A flow diagram to facilitate selection of interventions and research for health care.

Update I Le point A flow diagram to facilitate selection of interventions and research for health care L. Irwig,1 M. Zwarenstein,2 A. Zwi,3 & 1. Chalmers4 Decisions about health care should be informed by systematic review of valid research evidence on the effects of interventions on health outcomes that matter. If systematic review suggests it is likely that a health care intervention does more good than harm in some settings, questions must be addressed about the local applicability of the intervention, its cost-effectiveness, and feasibility of implementation. If systematic review suggests that it is unlikely that an intervention does more good than harm in any setting, its use should be discouraged, while existing interventions are improved or alternative interventions developed. If it is uncer- tain whether an intervention does more good than harm, further analysis of existing data or new controlled trials are required. The article contains a flow diagram, which provides a structure for making such decisions. Introduction In poor and rich countries alike, planners and pro- viders of health services and health research face the challenge of deciding how to make effective use of limited or contracting resources. Promoting equi- table access to health services is an important health policy objective; however, better access will not improve health unless the care provided does more good than harm. By "good" we mean all posi- tive benefits of health care: an improvement in the quality of life or a prolongation of life. By "harm" we mean the unintended negative consequences to individuals affected by the form of health care in question, whether or not they were the principal focus of the intervention. This criterion of effectiveness - doing more good than harm - can be applied to health care interventions across the board. Examples include the Associate Professor, Department of Public Health and Commu- nity Medicine, A27, University of Sydney, New South Wales 2006, Australia. Requests for reprints should be sent to this author. 2 Head, Health Systems Division, Centre for Epidemiological Re- search in Southern Africa, Medical Research Council, Tygerberg, South Africa; and Department of Public Health and Primary Care, University of Oxford, Radcliffe Infirmary, Oxford, England. 3 Senior Lecturer, Health Economics and Financing Programme, Health Policy Unit, London School of Hygiene and Tropical Medi- cine, London, England. 4 Director, UK Cochrane Centre, NHS Research and Development Programme, Oxford, England. Reprint No. 5818 following: curative medical and surgical treatments; the care offered by nursing and other professionals in health facilities or domiciliary settings; preventive immunizations; nutrient supplements; care offered by volunteers in the community; health education in schools; and regulatory or fiscal interventions to en- courage health-promoting behaviour. Policy-making in the health sector is complex and takes place within a given social, economic, and political context. Decision-makers may take differ- ent views as to whether a particular service should be provided by the public or private sector and whether it should be financed using public or private re- sources (1). However, it is becoming increasingly accepted that decisions about health service provi- sion should be informed by valid evidence of the effects of health care interventions, along with an appreciation of the costs of providing such interven- tions, the competing priorities, and the population's values and preferences. Rationally, this process of decision-making in health care begins with efforts to identify health problems of high priority; i.e. prob- lems that are common, serious, resource-consuming, of community concern, and potentially amenable to interventions (2-4). Decision-makers need reliable evidence to assess the relative merits of alternative strategies for preventing, treating, and researching the priority problems identified. In recent years, the importance has increasingly been recognized of improving the evidence base for guiding such decision-making by conducting system- atic reviews of research evidence (5). This is appar- Bulletin of the World Health Organization, 1998, 76 (1): 17-24 © World Health Organization 1998 17 L. Irwig et al. ent among those using health care (6) as well as those providing it (7), health research funding bodies (8), those assessing the academic performance of health research institutions (9), government advisory bod- ies (10), those promoting policy changes (11), and organizations responsible for funding and purchas- ing health services (12). The best evidence for the effects of an interven- tion comes from randomized comparisons of indi- viduals or groups since these avoid the selection biases inherent in