Advanced guide to research methods | |
Advanced handbook of methods in evidence-based healthcare. Andrew Stevens, Keith Abrams, John Brazier, Ray Fitzpatrick and Richard Lilford (editors). London: Sage, 2001 (xxx + 507 pp). ISBN 0 7619 6144 5. |
This book is largely a product of the Health Technology Assessment Research Programme in Britain, so it is odd that the words "health technology" don't appear in the title. It has a distinguished list of contributors, largely from the UK and Australia, and they cover a broad range of disciplines, including biostatistics, epidemiology, social science, health economics and ethics. The book summarises the research base which underpins evidence-based health technology assessment (HTA) with a mix of methodological approaches.
While generally falling short of providing checklists for activities such as designing instruments and outcome measures, the book does provide authoritative and informative background material for those involved in HTA and health services research. Because it is a collection of reviews rather than an integrated text, there is inevitably some repetition (for example, issues of validity and reliability are dealt with several times, in different ways), but the book comes together well and includes some important messages. The book would work best as a reference text. Although the reviews are not comprehensive, readers are pointed in the direction of more definitive texts. Areas of particular interest included sections on ethical issues in randomised controlled trials, with probably the best discussion of equipoise I've seen (emphasising the balance of societal and individual interests). There is a comprehensive review of the factors which limit the number, progress and quality of randomised controlled trials, and Diane O'Connell and other Australian colleagues provide a good summary of their work on applicability, which will be familiar to many MJA readers. It includes key information on methods for individualising treatment decisions, taking into account baseline risk and balancing benefits and harm. A section on quasi-experimental and observational research concludes that these trial designs do have a place, even though many studies are of poor quality. There is an intelligent discourse on when qualitative methods might have a place in HTA, and patient-based outcome measures are well covered, with helpful criteria in assessing these measures. The use of health status measures (such as SF-36) in economic evaluation are analysed in some depth. The section on conduct and design of questionnaires provides an excellent guide for researcher instrument designers. For the Bayesians* there is an excellent summary of the relevance of these methods in HTA, and clustered trial designs are well covered, with good examples from general practice. Some chapters go into considerable technical detail (particularly Billingham's contribution on simultaneous analysis of quality of life and survival data), but material such as this does belong, and broadens the book's appeal to a methodologically advanced audience. The section on consensus, reviews and meta-analysis includes some reasonably engaging descriptions of difficult concepts such as fixed versus random effects. There is a good section on quality assessment of clinical trials, with evidence supporting various quality assessment criteria, and what to actually do with quality assessment information. The final section concerns identifying and filling gaps in the evidence, and this should be required reading for all health technology agencies — it includes a particularly valuable description of horizon scanning for new technologies. A unique compilation of authoritative information on research methods as they apply to HTA, many will find it a valuable addition to their bookshelf. At a cost of $200, it is likely to have a specialised audience — possibly health service managers and advanced practitioners of health services research. David Weller * Bayesian methods are based on the idea that unknown quantities, such as population means and proportions, have probability distributions. The probability for a population proportion expresses our prior knowledge or belief about it, before we add the knowledge which comes from our data. BMJ 1998; 317: 1151
|