CHAPTER 1
The importance of clinical evidence
Combining the skill that an individual clinician has gained through experience and practice with the best available external clinical evidence is called evidence-based practice (EBP). It is also known as evidence-based medicine (EBM) and evidence-based healthcare (EBHC).
Definition
Clinical evidence is the information from research studies that helps decide the value of screening programmes, diagnostic tests, management plans and treatments.
Evidence-based practice
Integrating evidence-based medicine with patients' own values and preferences is the basis of good clinical decision-making.
Expert opinion or evidence-based practice?
In the past, clinicians relied on their own experience and the opinions of experts to make clinical management decisions. While these judgements were often correct, the careful application of research evidence has been found to improve patient care, making it safer and more effective.
Grounding care in clinical evidence has now become routine. Concepts like P values, meta-analysis and randomized controlled trials (RCTs) are such an important part of decision-making that clinicians need to understand them.
The types of clinical evidence
Researchers design a study to answer a specific clinical question, for instance "Is a steroid injection likely to be more effective than physiotherapy for tennis elbow?".
Early in their planning, researchers need to check whether their question has been answered already. If it has, there is no point in wasting resources on replicating the work of previous researchers.
A study is therefore designed to answer either a new question, one that hasn't been adequately answered before, or to confirm an answer that is unexpected or is particularly important.
The research method used depends on the question. The different methods are described in more detail in the following chapters, but here is a summary list of common ones.
Randomized controlled trials(Chapter 8) are a type of prospective study, i.e. one whose patients are identified and then followed up into the future. In these studies the risk of bias (see Chapter 5) is minimized by randomly allocating the subjects to one of two or more treatments; it could be used to compare the effectiveness of a new treatment for temporal lobe epilepsy with the existing best treatment, for example. The P value (Chapter 7) gives information on how likely it is that any difference in outcome was due to chance.
Cohort studies(Chapter 9) are another type of prospective research. Here, researchers follow groups of patients who are alike in many ways but differ in one or more particular characteristics. They may follow them for some years to look at outcomes, i.e. what happens to them. For example, researchers may study a group of patients to compare the effect of high and low levels of computer use on long-term visual acuity. A statistic called a risk ratio is used to compare the outcomes of these two 'cohorts' of patients.
A case–control study(Chapter 10) is an example of a retrospective design, i.e. one that looks back over time: the outcome of interest has already happened and the researchers want to know what factors may have influenced it. These studies compare a group of patients having a condition, the 'cases', with a group without it, known as the 'controls'. This method could be used to assess whether hospital patients who have developed a deep vein thrombosis (DVT) during their stay are more (or less) likely to have taken aspirin in the last month than similar, 'matched' patients without a DVT. The odds ratio is a statistic that compares these groups.
Research on diagnostic tests(Chapter 11) typically compares a test with a 'gold standard' test, for example how a new blood test for coeliac disease performs compared with the results of small bowel biopsies in the same patients. The resulting specificity and sensitivity values show how often the test is negative in patients without the condition, and positive in those with the condition, respectively.
Qualitative research(Chapter 12) finds out about what people are thinking and why. It could be used to find out why patients do (or don't) see their GPs if they have a productive cough. Researchers find the key themes and can develop a hypothesis that fits them together.
Health economic evidence(Chapter 15) typically examines the costs and benefits of different healthcare interventions, treatments or policies. When applying economic tools to the world of healthcare, researchers may try to answer broad questions on health policy. They may also look at a more specific question; for instance, one that asks which treatment is most cost-effective for control of hypertension, when compared to an alternative. Our financial resources are limited, so health economists work out how to optimize the provision of healthcare.
Can one paper do it all?
Very occasionally a single research paper causes a big change in practice. However, more often individual pieces of research don't give the whole answer, but are small building blocks to it.
One way for a clinician to plan an evidence-based approach to a clinical problem is to appraise and synthesize all the relevant papers. This is very time-consuming, so many rely on the work of others that have done that for them.
