Herramientas de usuario

Herramientas del sitio


# 3. Searching for studies

## Planning the search process

### Role of Information Specialists

#### Working with Information Specialists Conducting a systematic review requires a very thorough search for potentially relevant studies. A Cochrane Information Specialist or professional healthcare librarian, preferably with experience in supporting systematic reviews, has the expert knowledge to help you plan and conduct your search.

What they do Information Specialists are generally qualified librarians by background, with past experience of searching databases and other resources in a healthcare library, information service, or other similar setting.

This experience allows them to help you find the most relevant studies for your review.

What you need to do As a review author you need to have a good understanding of the overall search process so you can:

  • contribute to the search strategy design, for example by suggesting a range of search terms for your topic, or providing references to relevant studies you have already identified;
  • understand what the Information Specialist is doing;
  • identify and assist with any additional searching that can or should be done;
  • work with the Information Specialist to report the search strategy accurately in the review.

You and the Information Specialist will work together collaboratively.

When to start You should contact an Information Specialist and ask for assistance early, when you are considering the topic for your review.

#### Importance of a rigorous approach to searching A rigorous approach to searching is a defining feature of systematic reviews, to ensure you find and include as many relevant studies as possible.

A systematic literature review search is probably one of the most defining features for a systematic review. It's not that difficult to find a few things that answer a clinical question, but it's much more challenging to find all that there is to find. And one of the things that is really important, particularly for new authors, is not to fall into the trap of thinking that because you are very familiar with the literature, that there is no need for a systematic search.

A Cochrane systematic review literature search needs to be rigorous, it needs to be reproducible, and it needs to look across a wide variety of sources. In particular you have to think about somebody coming behind after you and reproducing that search. Could somebody take what you have done and run the same search, and maybe not get exactly the same results, but essentially reproduce what you did?

Why a limited search is not sufficient

If you do a limited search you get a limited number of trial reports and your systematic review is fundamentally incomplete. And everything that comes after that – no matter how good your methods are – if your review is not based on the greatest number of trial reports, then everything that comes after that is flawed in some way. So that's the most obvious thing that you are trying to avoid.

We are also trying to avoid selection bias. It's much easier to find some reports then it is to find others. Reports published in English, reports that are indexed in a database like MEDLINE, reports that have been published in high impact journals – those are all quite easy to find, but of course they may not be representative of all of the studies that are out there.

You also have the challenge of reports with positive results being easier to find. So, those kinds of studies that report positive findings make their way into the journal literature much more quickly – and therefore into the databases – and therefore are much easier to find. So, a systematic review literature search guards against those kinds of biases.

You also have got the challenge of applicability: the fewer studies you find the potential is that you won't capture studies that relate to all the populations, potentially, in which the intervention has been tested. And that will lead to people thinking that your systematic review doesn't apply to the populations that they're dealing with and is therefore less generalisable.

Sensitivity and precision in searching

The starting point is that you have to have a wide variety of sources to search. And then you need to develop a search strategy. This is particularly important for the bibliographic databases. And developing those strategies, in doing that, we take what we would call a sensitive approach. Basically what we're looking to do is try to find the largest proportion of reports that answer your question out of all of the reports that are in a particular source. The consequence of that for author teams is that very often you could have thousands, sometimes even tens of thousands of records to scan through. But that minimises the risk of you missing something relevant. The alternative to that, of course, is to do a much more focused search, what we sometimes call precision. And precision obviously gets you a number of records that are closer to mark. But the risk is that you will miss something that's relevant. There's always a balance and it's important that when you are developing a search strategy that you get expert input, particularly from your Review Group’s Information Specialist or any other Information Specialist.

### Sources to search

#### Key databases Databases are where your search for studies will start. You should search multiple sources, to identify as many relevant studies as you can. There are a few key databases that make a good place to start.

##### Where to start You should start with the sources that are likely to be the most productive, and return the most studies. From here you can then move to additional sources.

If your review will include randomized or quasi-randomized trials, start your search with CENTRAL (the Cochrane Central Register of Controlled Trials), as this is the most comprehensive source available for identifying controlled trials. Then, search other electronic bibliographic databases such as MEDLINE (which is freely available via PubMed) and Embase (if you have access). These databases are likely to provide most of the studies you use.

Once you’ve searched these major databases you can then add additional sources as appropriate to your review question, until you have built up a set of sources to search that is as comprehensive as possible, whilst maintaining an achievable workload.

##### Searching CENTRAL Known as CENTRAL, the Cochrane Central Register of Controlled Trials includes over one million records of randomized and quasi-randomized trials of interventions.

