Systematic Reviews in the Social Sciences Systematic Reviews in the Social Sciences A PRACTICAL GUIDE Mark Petticrew and Helen Roberts (cid:2)2006byMarkPetticrewandHelenRoberts BLACKWELLPUBLISHING 350MainStreet,Malden,MA02148-5020,USA 9600GarsingtonRoad,OxfordOX42DQ,UK 550SwanstonStreet,Carlton,Victoria3053,Australia TherightofMarkPetticrewandHelenRobertstobeidentifiedastheAuthorsofthisWork hasbeenassertedinaccordancewiththeUKCopyright,Designs,andPatentsAct1988. Allrightsreserved.Nopartofthispublicationmaybereproduced,storedinaretrieval system, or transmitted, in any form or by any means, electronic, mechanical, photo- copying, recording or otherwise, except as permitted by the UK Copyright, Designs, andPatentsAct1988,withoutthepriorpermissionofthepublisher. Firstpublished2006byBlackwellPublishingLtd 1 2006 LibraryofCongressCataloging-in-PublicationData Petticrew,Mark. Systematicreviewsinthesocialsciences:apracticalguide/MarkPetticrewandHelen Roberts. p.cm. Includesbibliographicalreferencesandindex. ISBN-13:978-1-4051-2110-1(hardcover:alk.paper) ISBN-10:1-4051-2110-6(hardcover:alk.paper) ISBN-13:978-1-4051-2111-8(pbk.:alk.paper) ISBN-10:1-4051-2111-4(pbk.:alk.paper) 1. Social sciences—Research—Methodology. 2. Social sciences—Statistical methods. I.Roberts,Helen,1949–II.Title. H62.P4572005 300’.72’3—dc22 2005011632 AcataloguerecordforthistitleisavailablefromtheBritishLibrary. Setin10.5pt/12.5ptBembo bySPIPublisherServices,Pondicherry,India PrintedandboundintheUnitedKingdom byTJInternational,Padstow,Cornwall The publisher’s policy is to use permanent paper from mills that operate a sustainable forestry policy, and which has been manufactured from pulp processed using acid-free andelementarychlorine-freepractices.Furthermore,thepublisherensuresthatthetext paperandcoverboardusedhavemetacceptableenvironmentalaccreditationstandards. Forfurtherinformationon BlackwellPublishing,visitourwebsite: www.blackwellpublishing.com Contents Foreword – William R. Shadish vi Acknowledgments x Preface xiii Chapter 1 Why do we need systematic reviews? 1 Chapter 2 Starting the review: Refining the question and defining the boundaries 27 Chapter 3 What sorts of studies do I include in the review? Deciding on the review’s inclusion/exclusion criteria 57 Chapter 4 How to find the studies: The literature search 79 Chapter 5 How to appraise the studies: An introduction to assessing study quality 125 Chapter 6 Synthesizing the evidence 164 Chapter 7 Exploring heterogeneity and publication bias 215 Chapter 8 Disseminating the review 247 Chapter 9 Systematic reviews: Urban myths and fairy tales 265 Glossary 277 Appendix 1 The review process (and some questions to ask before starting a review) 284 Appendix 2 MOOSE Guidelines 288 Appendix 3 Example of a flow diagram from a systematic review 291 Appendix 4 Example data extraction form 293 Appendix 5 Variations in the quality of systematic reviews 296 Bibliography 298 Index 324 Foreword The idea of evidence-based practice is not new. For example, nearly 60 years ago, the scientist-practitioner model of clinical psychology in the US was based in part on the idea that clinical psychologists would scientifically investigatetheeffectivenessoftheirowntreatments,andusewhatprovedto beeffective.Around100yearsago,theBritishArmyaskedSirKarlPearson to review the evidence about the effects of a new typhoid vaccine, to help them decide whether to adopt that vaccine more widely.1 Such relatively isolated examples are common for hundreds of years. What is new is the widespread acceptance that evidence-based practice is practice whose time has come. Whether it be through the Cochrane Collaboration, the Camp- bell Collaboration, the What Works Clearinghouse in the US Department of Education, or any of a host of similar efforts, the idea that practices and policies that are supported by scientific evidence can be identified and disseminated has now been institutionalized. That in itself is a remarkable accomplishment. In