Evidence Based Medicine
Search for the evidence
Sensitivity/Specificity or How do I get the Right Stuff and Avoid Getting the Wrong Stuff!
You have probably seen the terms sensitivity and specificity used in terms of test characteristics. In this context sensitivity is the percentage of patients with a known disease who test positive for the disease. Specificity is the percentage of patients without the disease who test negative. In searching, these two terms have similar meanings. Sensitivity is the likelihood of retrieving relevant items while specificity is the likelihood of excluding irrelevant items. For example if we are searching for articles on hypertension, a sensitive search would find all articles that mention the term hypertension while specific search would exclude those articles that mention hypertension, but hypertension is not the main focus of the article.
The ideal search finds the most relevant articles with the least amount of work. If you searched on the term hypertension in medline, you would quickly become overwhelmed with the number of ‘hits' or articles that mention hypertension. In this case you want a search with a higher degree of specificity and thus you would want to limit the search in some way to focus on some specific aspect of hypertension such as therapy with a particular antihypertensive. If you search for articles on a relative uncommon condition, you would want to do a search with high sensitivity because you don't want to miss potentially important articles. In this case you might not want to place limits on the search. Put another way if your search yields too many hits, you will want to increase the specificity of your search. If your search yields too few articles, you will want to increase the sensitivity of your search.
You can Increase specificity of your search by:
- Focusing or narrowing your question
- Limit your search by age of the patient (pediatric, elderly), type of article (randomized clinical trial, review, meta-analysis), language (English), type of publication (major clinical journals [AIM])
- Use the MESH heading instead of free text or textword search
- Use MESH subheadings
- Combine searches such as hypertension and diuretics or asthma and corticosteroids
You can increase the sensitivity of a search by:
- Making your question more broad (instead of looking at hypertension in just diabetic patients, look for hypertension irregardless of associated diseases)
- Don't employ the strategies listed above
- Expand the time frame of your search (in Medline you can search all the way back to 1966)
Levels of Evidence
Different types of evidence are better than others. A well done randomized clinical trial can provide better justification for instituting a therapy than a physician's experience with one or two patients. You will want to seek out those types of articles that provide the best evidence. Sackett et al have proposed a hierarchy of evidence as follows:
- N of 1 randomized trial
- Systematic Review of RCTs
- Single RCT
- Systematic Review of Observational Studies addressing patient-important outcomes
- Single observational studies
- Physiologic Studies
- Unsystematic clinical observations
The important point is that you will want to search for high quality studies. That usually means randomized clinical trials or systematic reviews. Well-done clinical guidelines may also provide you with a systematic review of the evidence.
Searching For Best Evidence
If you have access to PSHMC, there are excellent evidence-based sources.
- Best Evidence – is the electronic version of American College of Physicians (ACP) journal club. Articles of interest to internists are reviewed.
- Cochrane Library – Is a collection of meta-analyses on a variety of clinical issues
- Database of Abstracts of Reviews of Effectiveness (DARE) – is a collection of high quality systematic research reviews.
- Clinical Evidence – Clinical questions relevant to clinicians are posed and then the literature is reviewed for the best evidence.
- UpToDate – is an excellent electronic textbook. It is very clinically oriented and updated on a regular basis. Currently it is limited to Internal Medicine. It tends to be evidence-based; however, it does not present its methodology for searching or appraising the evidence. It can be used to obtain a clinical perspective and references for the original literature.
- OVID – is the product used to search various databases including the evidence-based sources listed above as well as medline.
Other Free Resources
- American College of Physicians (ACP) Journal Club Online - now available online for free. Searchable back to it's origins in 1991.
- NHS Centre for Reviews and Dissemination - produced by the University of York, England and the National Health Service (NHS) Centre for Reviews and Dissemination. This site allows you to search abstracts of Cochrane reviews, the Database of Abstracts of Reviews of Effectiveness (DARE), and a collection of health technology assessments.
- TRIP – is a collection of databases, mainly British, which are intended to improve the evidence-base of primary care
- University of Michigan Evidenced-based Pediatrics website - has a collection of CATs (Critically Appraised Topics). Another source of pre-appraised reviews of current pediatric literature.
- University of Washington Evidence-based Pediatrics - has a collection of CATs on pediatric topics
- BestBets - Emergency Medicine EBM at its Best - has a collection of CATs on questions asked in the emergency room setting (and often seen in the primary care office)
- Drug Effectiveness – Evidence-based drug therapy from the University of British Columbia
- Bandolier – is a monthly evidence-based electronic newsletter from the Centre for Evidence-Based Medicine in England
- PubMed – basically a medline search but has some built-in filters to emphasize clinical resources. It is usually advisable to start with another data base like Best Evidence. PubMed (like any other medline search engine) can overwhelm you especially if you are not familiar with all ins and outs of searching medline. This should be used as a last resource when other evidence-based resources do not yield answers to your questions
- Guidelines – well-done, evidence-based guidelines can be an excellent source.
- National Guideline Clearinghouse – a federal supported site that posts evidence-based clinical practice guidelines. It is sponsored by the Agency for Health Care Research and Quality (AHCRQ), the AMA and the American Association of Health Plans.
- Primary Care Clinical Practice Guidelines – a compilation of guidelines from around the world.
- HSTAT – Health Services Technology Assessment – federally sponsored collection of resources including CDC guidelines, NIH consensus statements, US Guide to Clinical Preventative Services, AHRQ guidelines.