Welcome back. Here we are now in Section B. We're going to cover Identifying Key Sources and the Techniques that you use for searching those sources. That will include a variety of electronic databases. I'm going to introduce at this point one of the features of the database called a Controlled vocabulary to help mem be able to describe the various databases and indexing, particularly for randomized control trials and observational studies because I know those are the types of studies that you all are going to be using in your systematic reviews. I'll also be going over the other sources that need to be covered in order to do a comprehensive search. The next two slides present a list of Major Bibliographic Databases that you would use to find randomized control trials and observational studies. The first set are focused within the medical domain, the next one is an example of regionally based database that covers literature produced within particular areas of the world. And then there are a whole series of subject specific databases. And I've picked out a few to illustrate this for you. You may be wondering whether really is important or really necessary to do less extensive a search. What this slide shows you are results from a study by Lawrence on five different areas, key areas, in public health injury prevention. He looked at four different databases and identified the studies that would be relevant to those five areas. And if you look at the right hand column, you see that It's labelled Unique articles. Every single one of those areas, every single one of the databases contributed something unique to the study. So it was in fact necessary to search all of them in order to get all of the literature relevant to the questions. Almost all of the large, bibliographic databases, at least the ones that are subject specific, include something called a Controlled Vocabulary. What this is, is a standardize list of terminology that's used in indexing information to facilitate retrieval. So that means that regardless of whether a paper is published in English or French, if it's about heart attack and the controlled vocabulary term is myocardial Infarction. When you use that you will get all of those articles in your search. So it provides a Consistency. In doing a systematic review though, it's important to use both controlled vocabulary and keywords. By keywords I mean any kind of free text searching, that you would do to look for words in articles or titles and other fields. Little more on PubMed. People often use the term MEDLINE and PubMed interchangeably. PubMed is the online access that the national library of medicine provides to the MEDLINE database and additional citations that are ruled into the product. MEDLINE consists of citations and abstracts from over 5,600, journals. And each of those citations has been indexed. I just described controlled vocabulary. There are actually PhD professionals at the National Library of Medicine who review the article. Select terms from that controlled vocabulary and add them to the record so that, whether or not something is mentioned in the abstract, or even if it doesn't have an abstract, you'll still be able to find the relevant article. Currently, there are over 19 million records in MEDLINE, total of 22 million in PubMed, that includes newer articles that haven't been indexed yet and some of the older articles that have been rolled in. A little more about PubMed. The controlled vocabulary that's used in PubMed is the medical subject headings or MeSH that consists of Descriptors and subheadings. There's a thesaurus available that you can search for these terms, and you can find also the definition that the indexes use when they apply a term. The list is not just one straight list, there are narrower terms and broader terms. And the indexes, index at the most specific term that would apply to the article. Particularly for those of you looking for randomized trials, publication type is something to pay attentions to. And in terms of keyword searching, there are titles available back to 1966 and abstracts to 1975. This slide shows you an example an entry for a mesh term, a medical subject heading for Macular Degeneration. See the term at the top. The definition that's used by the indexer. Narrower terms called subheadings that are used when an article is specifically about for example Microbiology related to Macular Degeneration. For the purpose of a systematic review, I would stay at the descriptor level, that is the Macular Degeneration. You're going to get too specific if you're using subheadings. If you scroll down to the bottom of the page, which is on this slide, you'll see what are called Entry Terms. For indexed articles, any of these terms that appear this article will have the MeSH term Macular Degeneration applied to it. You can use this list of entry terms though to find good keywords with non-indexed articles that have for example, Maculopopy, wouldn't necessarily come up with Macular Degeneration. And at the bottom of the page, you see the tree I described with narrow terms and broader terms. If the article is about Macular Degeneration, the indexer should use that term. For searching purposes, know that if there are narrower terms below the term you used, they will also be included automatically unless you tell PubMed not to do that. Now I'll say a few words about indexing specifically for randomized control trials and observational studies. Despite the importance of randomized trials to clinical medicine, the has been a challenge to develop indexing to find these studies effectively. Up to 1977, there were no specific terms for randomized trials and as you can see from since then, the terms have changed over time. Again the point here is if you want to get all the literature, including the older literature, you need to include in your search strategy the older term as well as the current term. The Cochran Collaboration here at Johns Hopkins was involved in a partnership with the National LIbrary of Medicine to improve the indexing for randomized-controlled trials. So, it has, indeed, got better over the years. When it comes to observational studies, there's an even broader range of terms available. Depending on the topic that you have, you'll want to check and see what the terms would be, here is a list of some of the terms that are in use for Epidemiologic studies, Case control, Cohort studies, and so on. Earlier indexing is shown here, as well as the website that you can go to, to get into the MeSH browser. And I would encourage you to take a look and see which terms would apply for your study. There hasn't been a whole lot of research done on finding Observational Studies in PubMed. There was one pilot study done by Susan Wieland, who looked at how well indexed these papers were, and how easy it was to find them. Here's a table that shows some of her results. She found that Outcome