Clinical practice guidelines: on what evidence is our clinical practice based?

Many sources of evidence inform clinical practice, including research findings, patients' views, and clinical experience. This article describes recent progress in developing what will be the College's firstguideline, The Assessment and Management of Violence In Clinical Settings.The design of the programme acknowledges the diversity of evidence in this area, by drawing a distinction between different types of evidence. Evi dence is systematically identified and Its quality ap praised by the Work Group, through commissioning literature reviews.Initial recommendations will be based on experimental data. The methods used to consider non-experimental data, particularly expert opinion, will be described in more detail ina later article inthisseries. The

The Clinical Practice Guidelines Programme's aim Is to Influence routine health outcomes by supporting clinicians' decisions about patient care. A cornerstone of the programme Is the task of gathering the evidence on which clinical practice Is based, and bringing it to the attention of clinicians. Sources of evidence include the basic sciences, patient-centred research findings, the law, medical ethics, patients' views, and clinical judgement and experience. The topic for the first clinical practice guideline, The Assess ment and Management of Violence in the Clinical Setting, was selected following a survey of those Involved in mental health care (Palmer, 1995).

A diverse range of evidence supports clinical practice
There are many different types of evidence which Inform clinical practice. Evidence can be de scribed as either experimental or non-experimen tal. Experimental evidence may be of either a quantitative or qualitative nature. Where quanti tative data is sought, there is a burgeoning methodology of trail designs, all varying consider ably In their 'power', or the extent to which they are likely to minimise both the operation of systematic bias and the opportunity for chance findings. The use of a control group, and methods which randomly allocate participants to a treat ment group strengthen the trial design. A randomised controlled trial (RCT) combines both of these features. It is widely regarded as the 'gold standard' of quantitative research methods. However, the quality of a controlled experiment does not depend solely on Its design, but also on how it Is conducted. Important questions in appraising a controlled study include those relating to systematic bias in the match between control populations and the treatment group, in the care provided to each aside from the Inter vention being evaluated. In those lost to follow-up during the course of the experiment, and in the assessment of outcome. Methods of allocation are also Important, particularly the extent to which those who evaluate outcomes are unaware (blinded) of the process or intervention preceding it. Where a number of RCTs address a virtually identical hypothesis, the results can be pooled Into a meta-analysis. If this Is well designed, it provides still stronger evidence because it summarises findings from substantially larger populations.
Unfortunately, there are neither sufficient welldesigned controlled trials to provide a compre hensive evidence base for the increasingly com plex world of routine practice, nor is their design always appropriate for this purpose. For Instance, to test the hypothesis that restriction of the most acutely disturbed patients is associated with an Increased risk of harm to others, It would obviously be practically difficult, as well as unethical, to randomise patients at the time of admission (including those on a section of the Mental Health Act), between a treatment group which received nursing care In a ward environ ment and another nursed In a community setting. Other quantitative designs can be adopted, and although not as robust, the findings from non-RCT designs can be so dramatic that it does not make sense to ignore them.

Good clinical practice is not only a matter of what is done, but also of how and why it is done
Qualitative research. Qualitative research designs are concerned with making sense of Information, Psychiatrie Bulletin (1996), 2O, 363-366 Table 1. Criteria for evaluating the quality of a controlled study Is the study hypothesis clearly defined? Is the study population a representative one? Was the assignment of patients to treatment random ised? Were patients, practitioners, assessors blind to the experimental Intervention? Were the groups similar at the start of the trial? Aside from the experimental investigation, were the groups treated equally? Were all those who entered the trial accounted for at its conclusion? Was this in the groups to which they were originally allocated? Are all clinically important outcomes considered? Whose perspective do they reflect? Are the techniques used for data analysis appropriate? What is the size, and precision of the treatment effect? Do the likely benefits outweigh the harms and risks? Is the conclusion supported by the results? and as much with generating hypotheses as with answering them. As a result, quantitative and qualitative approaches are often complementary.