A systematic review(Chapter 13) is a literature review that identifies, assesses and synthesizes all the relevant research evidence. An author might review all the papers on the non-surgical management of appendicitis, for instance.
A meta-analysis(Chapter 13) combines the numerical data from different studies. Pooling the results of different studies on survival after cardiac stents, for example, might give a more reliable idea of the size of the effect than any of the individual underlying studies.
Clinical guidelines(Chapter 14) are recommendations that authors have made as a result of their own systematic review. An example is a guideline on the management of type 2 diabetes by diet alone.
The five 'A's
The key steps in using clinical evidence are as follows.
• Ask: define the clinical question, for instance, whether or not a specific treatment is better than placebo for a particular condition.
• Acquire evidence: make a systematic retrieval of the available literature and data.
• Appraise: critically appraise that evidence to assess its validity (how close it is to the truth) and clinical applicability (how useful it is likely to be for our own patients).
• Apply the results: make changes in clinical practice as a result of our appraisal of the evidence.
• Assess the outcome: evaluate the effect that those changes have on patient care.
CHAPTER 2
Asking the right questions
Before we can make any progress with finding the most appropriate clinical evidence and then working out if this evidence is valid and useful to us, we need to frame the question that we want to answer. In day-to-day clinical life we have lots of clinical questions to answer, so it also helps to know how to prioritize them.
See Chapter 18 for a tool that will help us to design our own clinical questions.
Definition
Clinical questions can be divided into two categories:
• foreground questions – these are about decisions that need to be made regarding a patient's management;
• background questions – these look for general knowledge on a condition or an aspect of health status.
How easy is this?
Asking a question that is precise enough can be surprisingly difficult, but it is one of the most important steps in a critical appraisal. Once we have drafted our question it may need further development to ensure it meets our objectives. We therefore need to be prepared to return to it and, if needed, start our literature searches again.
Foreground questions
Foreground questions usually relate to prevention, diagnosis, treatment (or other clinical management), risk (including causation and risk of harm), or how people think.
These questions usually have four components:
• the Problem, Patient or Population (in this context the word 'population' means all patients with the same problem);
• Intervention, which may be a diagnostic test, a treatment or other management strategy;
• Comparator, i.e. the alternative to the intervention being considered;
• Outcome, which is the end result of interest.
These are known by the acronym PICO.
Not all foreground research questions will have a comparator, in which case the format becomes PIO.
Some questions about how people think fall into the PIO category.
Some questions need a fifth component: Timeframe, the period of time during which we want to study the impact of the intervention.
Defining the problem
When deciding which patient, problem, or population our question is to be based on, it may help to think about:
• How would we describe the patient to a colleague? "A patient with a central abdominal mass", for example.
• How might we describe a group of patients with a similar problem? "Patients with rectal bleeding".
• What are the important characteristics of these patients? "Patients over 80 years of age with weight loss and change in bowel habit".
Identifying the intervention
In the second PICO step outlined above, we need to be clear about which intervention we are interested in. While this may be a test or treatment, it could instead be about screening for disease, lifestyle advice or any other aspect of patient management. We do need to be quite specific about this. For example, are we interested in patients who have had screening for bowel cancer, or patients who have been offered that screening?
The 'I' of PICO may also stand for Indicator, for example patient demographics.
Comparison
If we are to have a comparator, defining it needs careful thought. Is it no intervention? A different treatment? Or a diagnostic test, and if so, what exactly? For instance, in our example of patients with possible cancer of the colon, do we want to compare virtual colonoscopy with conventional colonoscopy, or with barium enema? If the latter, with or without flexible sigmoidoscopy?
Outcome
This states the result that we are interested in: what are we trying to accomplish, measure or improve? Examples include the risk of illness, the rate of complications and the accuracy of a diagnostic test. Outcomes may be disease-orientated or patient-orientated.
Again, it is important to be specific because we will then get more relevant search results when we look for the clinical evidence that will help us to answer the question (see Chapter 3). In a question comparing the effectiveness of two different methods of screening for bowel cancer, are we interested in pick-up rates, false positives and negatives, or in preventing deaths?