terms and Design terms were fairly well indexed but Exposure terms less so. If you look at the bottom of the table, you'll rows for Precision and Sensitivity. Precision refers to how well the search did at Identifying only the studies that were relevant and not getting false hits or inappropriate articles. And Sensitivity refers to how good the search was at being comprehensive of getting all of the citations that were relevant. And she could calculate this because she knew how many studies were available for the various searches that she was doing. PubMed has over 22 million citations and covers over 5,600 journals but there is still more material out there. The other major medical bibliographic database to consider is EMBASE which is like a European Medline that was based on the index called Excerpta Medica, and is now provided by Elsevier. We have access to it through endbase.com, and they actually licensed the MEDLINE course, so when you search endbase.com you're searching both of them, unless again you tell it otherwise. There's nineteen million citations here, and over 7,000 periodicals indexed. There's also a controlled vocabulary, but it's not MeSH terms, they call it EMTREE terms, and it's a somewhat different vocabulary. The other very important source to use is the Central Register of Controlled Trials that's provided by the Cochrane Library by the Cochrane Collaboration. This is a group of researchers that a voluntary organization that's committed to doing high quality systematic reviews. They're arranged in review groups based on topic area. And each of those groups combs the literature for controlled trials and then contributes them to the Central Register of Trials. So you can benefit from their work by doing your search in the central registry to compliment other searches in PubMed and Embase. They index their articles using MeSH terms. At the beginning of this section, I talked about a set of databases or two databases, the Web of Science and Scopus that allow you to do citation searching. Let me just say a little bit more about the Web of Science. It differs from the first three databases in two major ways. One is, it's much broader in terms of the disciplines covered. You have articles from the social sciences, arts, and humanities, as well as all of the sciences. For every single article in this database, some of which go back to the 1800s, the reference list is included as well. What that does for you is create a web. If you have a key article that's on topic, from say ten years ago, not only can you look at the reference list online, but you can look and see who has cited that article since it came out. So you can follow trends of research over time and augment, the search strategy that you do with specific terms. Because it's so broad, there is no controlled vocabulary and I've given you some information about some of the ways to use the database on the slide. What I was just describing is sometimes called Snowballing. You can start from one reference and then select other relevant references from the reference list or move forward to the citing articles. In PubMed you can also do that using the related articles feature which other databases sometimes call Find Similar. All of these sources I have talked about now present you with peer reviewed published literature. It's also important in doing a comprehensive, sensitive search for a systematic review, to cover sources of unpublished literature. There's been a move in the last few years to requiring systematic reviewers, to search clinical trials registry, and I've given you a list of three, well, four possibilities to go look for trials. Additionally government sites for example the FDA are an excellent source for literature for references to studies. At least for those drugs that the manufacturer has put forward for licensing for use. Identifying key Organizations and foundations, I might also be publishing literature in this area and then finding their websites and searching for literature is a good idea. In doing sort of a general, quote unquote gray literature search online you might consider using scirus.com. It's tailored to find high quality literature and various sources from the open web. And here I've given you the web address for the FDA site. So again, there has been some research done on, whether it makes a difference to include unpublished literature. Hart and his colleagues did a meta-analysis of citations related to nine new medications and compared the results, with and without including the articles of the studies that were identified through the FDA site. As you can see from the slide, in several cases, including it or not made a difference. Sometimes, it actually showed a higher estimate of efficacy, sometimes it showed a lower indication of efficacy. In general however, the estimates of harm was always greater when unpublished data was included in the meta-analysis. A few more words about Hand-Searching for Cochran reviews, this is one of the mandatory aspects of doing a comprehensive search to support a review. What you need to do is identify key journals that will cover your topic area and select a number of years. And then review the table of contents. It's no longer necessary to do this by hand. The journal table of contents are available online, so it's easier but it still needs to be done. Conference proceedings are available from Biosis and open web searching, and you also need to review the reference list of any study that you have included in your study. And again, I've listed here a couple of sources for Gray literature, Dissertations, and reports. There's also been research done on whether doing this kind of Hand Searching, which takes time and effort makes a difference. Sally Hopewell and her colleagues who looked at systematic reviews to see whether all of the RCTs relevant to the topic would be included, Hand- Searching identified those reviews very effectively 92 to 100% accuracy. You can see from the list that the search is in MEDLINE, Embase and PyscINFO were less sensitive. There's something I'll be showing you a little later called the Cochrane Highly Sensitive Search Strategy for RCT's that was developed initially in 1994, and that was more effective at finding them. Unfortunately it takes too much time to just hand search all of the literature. So what we wind up doing is establish search strategies for these databases and then augment that with enough hand searching to feel that it's being fully comprehensive. And don't forget also to use your personal contacts whether it be experts in the field that you're working in or any contacts you have within the industry to help you identify and find articles or studies that are relevant to your topic. What I'm going to do in the next section, is explain how to put together the kind of strategy that you would use in the large databases. High -quality electronic search strategy.