this context, this book by Petticrew and Roberts is a timely gem, full of detailed history, pithy quotes, marvelous examples, compelling argu- ments, practical advice, and transparent demonstrations of how to do one of the most common ways of putting the evidence into evidence-based practice–thesystematicreview.Practicaladviceleapsoutatthereaderfrom every chapter, with novel information I have not seen in related works. Letmestartwithreferencetothekeyjustificationforsystematicreviews, that ‘‘single studies taken in isolation are often seriously misleading’’ (chapter 1, this volume). Or as I tend to put it: Never trust the results of singlestudies.Oneneedonlygraphafrequencydistributionofeffectsizesto see that some studies will show unusually positive effects, and some un- usuallynegativeeffects,substantiallybyvirtueofchance.Yetthepublic,the media,andeven other scientists routinelyplace greatfaith in single studies. FOREWORD vii Asaresult,ineffectiveorinaccuratepolicies,practices,orideasareadopted, accurate or effective ones are dismissed, and all parties get confused. The analogy that Petticrew and Roberts make with a survey is a good one. We would never take seriously the results of a sample survey where the sample was of a single person, and we all understand why – we will often get a differentanswerbyaskingadifferentperson.Thesameistrueinsystematic reviews. We will often get a different answer when we look at a different study.Insamplesurveys,onlyalargesampleiscapableofgivingusareliable answer. In reviewing literature on a question, only the systematic review process is capable of helping to clarify where the answer really lies. Yet as Petticrew and Roberts point out in chapter 2 and elaborate in chapter 3, we have had an unnecessarily narrow view of what a systematic review can do – answer questions about cause and effect relationships. Systematic reviews need not only address what works, as important as that question is. Such reviews can address the evidence about nearly any kind ofquestion,forexample,abouttheprevalenceofaproblem,thecorrelation of two variables to each other, the implementation of a program, the meaningprogramstakeholdersascribetoaprogramoritscontext,thecausal mediating processes that help us understand why it works, and program costs, to name just a few. Similarly, our understanding of the question of whatworkshasoftenbeentoonarrow.Thatquestioninvolveshighquality evidence about cause and effect relationships, to be sure. But we are also informed about what works by, for example, knowledge that some stake- holdergroupsrefusetoparticipate,thatcostsaretoohightobepractical,or that the need for the treatment is questionable. What is less clear is whether we have well-developed technologies for addressing some of these questions. The technological problem is twofold, howtointegratestudiesthatareallofonekind(e.g.,aboutneed)witheach other,andhowtointegratesstudiesofonekind(e.g.,need)withstudiesof another kind (e.g., treatment effectiveness). The former problem is relat- ivelyunproblematicformostkindsofquantitativestudies,butisfarlessclear forqualitativestudies.Thelatterproblemisalmostcompletelyunaddressed. It is this context that makes me so pleased to see the effort Petticrew and Roberts have put into both these problems. Many people talk about the value of integrating quantitative and qualitative studies, for example, but actualexamplesof such integration are few andfarbetween.Petticrew and Roberts lay out the logic of such integrations, some of the methods that might be used for doing them, along with examples. And they don’t shy away from the difficulties such as a lack of standardized search terms to use inlocatingqualitativestudiesonagiventopic.Formanyyears,whencritics havedecriedthenarrowfocusonexperimentsinsystematicreviewsofwhat viii FOREWORD works, I have responded by saying we are working on that problem. Now I can actually point to the work. Iwasalsopleasedtoseethehugenumberofreferencesinchapter5about