observational studies (13). Randomized controlled trials have shown, for exam- ple, that aspirin reduces mortality after myocardial infarction by about 23% (14) and that improved treatment of sexually transmitted diseases in rural Africa reduces spread of human immunodeficiency virus (HIV) by up to 40% (15). Analysis of the re- sults of all well-controlled trials in systematic reviews provides powerful evidence for the effects of inter- ventions (16-18). Access to the results of systematic reviews, and to information about reviews being prepared, has now been facilitated through publica- tion of sources such as The Cochrane Database of Systematic Reviews, The York Database ofAbstracts of Reviews of Effectiveness and the journal, Evidence-based medicine (19-21). Against this evolving background, we present in this article a flow diagram as an aid to selecting health care interventions and identifying research needs. The diagram should be helpful as an aid in system-wide planning, in assessing interventions for a single group of health problems, in evaluating a single intervention, and in prioritizing research on health care interventions. In the flow diagram (Fig. 1), the crucial step, shown in double-edged boxes, is determining whether a systematic review shows the following: - it is likely that an intervention does more good than harm in some settings; - it is unlikely that an intervention does more good than harm in any setting; or - it is uncertain whether an intervention does more good than harm. When it is likely that an intervention does more good than harm in some settings An intervention does more good than harm when compared with doing nothing or with an alternative intervention (and after taking into account the confi- dence intervals around the estimates of effect)- the beneficial effects of the intervention outweigh the harmful effects in some settings, populations, or pa- tient groups (22). Examples of such interventions include use of low-dose aspirin for reducing the bur- den of nonfatal myocardial infarction and stroke (23); vitamin A supplementation for reducing child- hood mortality in developing countries (24); and home visiting to prevent childhood injuries in Eu- rope and North America (25). Applicability If a systematic review shows that an intervention is likely to do more good than harm in some settings, populations or patient groups, its applicability to the particular setting of interest needs to be examined. There is sometimes reluctance to accept that the re- sults of research carried out in other places and at other times are applicable to local circumstances. However, the systematic review may include studies of the intervention that have been conducted in a range of settings, suggesting wide applicability. The use of antenatal corticosteroids in women expected to deliver preterm provides such an example (26). Reluctance to accept strong evidence of effective- ness from research performed elsewhere is difficult to justify unless it can be proved that the effect in local populations is likely to be different from that in the trials included in the review (27). For most drug interventions, this is unlikely to be the case. For social interventions, however, the effectiveness may vary substantially from one group or area to another, depending on factors such as literacy, income, cul- tural values and access to media and health services. For example, home visiting to prevent childhood in- juries, which has been shown to be effective in a systematic review of controlled trials conducted in the USA, Canada, the United Kingdom and Ireland (25), may not be applicable in low- and middle- income countries. Local populations or subgroups may differ in the extent to which the benefit of the intervention outweighs harm (28-30). For example, a reduction in embolic stroke through use of anticoagulants in individuals with atrial fibrillation carries a large benefit for those at high risk of stroke; however, the small benefit to those at low risk is outweighed by the harmful effects of anticoagulation, such as intracranial haemorrhage (28). Also, cholesterol- lowering drugs may be beneficial to those at particu- larly high risk of coronary heart disease, but for those at low risk the harm may outweigh the benefit (31). On occasions, there will be insufficient evidence to decide whether the intervention will do more good than harm in local populations, and extra infor- WHO Bulletin OMS. Vol 76 199818 A flow diagram for health care decisions Fig. 1. Flow diagram for selecting health care interventions and identifying research needs. Is there an adequate systematic review NO of the effects of all altemative interventions to prevent or treat the health problem? * Prepare a relevant YES systematic review What can be conctuded from the systematic review? -* UNCERTAIN Is it plausible that *_ the intervention does more good than harm (in general, or in subgroups)? - NO NO . * Discourage use of intervention K NO * * Improve existing interventions * Develop new interventions Gen * RUi to S Further analysis of existing data ierate valid new evidence by: Indertalsng new randomized trials lestricting availability of intervention randomized trials * NO/UNCERTAIN * Research and redesign method of implementation mation will be required about the frequency and severity of side-effects, level of risk of the outcome the intervention is intended to prevent, and popula- tion preferences or values (28). Further analysis of pooled data from all the valid studies may also be needed to assess effects in relevant subgroups (23). 