CENTRAL contains:

  • all records from MEDLINE that are indexed with the publication type term ‘randomized controlled trial’ or ‘controlled clinical trial’, except those that are indexed solely as animal studies;
  • all the records of the trials identified to date from Embase;
  • the results of additional searching for eligible trials conducted by Cochrane Centres, Fields and Review Groups (CRGs), including searching other databases and handsearching of individual journals.

CENTRAL is available through the Cochrane Library, and you can find further information on the Cochrane Library website. Detailed guidance on how to use and search the Cochrane Library, including CENTRAL, is available from the link in the Resources.

#### Other sources of published studies Beyond the key bibliographic databases such as CENTRAL, MEDLINE and Embase, there are other sources to search for published studies.

##### Which database would you use? Imagine you were writing a review of pharmacological interventions for treatment of Chagas disease, a tropical parasitic disease occurring mostly in Central and South America. Which of the following databases would you refer to? (Have a look at the Searching for studies document in the Resources if needed).

Published reviews Other published reviews and guidelines are also a useful source of information about the individual studies they include. Previous systematic reviews are especially useful, as they provide information about the search strategy used, which may inform your own review.

There are sources available to help you look specifically for reviews and guidelines. You can find some useful links in the Resources.

Reference lists Reference lists in individual studies included in your review should be assessed for information about other relevant studies. This will be particularly useful where the terms used to describe a condition or intervention change over time, or where different disciplines use different terms, making it more difficult to include all the relevant terms in your search.

You can also use citation indexes like Web of Science or Scopus to find later articles that cited those you’ve already found. Some databases have a feature that will find related articles to those you know are of interest.

Searching individual journals The full text of many journals is available electronically on the internet. Access may be on a subscription basis or free of charge. In addition to providing a convenient method for retrieving the full article of already identified records, you can also search full-text journals electronically, depending on the search interface, by entering relevant keywords in a similar way to searching for records in a bibliographic database.

Another approach sometimes used is handsearching: going through each issue of a journal or conference proceedings page by page to identify studies to include in your review.

These approaches can be useful alongside your searches of other sources, as not all journals and articles are included in bibliographic databases. Even when they are included, individual studies may not be well described or indexed, and they may be missed by your search terms.

This work is very time-consuming and may not be worthwhile given the small number of relevant papers that are likely to be identified this way. Consult your Information Specialist if you feel that there are key journals (whether indexed in MEDLINE or not) that you would like to search directly for your review. Remember that over 3,000 journals have already been handsearched by Cochrane, and these results have been included in CENTRAL, so be sure you are not duplicating this effort.

##### Grey literature Grey literature is work that “is produced on all levels of government, academia, business, and industry in print and electronic formats, but which is not controlled by commercial publishers, i.e. where publishing is not the primary activity of the producing body”. (Source: Sixth International Conference on Grey Literature, New York, USA, 2004)

##### The challenge of identifying grey literature Locating grey literature can often be challenging, requiring the use of a number of sources from various host providers or websites, but it is important to put in this effort. Failing to identify studies reported in grey literature might introduce bias, and affect the results of your systematic review. Therefore, it is essential to search at least in trials registers to look for ongoing and unpublished studies.

There are some resources that can help you plan a manageable and thorough approach to the grey literature for your topic. You can find some useful links in the Resources.

#### Designing a search strategy

##### Structure of a search strategy Before you start your search you need to design a carefully structured search strategy.

##### Specific expertise required Systematic reviews require a complex, rigorous search process, so you should not attempt this alone unless you have specific expertise. Getting assistance from an Information Specialist is strongly recommended.

The basis for the search strategy for reviews of interventions is the PICOS format, which is used to map out the question for the review, as explored in module 1: Introduction to conducting systematic reviews, and to define eligibility criteria as seen in module 2: Writing the review protocol.

To design a highly sensitive but adequately precise search, start by focusing on the two or three most important concepts for your topic. P + I + S

#### Turning concepts into search terms The search terms are built around the selected key concepts from the PICOS format.

##### Populations and interventions Imagine that you are working on a systematic review entitled ‘Helmets for preventing head and facial injuries in cyclists’.

What would be the population, intervention, and study design of interest?

##### Turning concepts into search terms Now let’s see how these concepts can be turned into search terms.

You want your search to be as sensitive as possible, so it finds as many relevant studies as possible. This means that when thinking of your search terms, you need to think of all the possible ways that your concepts of interest might be described. This will help you to minimize the risk of missing a study because it used different terminology or expressed the concept in a different way.

