Many early career scientists struggle with the task of screening, reading and collecting biomedical literature. Lack of familiarity with search tools, article structure/organization, scientific terminology and time management are obstacles that inexperienced researchers face when looking for information. No single approach will overcome all the hurdles but with practice, users can identify strategies that work best for them and increase the likelihood of finding the information needed, quickly and efficiently.
The first and most important step is to know what you are looking for. How specific is the information? Is it very detailed e.g., an antibody protocol or a broad question such as the frequency of a marker expressed in a disease? You should know what you are looking in order to help organize a search and narrow down the results to a manageable size. Searching by a keyword using broad terms i.e. immunohistochemistry will return hundreds of thousands of article citations, many more than is physically possible to sort through to find specific information. By adding more specific terms e.g. a specific antibody, to the search string, the list will narrow down significantly. The aim is to reduce the citation list to a manageable and representative size, perhaps 100 or fewer citations. This number can be screened in a relatively short time although longer and shorter lists can also be vetted. Regardless, since each search is unique, users should experiment until they become familiar with different approaches to keyword searching (using Boolean operators and trying different terms) by comparing the output lists to find the most viable solution or approach. Searching by author(s) name tends to return shorter citation lists. However, you need to know the author’s name to execute. PubMed follows the journal’s syntax for author’s names while Google Scholar does not and common names will return very large citation lists in both engines. Despite this, author(s) name and keywords can be combined to narrow down searches. Lists generated by either search engine can be limited or filtered in a number of ways but care must be taken to not restrict the search output to only a few articles. Depending on the information being sought, narrowly constructed searches will be limited in scope and exclude valuable references that may contain the needed information.
If you compare PubMed and Google Scholar searches using the same keywords, notice that both engines return similar citations when searching broad topics but as the search terms become more specific, the engines return different, smaller citation lists. Often, there is a significant numerical gap between the number of citations returned, with Google Scholar returning the numerically higher amount. This is not only due to the larger database google searches but also the different algorithms used by the engines. PubMed uses weighted terms mapped to medical subject headings to rank order results. Google Scholar uses a combined weighted algorithm that places more weight on citation counts to rank results. While both engines return valid results, users should be aware that output lists are ranked differently and therefore require different refinement strategies to narrow down these results and find the appropriate information.
Once a reasonable list is generated, the next step is to begin screening the list to choose articles that contain information relevant to what you are trying to find. Many use the “Top Down” approach, reading only the titles first to see if they contain information relevant to pursue. If the title seems pertinent, the citation is moved to a “hold” list. Using this process, the original citation list may be reduced to contain 20-30 articles. The next screening step takes place by reading the abstracts. Obviously, a more time consuming step but once again, articles are selected or removed from the hold list based on how pertinent the information in the abstract is to your search. After screening the abstracts, the hold list may contain as few as 10-15 articles. If full-text is available, the articles are downloaded to be scrutinized more thoroughly.
Once a set of manuscripts has been collected, they are screened using Introduction, Methods, Results, and Discussion (IMRaD) as a guide seen in Blog #I in this series. There are a number of different approaches to reading an article that will depend on what specifically you are looking for. These approaches can be 1) skim specific sections first to check for the relevant information, 2) quickly read the entire paper to determine applicability, 3) examine the results section to see if data is applicable to address the question, 4) read the introduction to look for significance or background information, 5) examine the figures or tables, and 6) read the material and methods. Articles can be excluded or retained based on the information contained within and its value to the question being asked. At the end of the process, you will have a handful of articles containing the content that you are seeking. Finally, these articles are read in extreme detail in order to fully understand the experimental rationale and to extract the needed/sought after information. And yes, sometimes an article must be read two or three times to fully comprehend the subject matter.
It is important to remember that searching and reviewing the literature is a process and there is no single best approach to searching, filtering, screening and sorting. It can be done in a number of ways, each as unique as the individual performing the search. The same is true for the approach to reading articles and finding the needed information. Keep in mind, it is possible that the information you are trying to find does not exist or there may not be an answer to your question.
In the fourth post of this series; Cruising the Biomedical Scientific Literature: Managing References, learn ways to download and collect literature as well as tools to help build a personal reference library that you can use as a resource.
Written by Luis Chiriboga, PhD, HT(ASCP), QIHC