0 Cost-effectiveness If interventions have been shown to be effective, either overall or in subgroups of a population, their cost-effectiveness must be considered. Different methods of organizing interventions vary in their WHO Bulletin OMS. Vol 76 1998 I 19 L. Irwig et al. labour and time costs; also, factors such as the scale of provision, and whether or not they are provided alongside other interventions and health service ac- tivities, may influence the cost-effectiveness (32). Some interventions remain cost-effective across a wide cost range, e.g. antenatal corticosteroids (33), and this evidence needs to be translated into action. Others may be considered sufficiently cost-effective in some circumstances, but not in others. For exam- ple, mammographic screening of women over 50 years of age is considered sufficiently cost-effective in many wealthy countries but not in poorer coun- tries where breast cancer incidence is lower, and there are more pressing demands on scarce re- sources (34). Implementation and ongoing audit If it has been shown that an intervention is likely to do more good than harm and that it is likely to be cost-effective if implemented, policies or guidelines based on the evidence should be formulated (35) and intervention plans developed and piloted, taking account of local circumstances and views. If pilot implementation establishes that the proposed in- tervention is logistically and politically feasible in the local setting and is acceptable to users, it should be introduced more widely, together with an ongoing audit to ensure that those who can benefit from the intervention receive it and that its quality is ad- equate (36-39). On the other hand, if the pilot im- plementation programmes fail, further research may be needed to assess how to improve methods of implementation. When it is unlikely that an intervention does more good than harm in any setting Systematic review may show that it is unlikely that an intervention does more good than harm in any setting, or that it is unlikely to be superior to existing alternative interventions. Also, although an inter- vention has been shown to have the potential to do more good than harm, it may be either inapplicable or not cost-effective in the population under consid- eration. In any of these circumstances, use of the intervention is an undesirable drain on limited re- sources and such use should be discouraged. If an intervention with any of these characteris- tics is not yet in widespread use, it may be possible to prevent its further dissemination; however, the inter- vention may already be in wide use, for example electronic fetal heart rate monitoring as a screening test in the absence of facilities to confirm a diagnosis of suspected fetal hypoxia (40). Persuading people to abandon an intervention is often not easy, particu- larly if no acceptable alternative is readily available. Levers to discourage use of ineffective interventions have been developed most effectively in pharmaceu- tical prescribing, using reimbursement schedules such as those in Australia (41). In the United Kingdom, purchasing authorities are often in a position to specify the types of services they are willing to fund, although this may at times conflict with advice offered by local health professionals (42). A variety of other techniques have also been shown to change practices, such as clearly stated and widely disseminated clinical guidelines, systems of audit, and continuing professional development activities (37, 43-46). Formal demonstration of the inadequacies of an intervention should also prompt further research to improve existing interventions or basic research to develop new alternatives. When it is uncertain whether an intervention does more good than harm Systematic reviews of research findings often reveal uncertainty about the effects of health care inter- ventions. Even if there are high quality trials of interventions in which biases have been well con- trolled, the confidence intervals around the esti- mates of the effects may be wide, making it impossible to decide whether the benefits outweigh the harm. Also, there may be doubts about the appli- cability