Using a broad range of terms will, however, reduce the precision of your search – you increase the risk of finding irrelevant studies. You have to balance the risk of missing studies with the cost of having to screen through more search results.

Note that although using too many different concepts in one search is not recommended, you should use a wide variety of individual terms within each concept to describe your search.

##### Search terms You need to use two different types of search terms to make sure you’re as thorough as possible in including all the ways your concepts can be expressed.

Text words Text words (also known as 'natural language terms') are used by authors and appear in the title and/or abstract of a publication. When searching, remember to:

include synonyms, related terms, opposites, international terms, alternative spellings (e.g. brain injury, head injury, skull fracture); use tools such as truncation and wildcards (e.g. * $ ?) to identify variations in the words you're looking for (e.g. injur* for injury, injured, injuries); use proximity operators (e.g. NEAR, NEXT, ADJ) to identify search terms that are near to each other (e.g. (Fracture adj2 skull) for fracture of the skull); Bear in mind that the way these tools work will be different according to the interface, or service provider, you are using to access a database (e.g. if you use Ovid or PubMed to access MEDLINE). Your Information Specialist can help you get this right.

Controlled vocabulary Controlled vocabulary (also known as 'index terms'), or subject headings, are pre-defined terms that are added by the database indexers to describe the topics mentioned in the publication. Each database has its own system of terms (thesaurus), for example MeSH in MEDLINE and Emtree in Embase. They are usually organized in a tree structure from very broad terms down to increasingly specific terms.

A database will use the same set of vocabulary no matter which interface or service provider you use – for example, you will use MeSH (Medical Subject Headings) for searching MEDLINE, no matter whether you use PubMed, Ovid or another service provider.

It is important to use controlled vocabulary, because it will find the concept whether or not the specific phrase you have searched for appears in the article. For example, if you search on 'dental caries' as a MeSH term, the search will also retrieve records on tooth decay, even if it's not referred to as dental caries anywhere in the title or abstract.

Most database interfaces will suggest controlled vocabulary terms that match text words if you enter them into a search. For example, to do this in PubMed, you can try searching directly in their MeSH database.

Let’s take for example, a search that’s being undertaken for a review on protective equipment for all kinds of sports. You could start with the term ‘Sport’ and you will find that the corresponding MeSH term is actually ‘Sports’.

It’s also useful to check the definitions of the terms suggested, as they may not be what you expect.

Also, you should think carefully about the suggested terms – is there a more specific term that would be better?

When using controlled vocabulary in this way, you can ‘explode’ the results. The term ‘explode’ is used to mean searching a heading and including all its narrower, i.e. more specific terms.

So, for example, if you ‘explode’ the the MeSH term ‘Sports’, you will find a list of more narrow, specific terms listed under the heading ‘Sports’ in the tree, such as Baseball, Boxing and Soccer.

Developing a search strategy is an iterative process in which the terms that are used are modified, based on what has already been retrieved. Some preliminary test searching may help you with this.

So, for example, you may try searching the term ‘sports protection’, and you will find the Mesh term ‘Sports equipment’.

Once you've identified relevant MeSH terms, you can add them into your overall search strategy for use in your review

#### Study design filters Search filters are sets of search terms intended to identify a particular study design or topic. Those designed to identify particular study designs are called study design filters or methodological search filters. These can be useful in reducing large numbers of records retrieved by a search.

##### Applying filters When applying these filters, you will need to select the right study design filter for your review question (e.g. for RCTs or other study designs) and for the database and interface you are using. You can then add the study design filter to the search terms you wish to use for your review. Your Information Specialist can suggest an appropriate study design filter.

More information is available in the Cochrane Handbook Chapter 6.4.11.

##### Word of warning Database search interface often change and controlled vocabularies are frequently updated. Keeping this in mind, you should check that the search filters you are planning to use are still accurate, relevant, and effective before using them. Your Information Specialist can advise you on this.

#### Boolean operators Boolean operators are words that are used to combine the individual search terms within each concept, and to combine the major concepts of your search into a complete search strategy.

There are two main Boolean operators, the first is OR. If you say you want to search for records that contain the word bicycle or the word cyclist. This makes your search broader. It will find any record that has one or the other word present or both. One way to remember this is, OR is more.

The second Boolean operator is AND. If you say you want to search for records that contain the word bicycle and the word helmet, this narrows down your search. It will only find records where both of those words are present. To bring all your search terms together you need to use a combination of both OR and AND.

