Data Handling in Epidemiology

EPIDEMIOLOGY. Co-ordinating Editor, W. W. Holland. (Pp. viii+212; figs. 26. £3.25). London: Oxford University Press, 1970. THIS book deals, in 15 chapters by different authors, with seven topics: asking the question, study design, collection of data, record linkage, data processing, analysis, and implications. They discuss most aspects of epidemiological research, from formulating hypotheses to interpreting statistical analyses of results. Thus, many, if not all, chapters should interest not only epidemiologists but also other medical research workers. Several authors of the early contributions make the same plea, namely, that rigorous consideration must be given to the design of studies and the data collected in them. Several authors emphasise the need to define study objectives and formulate hypotheses before considering other aspects of a study. As one author comments, "A casual 'look-see' attitude usually leads to nothing but confusion, and the use of data collected during a study to test a hypothesis which is formulated after the event, may invalidate the conventional methods of testing statistical significance." Part 4, on medical record linkage, includes two contributions. The first is concerned with the possible application of a system of linked medical records; for example, as a tool for measuring morbidity; as a means of retrieving individual records easily and quickly; and as an aid to patient management. The second contribution describes briefly a series of studies undertaken in the last two decades among some French-speaking African communities south of the Sahara. The fifth part of this book is concerned with data processing. The first chapter in this section describes the use of a computer for processing of information collected during a study carried out in the Greater London Borough of Lambeth (details are also given of the design of the study), while the second chapter deals in general terms with computer's role in data processing and its impact on epidemiological studies. Analysis of data is considered in Part 6. While not claiming to be fully comprehensive, the three chapters in this section give useful examples of building models for infectious diseases; the use of multivariate analyses in psychiatry; and the use of models to study survival. The non-mathematical reader may find parts of these chapters rather heavy going. The findings of epidemiological studies may have certain implications with regard to health policies and these are considered in Part 7. The example chosen, automobile accidents, demonstrates the three major ways of attacking health problems, namely, by environmental health measures, by medical care, and by health education. The uses and implications of computerised data processing techniques for automated multiphasic screening procedures are discussed. In a book which covers so much ground it is obviously impossible for each author to have sufficient space to consider fully all aspects of his topic. An additional bibliography, or recommended reading list, at the end of each chapter would have been a useful addition to help the reader whose interest had been aroused. The reader's attention is drawn to the following errors in the book: l. Page 171, item 3, line 1. The formula referred to should be (3) and not (2). 2. Page 186. The "is" in line 20 and the "and" in line 21 should be deleted so that this definition conforms to that of the preceding paragraph. J.D.M.