of the available evidence to local populations or subgroups. Often, there may be no randomized trials of adequate quality or they may not have addressed important outcomes. In the first of these instances, a decision needs to be reached about whether such trials are or are not feasible (47). If they are not feasible, decisions have to be based on the most valid available evidence identified by systematic review of observational studies (47, 48). If no trials have ad- dressed important outcomes, causal pathways can be constructed, mapping out the necessary sequence of events between the intervention and final outcome (49, 50); the evidence can then be assessed for each step along the pathway. There may be randomized trial evidence for the effect of the intervention on an intermediate measure, and decisions can then some- times be made on the grounds that the evidence on the effect of the intermediate measure (e.g. stopping smoking) on final outcome (e.g. mortality) is well established. WHO Bulletin OMS. Vol 76 199820 A flow diagram for health care decisions Wide, uncontrolled application of new interven- tions of uncertain benefit is rarely justified, even though policy-makers, planners, and providers may feel under pressure to make the intervention avail- able. In the past, implementation without strong evi- dence has led to loss of life; for example, the widespread use of anti-arrhythmic drugs following myocardial infarction before the adverse effects on mortality outcomes had been demonstrated in randomized trials (51). The general implementation of interventions of unproven value consumes resources that could po- tentially be deployed more effectively elsewhere. This does not mean that such new interventions are not desirable, but it does call for their careful assess- ment. For example, systematic reviews of controlled trials of routine iron supplementation in pregnancy (52) and of routine antimalarial chemoprophylaxis during pregnancy (53) have so far been unable to detect important beneficial effects for either inter- vention. However, estimates for the effects of these policies on substantive outcomes are very imprecise - and certainly compatible with important benefi- cial effects. Furthermore, most of the trials of routine iron supplementation have been carried out in high- income countries but the benefits may be larger in low- and middle-income countries, where iron defi- ciency is more common and severe. If it is plausible that the intervention being con- sidered could do more good than harm, uncertainties such as those illustrated above might be addressed using data available from systematic reviews to ex- plore effects on subgroups. For routine iron supple- mentation during pregnancy, for example, it might be possible to obtain data for individual persons from all the relevant trials and conduct further analy- ses, stratified by the level of haemoglobin at the time of randomization. If this approach is not helpful, questions about the value of the intervention are best answered by conducting more randomized trials. Where an intervention is not already in wide use, its availability should be restricted to controlled trials to assess its effects, particularly if it is very costly (54). For example, until a randomized trial had yielded strong evidence that mortality could be re- duced without a reciprocal increase in devastating morbidity among survivors, use of extra-corporeal membrane oxygenation for severely asphyxiated neonates was initially available in the United King- dom only as part of a multicentre randomized trial (55). The evidence generated by new trials should be used to update existing systematic reviews or to ini- tiate new ones. The systematic review-research cycle then continues until there is sufficient evidence upon which to base decisions for health care. For example, a systematic review, recently updated in the light of evidence generated by a new trial, has made it clear that magnesium sulfate should now be regarded as the drug of choice for treating eclampsia (56). Conclusions Decisions on health care are complex and subject to many influences. Recognizing their complexity, how- ever, should not deter effects to encourage a more rational basis for decision-making, supported by evi- dence for the effectiveness, applicability to specific population groups, and cost-effectiveness of the pro- posed interventions. Clinicians, planners, managers, politicians and communities should be aware of the findings of sys- tematic reviews so that these are taken into account in the debates and negotiations surrounding the in- troduction, expansion, or reduction of health care interventions. Systematic reviews are becoming available in electronic format, enabling those con- cerned with health to identify regularly updated evi- dence that is relevant to their decision-making. The results of the few