Performing a database search and an initial analysis of the search results
In our previous blog we focused on understanding the goal of our literature search and on writing a good, solid literature search and review protocol.
In this second part of our three-part series, we will discuss how to conduct a database search, how to recognize any errors you may have made and finally how to achieve better search results.
Step One: Perform a search
The search must be conducted according to the literature search protocol by applying the keywords and terms and the Boolean operators you have chosen to the databases you have selected.
Let’s use the example from our previous post.
We want to perform a search to find articles on State of the Art related to traction splints for femoral fractures. In our Protocol we have selected the following MeSH terms: traction, splints, immobilization, femoral fractures and their combinations with Boolean operators:
Traction AND (Splints OR Immobilization) AND Femoral Fractures
We will now perform the search in the PubMed database by using the PubMed Advanced Search Builder which is available at this link:
You can browse by all fields or within specific fields such as MeSH Terms.
In the Search Builder where we add terms to the query box, we can enter a term and then click on Show Index to choose a term and perhaps also the subheading we want to include in our search. The index displays an alphabetical list of the resulting terms and the approximate number of citations for each of them.
For more options for an advanced search, please see:
In the query box our search will look like this:
(((“traction”[MeSH Terms]) AND (“splints”[MeSH Terms])) OR (“immobilization”[MeSH Terms])) AND (“femoral fractures”[MeSH Terms])
The Search Builder automatically adds round brackets so we must check if these are in line with the search concept from our protocol. If this is not the case, then for our example we would manually refine our search to:
((“traction”[MeSH Terms]) AND ((“splints”[MeSH Terms]) OR (“immobilization”[MeSH Terms])) AND (“femoral fractures”[MeSH Terms]))
We got 113 results which is probably too much so we can apply some filters (our inclusion/exclusion criteria) to achieve more specific results. So, we can select – Review or Systematic Review and apply a time span for the publication dates of between 2000 and 2020.
Click the filter you would like to activate from the sidebar.
Now we have only 4 results so we must broaden our search results.
Step Two: Broaden or narrow the list of obtained search results
At this stage, we can see how well the search strategy we have created works. The results from all the databases we have used should be precise with their range neither too high nor not too low but sufficient to support claims and to reach a valuable conclusion.
Since the research is dependent on our search strategy, if we are not satisfied with the results then we can add or remove inclusion and exclusion criteria, use different Boolean operators or other terms or keywords and finally we can refine the search protocol. We should adjust our search results to a valid scope that corresponds with the goal of the literature review.
Several searches with different search criteria or focus are usually necessary to obtain the necessary data.
For example, our previous search achieved a poor result, just 4.
We can remove the exclusion criteria but we must not forget to document this and any other changes in our literature search and review protocol.
For example, we can remove the filters Review, Systematic Review, and reduce the time filter from 2010 to 2021 (state of the art data should typically cover 5 to 10 years).
Now we have 15 results and this is an acceptable number since search results should provide between 10 and 50 articles.
In addition, it is a good idea to perform a quick check to ascertain that all important articles have been caught in the search. Use the same filters but different, similar terms or use keywords instead of MeSH terms. If significant works do then appear you will know you will need to refine the search.
In our example, we can replace MesH terms with keywords and in the Search Builder we can browse within the Title/Abstract field:
((traction[Title/Abstract]) AND (splints[Title/Abstract]) AND (fractures[Title/Abstract]))
Again, we now have 15 results and the results are identical so we have confirmed we can proceed with our initial search strategy using MeSH terms.
However, if the results show any further significant articles, then refine the search and use this new concept for the search strategy instead of using the MeSH terms. Again, be sure to document this in the protocol.
Tip 1: Most published works have citations that may lead you to additional relevant works and appropriate MeSH terms. However, be careful not to become lost following the path of bibliographical references.
Step Three: Do the First Level of screening
Now we can undertake the first level of screening the results based on inclusion and exclusion criteria and that involves screening the title or abstract.
In this step you don’t have to read the full texts of all the works found in the search. You can exclude publications that are obviously not relevant on the basis of the contents of the abstract which will indicate whether or not they are related to the clinical evaluation or don’t have any connection to the device under evaluation. Again, you must document this step since notified bodies may wish to confirm that all potentially relevant articles were considered for inclusion.
This means that a record should be maintained of all the literature returned by the search and this record will be added to the CER in the form of an Appendix, i.e., Relevant data of literature search identification.
Abstracts lack enough detail to allow issues to be evaluated thoroughly and independently but may be sufficient to allow a first evaluation of the relevance of a paper. Copies of the full text papers and documents should be obtained for the appraisal stage.
Step Four: Save the search
It is important to save the successfully performed search in the database you have used and then export the search results. In this way there will be a visible trace of the searches conducted and the data sources which can be provided to notified bodies.
Tip 2: It is recommended that reference management software, e.g. EndNote (www.endnote.com) is used to manage the literature search results.
After the search process has been done and we have saved the search in an appropriate format, we can then progress to the second level of screening: full text screening and appraisal of the literature.
In the next blog, we will focus on how to appraise the search results.
Literature Search and Review at BioReg Services
BioReg Services specializes in systematic literature searches and reviews in compliance with the MDR Regulation (EU) 2017/745. Our team can design a comprehensive literature search protocol with complete documentation of the literature screening and data extraction process. We can obtain clinical evidence for the literature review for your new medical device or update the literature review of a CE-marked or legacy device to create an MDR compliant document.
Feel free to contact us:
LLB. Legal & Literature Clinical Evaluation Report (CER) Expert
Founder & Managing Director, M.Sc. Mol. Biol., RAC, Regulatory Consultant