Time
We may also need to state the timeframe, for example how long it takes for the intervention to achieve the outcome. If we want to know about the morbidity caused by conventional colonoscopy, are we only interested in symptoms experienced while patients are still in hospital, or do we want to include what happens to them after discharge as well?
Background questions
These are about the general knowledge relating to a diagnostic test, disease process, intervention or thinking. Background questions have a question root (what, who, when, where, why or how?), then a verb followed by the test, disease, treatment or intervention.
Prioritizing the questions
Clinical questions arise out of our day-to-day clinical practice, but there are almost always more questions than we have time available to answer. We can prioritize them based on the following framework.
• How important is the question to our patient?
• How important or serious is the outcome?
• How likely is this scenario to recur?
• How large is the gap in our knowledge?
• How much time do we have in which to answer the question?
• How easy will it be to find the answer?
Watch out for ...
Time spent on developing our clinical question is more than made up for in the time we save by not having to peruse irrelevant literature. If the question isn't working well, it is best to review it and, if necessary, change it.
CHAPTER 3
Looking for evidence
The last chapter explained how we can formulate specific clinical questions that are relevant to the patients we are treating and use these to fill gaps in our clinical knowledge. To be able to appraise papers for quality and to evaluate the applicability of results given, we need to be sure that that we have found the relevant literature in the first place. So, the next step is to search the literature to identify the relevant peer-reviewed papers that can help to answer our question.
How easy is this?
The internet makes it easy for us to access a mass of data. However, selecting out only the papers that are relevant to our question, and ensuring that we have not missed any important ones, is a task that takes practice.
Looking for guidelines
One starting point is to search for the current guidelines within our own organizations. Hospitals or other healthcare providers usually have their own, locally relevant guidelines.
National guidelines are likely to be more authoritative. For example, in the UK one source of guidelines is the Scottish Intercollegiate Guidelines Network (SIGN) at www.sign.ac.uk, another is the National Institute for Health and Care Excellence (NICE) at www.nice.org.uk, which publishes guidelines that are recognized internationally. For primary care practitioners, NICE provides guidelines on many topics at its Clinical Knowledge Summaries website, cks.nice.org.uk.
Professional bodies also produce high-quality clinical guidelines. Look for guidelines posted on a relevant specialist site, for example the Royal College of Obstetricians and Gynaecologists' website, www. rcog.org.uk/guidelines.
Finding systematic reviews
If we search and find that no suitable guidelines are available in our area of interest, the next stage is to look for systematic reviews.
The best source for these is the Cochrane library, www. thecochranelibrary.com. This online database has systematic reviews of primary research that are generally accepted as the highest standard in EBHC. They investigate the effects of interventions for prevention, treatment and rehabilitation, as well as assessing the accuracy of a diagnostic test.
Hunting for original research articles
If we cannot find any relevant systematic reviews, then we need to look for original papers. These are indexed in medical databases such as MEDLINE. MEDLINE is a large database of clinical papers, comprising over 20 million references to original articles published in about 5600 current biomedical journals from over 80 countries. We can search this database using PubMed, www.ncbi.nlm.nih.gov/ pubmed, which is freely accessible to all.
NICE gives links to high-quality authoritative evidence sources at www.evidence.nhs.uk. It covers health, social care and public health evidence, all designed to help healthcare professionals make better and faster evidence-based decisions.
Checking out the references given in recent guidelines, reviews and primary sources can also be valuable because the authors of these will have spent a lot of time searching the literature for their topic.
Planning a search
When we have clearly defined our question (see Chapter 2), we need to choose keywords which best represent the question we are trying to answer. This can be the most challenging part of doing a literature search.
In general, it is best to search using the medical database's keywords; for example, it may use Medical Subject Headings (MeSH). Being familiar with the hierarchy of headings used by MeSH can be helpful.
In the text that follows, we are using square brackets [ ] to highlight search words and phrases.
Note that a keyword may be a phrase like [Respiratory hypersensitivity], or a single word like [Asthma].
We need to choose the most specific terms relevant to our query.