hundred systematic re- views that have already been carried out pose a substantial challenge to those responsible for com- missioning, managing, and providing health services: health authorities and decision-makers are failing to provide forms of care that are known to be effective, while acquiescing in the provision of other interven- tions that are very unlikely to improve health or are inferior to alternatives. Nevertheless, it is important to note that there are still large numbers of health care interventions whose benefit appears uncertain, even though they have been the subject of randomized trials. This information gap offers a sub- stantial agenda for systematic reviews. With support from health planning and research funding bodies, this challenge is now being tackled in a number of countries (19), but much more could be done. We hope that the approach outlined in the flow diagram we have presented will encourage the more rapid introduction of effective interventions, dis- courage the introduction or promote the withdrawal of ineffective interventions, and provide a frame- work for identifying interventions that should be investigated using randomized trials so that their effects can be assessed. Acknowledgements We thank Debbie Bradshaw, Ruari Brugha, Jonathan Craig, Lelia Duley, Paul Glasziou, Ronald Ingle, Jonny Myers, Olive Shisana, Chris Silagy, Bernie Towler and WHO Bulletin OMS. Vol 76 1998 21 L. Irwig et al. Jimmy Volmink for comments on earlier drafts of this pa- per, and Miranda Cheung and Lizi Holmes for help in preparing the paper and the flow diagram. The Health Economics and Financing Programme is funded by the British Overseas Development Administration, but the views in this article do not necessarily represent those of this agency. Resume Un ordinogramme pour faciliter le choix des interventions et des recherches a entreprendre en matiere de prestations de sante 11 importe que les d6cisions a prendre dans le domaine des prestations de sante reposent sur un examen syst6matique des conclusions de recherches sur les effets de diff6rentes interven- tions dans des secteurs importants. Si cet examen systematique donne a penser qu'une intervention a des chances de faire davantage de bien que de mal dans certains contextes, il faut chercher a d6terminer si cette intervention peut etre appliqu6e au niveau local, si elle offre un bon rapport coOt/ efficacit6 et comment elle peut etre mise en oeuvre. S'il ressort de l'examen syst6matique qu'une inter- vention risque de faire davantage de mal que de bien dans quelque contexte que ce soit, il faut en d6courager l'utilisation et am6liorer les interven- tions existantes ou mettre au point d'autres inter- ventions. S'il est difficile enfin de savoir si une intervention fera davantage de bien que de mal, il conviendra de proc6der a une analyse plus d6taill6e des donnees existantes ou a de nouveaux essais contr6les. L'ordinogramme (voir p. 24) fournit un schema pour la prise de ces d6cisions. References 1. Wait G. Health policy: an introduction to process and power. London, Zed Books, 1994: 3. 2. World Bank. Investing in health: world development report, 1993. New York, Oxford University Press, 1993. 3. Evidence-Based Care Resource Group. Evidence- based care: 1. Setting priorities: how important is this problem? Canadian Medical Association Journal, 1994, 150: 1249-1254. 4. Schaapveld K, van de Water HPA, Bergsman EW. Setting priorities in prevention. Health policy, 1992, 20: 277-287. 5. Chalmers I, Altman DG. Systematic reviews. Lon- don, British Medical Association, 1995. 6. Oliver SR. How can health service users contribute to the NHS research and development programme? Brit- ish medical journal, 1995, 310: 1318-1320. 7. Smith R. The scientific basis of health services. Needs a better balance between original and evalua- tive research. British medicaljoumal, 1995, 311: 961- 962. 8. Research and development in the new NHS: functions and responsibilities. Leeds, England, NHS Executive, November 1994. 9. Criteria for Assessment: 1996 research assessment exercise. Bristol, U.K. Higher Education Funding Councils. RAE 96 3/95. November 1995. 10. U.S. Congress, Office of Technology Assessment. Identifying health technologies that work: searching for evidence, OTA-H-608. Washington, DC, U.S. Government Printing Office, September 1994. 11. Mackenbach JP. Tackling inequalities in health. Great need for evidence based interventions. British medical journal, 1995, 310: 1152-1153. 12. Eddy DM. Rationing resources while improving qual- ity. How to get more for less. Journal of the American Medical Association, 1994, 272: 817-824. 13. Guyatt GH, Sackett DL, Cook DJ. Users' guides to the medical literature. II. How to use an article about therapy or prevention. A. Are the results of the study valid? Journal of the American Medical Association, 1993, 270: 2598-2601. 