Methods include delphi-techniques, in-depth interviews, and preference surveys.
Non-experimental evidence. This includes the beliefs, convictions, and judgements of indivi duals or groups, including clinicians, patients, and expert committees.
Unfortunately, whether or not evidence con clusively supports one course of action rather than another usually is influenced by a range of considerations rather than just one. For in stance, an important question is not just whether the expected outcome is good, but whether it is good enough when the possible risks are taken into account. Even with sophis ticated appraisal methods, the distinction be tween experimental and non-experimental, qualitative and quantitative studies, fact and opinion is not always crystal clear. Employing advanced statistical analyses may not be suffi cient to support a confident conclusion from a study's findings. A degree of judgement is always involved, and there is no substitute for common sense.

The quantitities of evidence are enormous, and can be hard to track down
The sheer volume of supporting evidence also poses difficulties. For instance, the first issue of the British Journal of Psychiatry this year contained 20 sdentine papers, two editorials, one review article, five letters, and five book reviews, each on topics related to clinical practice. It also reproduced the index for the most recent issue of the American Journal of Psychiatry, which included approximately the same number of items. These are just two of at least 1700 scientific and medical journals listed on the Mediine database alone. Other databases, such as Embase or PsychLit, include a signifi cant number of additional journals published at weekly, monthly or quarterly intervals.
The inexperienced electronic database user can expect to find less than 20% of the Information relevant to her search task, and a more experienced searcher will approach 50%. Systematic search techniques, employing a comprehensive range of subject headings (called MeSH in Mediine) and using specialist operator functions appropriately across the range of relevant databases are recommended. When these are supplemented with efforts to identify high quality, unpublished material, as well as material printed in unlisted publications, the sensitivity and specificity of searching can be improved substantially (Dickersin et al 1994). Hand-searching journals is very time consum ing, but sometimes reveals material 'concealed' from a searcher by inappropriate indexing terms. No. of articles: ', 1-5; ". 6-10; '". 11-20; "". 21-40. RCTs, randomised controlled trials.

Collating the evidence
The Work Group has met regularly since its first meeting in November 1995. It is chaired by Professor John Wing, and includes members from a range of professional backgrounds, and a service user1. It is supported by the members of the Clinical Practice Guidelines Project Team, based in the College Research Unit2. The Work Group's first task was to 'scope' the clinical topic. It adopted a dÃ©finitionfor violence, 'physically destructive or damaging behaviour'; for the settings in which the guideline might be applied, 'settings in which clinicians provide care'; and a clearly defined clinical scenario, 'a person who is actively violent, for example attacking others, breaking objects, or on the brink of such behaviour'. The clinical skills and decisions demanded by this clinical situation were considered in a wide range of practice settings, together with the desired (and possible undesired) outcomes. The resources and struc tures which might influence clinical processes and promote better outcomes were also explored.
Four project modules have been developed which together outline the guideline's content. These are prediction, prevention, assessment, and intervention. Before commissioning litera ture reviews, the Work Group required an estimate of the quality and quantity of relevant data. An overview of published literature was commissioned, and conducted by systematically searching a single electronic database (Mediine).
Literature reviewers have been recruited from a wide range of multidisciplinary backgrounds. Training in the techniques of developing and conducting a systematic literature review, sys tematic search and quality appraisal skills are provided by the programme, through the Clinical Practice Guidelines Office Team3. So far, six reviews have been commissioned, across the prediction and intervention modules. These are short-and long-term prediction of violence (prediction module), psychological intervention, seclusion and physical restraint, the human environment, the physical environment, and pharmacological Intervention (intervention mod ule). The protocol for each of these reviews is developed in consultation with the Work Group membership.
3. The Clinical Practice Guidelines Office Team Includes Ms Claire Palmer, Ms Victoria Thomas, and Dr Sarah Marriott.  Oxman, 1994).

The tasks ahead
The Clinical Practice Guidelines Programme is now in its second year. As its experience broad ens, It Is refining the framework within which practice guidelines can be developed and put to use (Marriott & Lelliott, 1994). This need not apply only to national projects, such as the College's programme. Many local clinicians are also keen to develop clinical systems, as well as the evidence base for the care they routinely provide. The Clinical Practice Guidelines Office Team is now working on educational and training materials covering multldiscipllnary project de velopment, literature searching and appraisal skills. Finally, the development of the College's first guideline, Assessment and Management of Violence in Clinical Settings is well under way. It is likely that the systematic approaches to estab lishing the evidence base for practice outlined here will reveal areas In which the research findings are far from robust. This alone will be valuable and will contribute to defining a future research agenda.