Within each concept you have a variety of terms relevant to that concept. For your review of bike helmets. You have a group of terms around your population, i.e. cyclists, a group of terms around your intervention, i.e. helmets, and your study design filter to identify RCTs or other study designs.

So, for our population concept we want each record to have at least one of the terms listed such as bicycle or cyclist or the mesh term bicycling. We use the Boolean operator OR to join together this set of free text terms and thesaurus terms. This is because we only need the record to have at least one of these terms for this concept. Using OR to combine terms widens the search to try to lessen the chance of missing relevant records. We do the same for the intervention terms.

Once we have grouped together our sets of terms within each concept and have a bottom line search strategy set number for each concept. We join all the individual concept bottom lines together using the Boolean operator AND because we want each record to have at least one term from every concept. Using AND narrows the search excluding the records that don’t have all three of our concepts.

What we hope to find are those records that have the right population and the right intervention, and the right study design without missing any, and we want to avoid those that have only one or two of the right concepts like RCTs about bicycles or editorials about bike helmets. Those records where all three concepts overlap are the ones that will be relevant to the review and the ones we want to retrieve.

#### Limits and restrictions As part of a systematic and sensitive search, you need to avoid excluding any possibly relevant study, regardless of where, when, or how it was published.

##### Missed results The problem with using limits and restrictions is that your search might miss some relevant studies. For this reason you should, generally speaking, avoid using limits and restrictions, and if you do use them, justify your decision.

Language Limiting your search to studies published in English risks introducing bias, as trials with positive results are more likely to be published in English-language journals (you can learn more about reporting biases in module 7: Interpreting the findings).

Don’t forget that even if you don’t add any language limits to your search, if you only search databases where the majority of the publications are in English, you will miss much of the non-English language literature. Think about including some of the national and regional databases (see more information in Resources).

Publication date Restriction by year of publication should be avoided unless there is a clear time period during which relevant studies might be found, for example if the intervention first came into use in a particular year.

Publication format Remember that important information about studies may be published in letters and commentaries, including information about studies not published elsewhere. Also, important concerns about previously published studies, such as errors or fraudulent data, may first be published as comments.

#### Errors and retractions It’s vital to remember that studies can be subject to errors or retractions.

##### Searching for errors and retractions When considering the eligibility of studies for inclusion in a systematic review, it is important to be aware that some studies may have been found to contain errors or to be fraudulent, or may, for other reasons, have been corrected or retracted since publication.

Including data from such studies can impact on the overall estimates in your systematic review. It is essential at the search stage to find out whether any retractions or errata have been published for the included studies.

Relevant terms can be incorporated into your search strategy to help retrieve retractions e.g.: (retracted publication or retraction of publication).

Your Information Specialist can advise on this.

#### Piloting the search strategy Before going ahead with your search strategy it is a good idea to test it first.

##### Why pilot? Piloting your search strategy lets you see whether the results it identifies are what you expect. If there are too many, or two few results you can revise your search strategy accordingly.

Too many results What if you retrieve lots of irrelevant results? You may be able to modify your terms to improve the precision of the search, but remember to be careful and seek advice – it’s easy to accidentally exclude studies you do want.

Too few results What if you retrieve too few results? Check that you haven’t made an error in your syntax. It may be that you have included too many concepts in your search, or that you have not thought of a wide enough range of terms for one or more of your concepts.

#### Managing the references and documenting and reporting the search ##### Managing the references Managing the references to studies identified for your review is extremely important, and can be complicated if you are managing a large number of studies in your review.

###### Using reference management software Reference management software is a good way to organize your search results. There are many options, and you can use whichever software package is available to you. You may have access to this software at work, or there are free options available. Let’s take a look at some of the benefits of using reference management software.

Bring all your search results together and remove duplicate records You can also use the software to organize your search results, and to identify and remove any duplicate records found in more than one database (for example, where exactly the same article is indexed in both MEDLINE and Embase – not two separate articles reporting the same study).

Remember, if your Information Specialist is running your searches, they may have done this for you already.

Store information about the methods and process of a search Reference management software can be used to store information about the methods and process of a search. For example, unused record fields can be used to store information such as:

  • the name of the database or other source details from which a study report was identified;
  • when and from where a full text copy of the publication was ordered, and the date it is received;
  • whether the study is included in or excluded from your review and, if excluded, the reasons for exclusion.

Import references to RevMan RevMan is Cochrane’s review production software. When you’re ready to write the review, you can import references from reference management software into RevMan or other review production software (such as Covidence or EPPI-Reviewer) without having to enter them all by hand.

#### Documenting and reporting the search You will need to report the search in detail in the review, so that your readers can see (and, if necessary, replicate) what has been done to attempt to find all the relevant studies.