14. ISIS-2 (Second International Study of Infarct Sur- vival) Collaborative Group. Randomized trial of in- travenous streptokinase, oral aspirin, both, or neither among 17, 187 cases of suspected acute myocardial infarction: ISIS-2. Lancet, 1988, 2: 349-360. 15. Grosskurth H et al. Impact of improved treatment of sexually transmitted diseases on HIV infection in rural Tanzania: randomised controlled trial. Lancet, 1995, 346: 530-536. 16. Chalmers I, Haynes B. Reporting, updating, and cor- recting systematic reviews of the effects of health care. British medical joumal, 1994, 309: 862-865. 17. Dickersin K, Scherer R, Lefebvre C. Identifying rel- evant studies for systematic reviews. British medical journal, 1994, 309: 1286-1291. 18. Oxman AD, Cook DJ, Guyatt GH. Users' guides to the medical literature. VI. How to use an overview. Journal of the American Medical Association, 1994, 272: 1367-1371. 19. The Cochrane Library. The Cochrane Collaboration, issue 1. Oxford, Update Software, 1997. 20. NHS Centre for Reviews and Dissemination, Uni- versity of York. Database of abstracts and reviews of effectiveness (DARE) website http://www.york.ac.uk/ inst/crd/info.htm. 21. Evidence-based medicine. London, BMJ Publishing Group. 22. Guyatt GH et al. Users' guides to the medical litera- ture. IX. A method for grading health care recommen- dations. Journal of the American Medical Association, 1995, 274: 1800-1804. 23. Antiplatelet Trialists' Collaboration. Collaborative overview of randomized trials of antiplatelet therapy - I: prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients. British medical journal, 1994, 308: 81-106. 22 WHO Bulletin OMS. Vol 76 1998 A flow diagram for health care decisions 24. Glasziou PP, Mackerras DEM. Vitamin A supple- mentation in infectious diseases: a meta-analysis. British medical joumal, 1993, 306: 366-370. 25. Roberts I, Kramer MS, Suissa S. Does home visiting prevent childhood injury? A systematic review of randomized controlled trials. British medical journal, 1996, 312: 29-33. 26. Crowley P. Corticosteroids prior to preterm delivery. In: Neilson JP et al., eds. Pregnancy and childbirth module of the Cochrane Database of Systematic Re- views (updated 5 December 1996). Available in: The Cochrane Library. The Cochrane Collaboration, is- sue 1. Oxford, Update Software, 1997. 27. Guyatt GH, Sackett DL, Cook DJ. Users' guides to the medical literature. II. How to use an article about therapy or prevention. B. What were the results and will they help me in caring for my patients? Journal of American Medical Association, 1994, 271: 59-63. 28. Glasziou PP, Irwig LM. An evidence-based approach to individualising treatment. British medical journal, 1995, 311:1356-1359. 29. Rothwell PM. Can overall results of clinical trials be applied to all patients? Lancet, 1995, 345: 1616- 1619. 30. Boissel JP et al. An effect model for the assessment of drug benefit: example of antiarrhythmic drugs in postmyocardial infarction patients. Joumal of cardio- vascular pharmacology, 1994, 22: 356-363. 31. Davey Smith G, Song F, Sheldon TA. Cholesterol lowering and mortality: the importance of considering initial level of risk. British medical journal, 1993, 306: 1367-1373. 32. Phillips M. Setting global priorities for strategies to control diarrhoeal disease: the contribution of cost- effectiveness analysis. In: Mills A, Lee K, eds. Health economic research in developing countries. Oxford, Oxford University Press, 1993: 214-249. 33. Mugford M, Piercy J, Chalmers I. Cost implications of different approaches to the prevention of respira- tory distress syndrome. Archives of disease in child- hood, 1991, 66: 757-764. 34. van Ineveld BM et al. How cost-effective is breast cancer screening in different EC countries? European journal of cancer, 1993, 29A: 1663-1668. 35. Evidence-based Care Resources Group. Evidence- based care: 2. Setting guidelines: How should we manage this problem? Canadian Medical Association joumal, 1994, 150: 1417-1423. 36. Evidence-based Care Resource Group. Evidence- based care: 3. Measuring performance: How are we managing this problem? Canadian Medical Associa- tion joumal, 1994, 150: 1575-1579. 37. Evidence-based Care Resource Group. Evidence- based care: 4. Improving performance: How can we improve the way we manage this problem? Canadian Medical Association joumal, 1994, 150: 1793-1796. 38. Mant J, Hicks N. Detecting differences in quality of care: the sensitivity of measures of process and out- come in treating acute myocardial infarction. British medical journal, 1995, 311: 793-796. 39. Tarnow-Mordi W. Comparing performance of hospi- tals. Lancet, 1994, 343: 1162-1163. 40. Neilson JP. EFM alone vs intermittent auscultation in labour. In: Keirse MJNC et al., eds. Pregnancy and childbirth module. In: The Cochrane Pregnancy and Childbirth Database. The Cochrane Collaboration; Is- sue 2. Oxford, Update Software, 1995. 