assessingthequalityofdifferentkindsofresearchdesigns.Justthecompila- tionitselfisamajorcontribution.Yetweshouldrememberthatqualitycan meantwodifferentthings:(a)qualityoftheimplementationofthemethod (e.g., did the primary study author implement the qualitative method properly); and (b) quality of the inferences from the method (do the methods used support a causal inference). My sense is that the latter kinds of quality scales are pretty common for experiments, though serious ques- tion remains about whether those scales are well enough developed to warrant widespread use. But inference quality scales are far less well devel- oped for qualitative methods, where the available scales seem more con- cernedwithwhetherthemethodswereimplementedwell.Theproblemis that good implementation of a method is orthogonal to whether that method supports a given inference. The best-implemented randomized experiment is rarely going to tell us much about the need for an interven- tion, just as the best implemented ethnography is rarely going to yield a strong inference about cause-and-effect relationships. Finally, chapter 8 on dissemination and use of systematic reviews raises a crucial issue. Too often we assume that policymakers and practitioners are likelytoadoptinterventions thatareshown towork. Thatassumptionisat bestonlypartlyright.Welearnedthatthehardwayinthefieldofprogram evaluation in the 1960s and 1970s. The early emphasis in that field was on experimentationtodetermineeffectiveinterventionsthatcouldbeadopted aspolicy.Wequicklyfoundoutthattheywerefrequentlynotadoptedeven whentheywereeffective,andbegantostudytheprocessofuseitself.2The fate of Fairweather’s Lodge for the chronically mentally ill is illustrative.3 Thatinterventionwasfarmorecosteffectivethanthealternatives,waswell- liked by patients and successful on most outcome criteria; but it was relatively little used despite extensive dissemination funding because it had no funding mechanism, conflicted with the extant treatment institutions, and required citizens to accept a house of patients next door. Adopting effective practices involves far more than knowing what works, and Petti- crew and Roberts do a commendable job of laying out the issues. The lessons learned are remarkably similar to those we learned in program evaluation years ago, and we would do well to revisit the modern versions of that literature (e.g., Patton)4 to help us understand more about how to make our work useful. As you can see, I am a fan of what Petticrew and Roberts have done in thisbook.Theylayoutabroadandambitiousagendafortheartandscience FOREWORD ix of systematic reviews, an agenda that cogently presages issues that I think will be at the cutting edge of this field of inquiry in the coming decades. I hope you enjoy it as much as I did. William R. Shadish University of California, Merced REFERENCES 1. Pearson,K.Reportoncertainentericfeverinoculationstatistics.BritishMedical Journal1904,3:1243–6. 2. Shadish, W. R., Cook, T. D., and Leviton, L. C. Foundations of program evalu- ation:Theoriesofpractice.NewburyPark,CA:SagePublications,1991. 3. Shadish, W. R. Policy research: Lessons from the implementation of deinstitu- tionalization.AmericanPsychologist1984,39:725–38. 4. Patton,M.Q.Utilization-focusedevaluation(3rdedn.).NewburyPark,CA:Sage Publications,1996. Acknowledgments Writingbooksincursalargenumberofdebts.Thisoneisnodifferent,and we would particularly like to thank Susan Kennedy and Rodney Barker, who have gone well beyond the usual demands of being our partners by reading, commenting, and constructively criticizing. Professor Will Shadish, one of the foremost meta-analysts in the USA readily and generously agreed to write a foreword. The anonymous peer reviewers of our proposal and, in particular, the reviewer of our final manuscript contributed a great deal to our thinking and the final product. At our request, Blackwell approached them to ask if they would forgo anonymity