##### The importance of documenting and reporting the search It is important that your search is well documented at every stage, not only so it can be replicated, but also because you too need to know what has been done. This will allow you to keep track of all your results during the review process, and will help you or others update the search in future.

Template texts for writing search methods in your systematic review are available in the Resources.

Databases and other sources searched Record the name and interface for each database (eg. MEDLINE (Ovid)) or MEDLINE (PubMed)); the date range of the database; the URL of web sources. Also, document and report the details of other sources that were searched.

Dates the searches were conducted When were the searches conducted? Details on reporting search dates in Cochrane Reviews can be found in the Editorial and Publishing Policy Resource.

Date limits applied to the searches If you applied any date limits to the searches, such as studies published after a certain year, or during a specific time period, you should document these limits.

Search strategies used In an appendix of your review you should include the full search strategies for every database and any other source searched. Include the set numbers of each row and the total number of records retrieved for each search strategy. Search strategies should be copied and pasted to avoid errors in reporting.

Total number of records retrieved You will need to report three figures in the PRISMA flow diagram:

  • the total number of records retrieved from all your sources;
  • the total number of records retrieved minus duplicates;
  • the number of records coming from sources additional to the database searches.

Find out more about PRISMA in module 4: Selecting studies and collecting data.

Non-database search activities Don’t forget to record non-database search activities, such as contacting experts and reference list searching.

Copies of information found on the internet Make sure you keep copies of search results and information found on the internet. You never know when a database might be closed down, a website moved, or your access rights changed. Do not rely on simply saving the URL.

#### Updating your search You may need to revisit and re-run your searches before the review is completed. Although the search is usually the first thing you start after the protocol is published, it can take a long time to complete a review. It will be necessary to re-run the searches if they were conducted more than 12 months before the review is ready for publication.

##### How to update a search

Before the review is completed you may need to revisit your search. Although the search is usually the first thing you start after the protocol is published, it can take a long time to complete a review.

If your search is more than 12 months out of date when the review is ready to be published, you should update it before submitting the review for publication. Best practice is to do this if your search is more than six months out of date if at all possible.

Whenever you update a search, you should make sure you revisit every source (not just the electronic databases). Having good documentation will help you with this. If you do not revisit every source you originally searched, you should report this with justifications.

If you have updated your search, you should screen through the results and identify any potentially eligible studies. Ideally, the results of eligible studies should be fully incorporated into the review before publication. This aim should be balanced against the potential delay in publication – if it would result in an extended and unreasonable delay, then at the very minimum you should list the potentially eligible studies in your references (for Cochrane Reviews, under the ‘Studies awaiting classification’ or ‘Ongoing studies’ headings).

It’s important to note that sometimes there can be changes in the electronic databases over time – controlled vocabularies and definitions change, and sometimes the search syntax can also change. Check with your Information Specialist to make sure there haven’t been any changes that could affect your updated search.

Note that this process is different from the searches you undertake for updating your review.

### Protocol considerations

#### What to include in your protocol You should outline details of your search plans in your protocol, developed in collaboration with an Information Specialist.

##### Considerations for your protocol Let's look at what to include in your protocol when considering searching for studies.

  • At the protocol stage, you should list all the sources you plan to search (including trials registers and other sources), and justify any limits you plan to use (for example, to exclude articles published before the invention of a drug). Don’t forget to include all sources, not just databases – including reference and citation checking, web searches, and people/organizations you plan to contact.
  • You should include at least one full database in your protocol, unless the search strategy has been published elsewhere. If this is the case, then the search strategy can be referenced rather then reproduced.
  • You should report search terms that will be used to search any sources other than bibliographic databases (for example trials registers and the web).
  • If you plan to include study designs other than randomized trials (for example, non-randomized studies, economic evaluations, or qualitative studies), your search strategy should be designed so that these types of studies will be retrieved.

MECIR standards

MECIR are Cochrane's methodological standards for developing protocols, reviews, and updates. The following MECIR items are relevant when planning and reporting the search methods in your protocol:

  • C19 – Planning the search.
  • PR17 – Search sources.
  • PR18 – Search restrictions.
  • PR19 – Searches for different types of evidence.
  • PR20 – Search strategies for bibliographic databases.
  • PR21 – Search strategies for other sources.

A detailed description of these items and the full list of MECIR standards is available in the Resources.

Template texts for writing search methods in your protocol are also available in the Resources.

cursos/cursocochrane/busqueda.txt · Última modificación: 2017/12/21 06:00 por