41. Guidelines for the pharmaceutical industry on prepa- ration of submissions to the Pharmaceutical Benefits Advisory Committee. Canberra, Australia, Common- wealth Department of Human Services, November 1995. 42. Littlejohns P, Dumelow C, Griffiths S. Knowledge- based commissioning: can a national clinical effec- tiveness policy be compatible with seeking local professional advice? Journal of health services re- search and policy, 1996, 1: 28-34. 43. Grimshaw JM et al. Development and implementa- tion of clinical practice guidelines. Quality in health care, 1995, 4: 55-56. 44. Haines A, Jones R. Implementing findings of re- search. British medical journal, 1994, 308: 1488- 1492. 45. Oxman AD et al. No magic bullets: a systematic review of 102 trials of interventions to improve profes- sional practice. Canadian MedicalAssociation journal, 1995, 153:1423-1431. 46. Johnston ME et al. Effects of computer-based clini- cal decision support systems on clinician performance and patient outcome. A critical appraisal of research. Annals of intemal medicine, 1994, 120: 135-142. 47. Liddle J, Williamson M, Irwig L. Improving health care and outcomes. Method for evaluating research guideline evidence. Sydney, NSW Health Depart- ment, December 1996. 48. Cochrane Collaboration of effective professional practice. Available in: The Cochrane Library. Oxford, Update Software, 1997. 49. Woolf SH et al. Developing evidence-based clinical practice guidelines: lessons learned by the U.S. Pre- ventive Services Task Force. Annual review of public health, 1996, 17: 51 1-538. 50. U.S. Preventive Services Task Force. Screening for adolescent idiopathic scoliosis. Journal of the American Medical Association, 1993, 269: 2664- 2666. 51. Echt DS et al. Mortality and morbidity in patients receiving encainide, flecainide, or placebo. The car- diac arrhythmia suppression trial. New England jour- nal of medicine, 1991, 324: 781-788. 52. Mahomed K. Routine iron supplementation in preg- nancy. In: Neilson JP et al., eds. Pregnancy and childbirth module. In: The Cochrane Database of Sys- tematic Reviews (updated 20 November 1996). Avail- able in: The Cochrane Library. The Cochrane Collaboration; Issue 3. Oxford, Update Software, 1997. 53. Gulmezoglu AM, Garner P. Interventions to prevent malaria during pregnancy in endemic malarious ar- eas. In: Garner P et al., eds. Infectious diseases mod- ule of the Cochrane Database of Systematic Reviews (updated 27 February 1997). Available in: The Cochrane Library. The Cochrane Collaboration, Issue 3. Oxford, Update Software, 1997. WHO Bulletin OMS. Vol 76 1998 23 L. Irwig et al. 54. Glasziou PP. Support for trials of promising medica- tions through the Pharmaceutical Benefits Scheme. A proposal for a new authority category. Medicaljoumal of Australia, 1995, 162: 33-36. 55. UK Collaborative ECMO Group. UK collabora- tive randomised trial of neonatal extracorporeal membrane oxygenation. Lancet, 1996, 348: 75- 82. 56. Duley L, Henderson-Smart D. Magnesium sulphate vs diazepam for eclampsia; and magnesium sulphate vs phenytoin for eclampsia. In: Neilson JP et al., eds. Pregnancy and childbirth module of the Cochrane Database of Systematic Reviews, (updated 5 Decem- ber 1996). Available in: The Cochrane Library. The Cochrane Collaboration, Issue 3. Oxford, Update Software, 1997. Fig. 1. Ordinogramme pour faciliter le choix des interventions et des recherches a entreprendre en matiere de prestations de sante Les effets de toutes les interventfons susceptfbles de prevenir ou de resoudre le NON problerne de sante font-ls l'objet d'un examen systAmatique adiquat? * Preparer un examen OUI systematique appropdi Que peut-on conclure de 1'examen systematique? PROBABLE qu'une intervention PEU PROBABLE qu'une UNCERTAIN si unefora davantge do bion quo intervention fera davantage do intervention fers davantage dede mat dans certainos bien que de mat dans quelquo bien que de mat contoxts contexft que ce soit Lintervonton fera-tele davantagp Est-il plausible que l'intervention de bien que de mal - INCERTAIN tfera davantage de bion que appltquoe A des p ions locales de mal (on general, ou au(on ginral, ou au sein de sein de sous-groupes ? sous-groupes)? NON NON OUI OUI * Decourager le recours a * Entreprendre une analyse plus Celte intervention efficace l'intervention poussee des donnees oxistantes et appHcable est eole d'un -* NON - * Ameliorer les intervenffons bon rapport coOt/efficacit ? existantes Obtenir de nouveles preuves valides:* Mettre au point de nouvelles oen efctuant de nouveaux essaisinterventions randoris6s S en limitant l'intervention aux essaisOUI randomises Ceote interventIon d'un bon rapport cot/efficaciti peut-alle Atre mise. NON/INCERTAIN en cauvre ? Passer A la mise en ceuvre en la soumottant 0 ProcAder a des recherches et revoir les methodes a un contr6lo continu d'executon 24 WHO Bulletin OMS. Vol 76 1998