression This is not listed although the relevant distribution is shown on p. 83 (no cross link of ﬁgures to terms).
The dictionary is a credit to the collegiate approach of epidemiology and to the International Epidemiological Association. It is a vital resource for students, practitioners, teachers, and researchers. This sixth edition is a work in progress but one that hopefully will never ﬁnish, and apparently, work on the seventh edition is already underway. In addition to bringing us explanations of the new terms arising from the rapid advances of our discipline, I hope the seventh edition will also bring a greater harmonization of entries, simpliﬁcation (perhaps to the level of introductory epidemiology in a masters of public health course), and less technical and more natural language in the explanation of terms, especially those from other disciplines such as statistics and genetics. Anyone can contribute to the dictionary. This democratic and inclusive approach, however, demands tight editorial control to maintain cohesiveness. The passion and energy of Porta and the previous editor, Last, are the driving forces behind the success of the dictionary. I, and I am sure all who study, practice and above all love epidemiology, thank them. Raj S. Bhopal, CBE, MD, DSc (Hon) Bruce and John Usher Professor of Public Health Edinburgh Ethnicity and Health Research Group Centre for Population Health Sciences University of Edinburgh Teviot Place Scotland UK EH8 9AG E-mail address: [email protected]