so that we could acknowledge them by name, and we are happytosaythattheyagreedtothis.Manythanksthen,toTrevorSheldon, at the University of York, who nobly reviewed both the proposal and the book; and to the proposal reviewers Jim Connelly, University of Leeds, Emma Davies, Auckland University of Technology, David DuBois, Uni- versityofIllinoisatChicago,LoraineGelsthorpe,UniversityofCambridge, David Gough, Institute of Education, University of London, Geraldine Macdonald, Commission for Social Care Inspection, Susan Michie, Uni- versity College London, Deborah Phillips, Georgetown University, and Maggie Savin-Baden, Coventry University. We are also grateful to Kate Pool from the Society of Authors. OurcolleaguesinGlasgowandCityUniversities,ourco-researchersand colleagues on the ESRC-funded project on narrative synthesis: Jennie Popay, Amanda Sowden, Nicky Britten, and the late Sally Baldwin, Lisa Arai, Mark Rodgers, and Katrina Roen, and our colleagues in the ESRC Evidence Network, and the Cochrane and Campbell Collaborations (par- ticularly those whose work we have quoted) have all been parties to discussions and conversations that have helped inform our work. ACKNOWLEDGMENTS xi MarkPetticrewwouldparticularlyliketothankcolleaguesintheESRC Centre for Evidence-Based Public Health Policy, and in his research pro- gram (funded by the Chief Scientist Office of the Scottish Executive Department of Health), whose work and thinking the book draws on: among them Pam Attree, Beth Milton, Clare Bambra, Frances Drever, Matt Egan, Hilary Graham, Val Hamilton, Sally Macintyre, Liz McDer- mott,DavidMorrison,DavidOgilvie,SianThomas,HilaryThomson,and Margaret Whitehead. Colleagues elsewhere who read and commented on partsofthebook,orwhokindlygavepermissiontousetheirwork,include IainChalmers,IainCrombie,the EPOC(Effective Practice andOrganiza- tionofCare)groupoftheCochraneCollaboration,SimonGilbody,David Gough and colleagues at the EPPI Centre, Rebecca Ivers and colleagues, Betsy Kristjansson, Geoffrey Nelson and colleagues, Anthony Petrosino, Steve Platt, Scott Reeves, Roberta Scherer, Fujian Song, Peter Tugwell, and Paul Wilson. Thanks for permission to reproduce various figures or tables go to John Wiley & Sons Ltd (the Forest plots in chapters 6 and 7, which are copyright of the Cochrane Library), The Sainsbury Archive (Figure 6.2 in chapter 6), Hawaii Medical Library, the National Portrait Gallery (for the portrait of Francis Bacon in chapter 4), the British Medical Journal (for permission to reproduce material in chapter 9), David Ogilvie (the photograph in chapter 6), and the Centre for Reviews and Dissemin- ation(CRD)attheUniversityofYorkforpermissiontoreproducematerial from their systematic reviews guidance (CRD Report 4). Thanks also to Mary Robins for library support. Helen Roberts is particularly grateful to colleagues at City University, especially current and past colleagues in the Child Health Research and Policy Unit, and to fellow members of the What Works for Children ‘‘node’’ funded as part of the ESRC evidence network – Trevor Sheldon at the University of York, Kristin Liabo, Madeleine Stevens, Greg Khine, and Alison Moore at City University and Di McNeish, Tony Newman, Sarah Frost, Angela Hutton, and Philippa Roberts at Barnardo’s. Shruti Uppal provided valuable assistance with reference and Web checking. Helen has learned a great deal from colleagues at the EPPI Centre at the UniversityofLondonInstituteofEducationandotherswhocontributedto the seminar series on research evidence organized by the Evidence Net- work, funded by the UK Health Development Agency. Other colleagues with whom Helen has worked on systematic reviews including Carol Joughin, Gabrielle Laing, Stuart Logan, Sandra Dowling, Catherine Law, Jos Kleijnen, Patricia Lucas, Janis Baird, and David Fisher have helped her thinking while working on joint systematic reviews. Ruth Gilbert, Ian Roberts,andDennisCheekallrespondedquicklyandhelpfullytorequests