PhD / researcher
PhD / researcher
For literature studies at this level, the full methodology of systematic reviews is utilized. We recommend making the research question as precise as possible.
Be aware that the literature search is only a small part of the total methodology that should be used when conducting a systematic review.
In our examples, we present a sensible solution for a literature search tailored to the research question. We want to emphasize that there is never a definitive answer for literature searches, and that multiple solutions can be effective.
Example from Medline & Embase
Our starting point here is the following example project:
The purpose of the project is to find out how accurately frontline healthcare physicians can detect heart valve diseases using auscultation. A possible title could be:
Diagnostic accuracy of heart auscultation for detecting valve disease
Step 1: From research question to searchable concepts
A simple method to use when defining the major search terms for your project is to ask yourself the following question: which main concepts in my project title do I want to find in relevant publications?
The main concepts in our example are:
- Heart auscultation
- Diagnostic accuracy
- Detecting valve disease
To systematize these main concepts and find the correct scientific terms, we recommend setting the main concepts as headings above individual 'boxes'. This provides a clear overview of the individual search terms and helps you conduct the search in a systematic and structured manner.
Step 2: Find search terms for each main concept
You can now start the process of finding correct scientific search terms. There are many different methods to use. For example, you can ask a supervisor, a fellow student, look up terms in a medical dictionary, or ask your supervisor for relevant articles and look them up in the relevant database, where you can easily find the controlled vocabulary search terms with which these are indexed. If you are an experienced researcher with a good understanding of the subject's terminology, there is an effective method to quickly identify relevant references and then check the indexing of these references in relevant databases.
For instance, you can copy the title of the project and paste it into the Basic Search search box in, for example, Medline.
Basic Search in Medline allows you to search with natural language, much like in Google Scholar. As you can see, you get over 10,000 references. Medline will sort these references and place the most relevant ones at the top of the results list. It may therefore be worthwhile to go through the top 10-20 most relevant references to check how they are indexed. Here you will find relevant controlled vocabulary search terms and text words that can be used in your own search under the Advanced Search tab. By clicking on the Complete Reference button, to the right of a relevant article title, you are taken to the page showing the abstract and which Medical Subject Headings (MeSH terms) the particular reference is indexed with.
During the initial work of finding relevant search terms and gaining an overview of the terminology used for your chosen theme, you will almost always find that several descriptive terms can be used under each main concept. These are the first and second steps in the 5-step method of building and conducting a structured and systematic search, which always takes time and require a substantial amount of work.
In our example, we find the following words:
- Heart auscultation, Heart murmurs, Heart sounds
- Sensitivity and specificity, Observer variation
- Echocardiography
We now add these words to our 'boxes' and end up with the following setup for our search:
As a researcher/PhD student, you are expected to run your search in the most relevant reference databases. If you are unsure which databases are the most relevant, you can contact the University Library, or discuss this with your supervisor or your colleagues. For this specific project example, we recommend Ovid Medline and Ovid Embase Classic+Embase, as the most relevant databases. It would also be wise to check the Cochrane database for updated literature reviews.
We start our first search in Ovid Medline. The first thing we do in Medline is to look up all the search terms under each main concept in the controlled vocabulary (in this case Medical Subject Headings or MeSH). This is a very important step in the process because we should always use controlled vocabulary search terms for a main concept, where they exist. If you are unsure what a controlled vocabulary is and what advantages it offers, you should take another look at our page about controlled vocabulary search terms.
Here exemplified with the first search term in our first box, "Heart Auscultation". In the search window above, you will see that Keyword and Map Term to Subject Heading are checked. This is the default setup in Medline and ensures that Medline searches in the controlled vocabulary for the word we have typed into the search box.
In the search window below, you find that your search for "Heart Auscultation" found the controlled vocabulary search term (MeSH term) "Heart Auscultation".
You should always check the definition of controlled vocabulary search terms that you are considering adding to your search. You do this by clicking on the information symbol to the right of a controlled vocabulary search term. You find this under the heading Scope.
Here you read the objective definition of this controlled vocabulary search term.
NB. Under the heading Used For: in this window, you will always find an overview of synonyms that can be used as text words. You should also consider using these in your search. When carrying out a systematic search, you should always refine your search by including synonymous text words. Indexing of articles with controlled vocabulary search terms, such as MeSH terms, takes time. There may therefore be delays of months, or longer. In addition, errors can occur in the indexing. It is, then, always a good idea to build a literature search that also includes one or more text words, in addition to controlled vocabulary search terms.
To ensure that you include all controlled vocabulary search terms further down in the hierarchical structure of the controlled vocabulary, it is important that you check the Explode box (see our page about controlled vocabulary search terms).
We then click on Continue, and are taken to an overview of subheadings. Here, you can check Include All Subheadings, unless you want to focus on one or more of these.
Next, we click on Continue and see that we have found all references indexed in Medline with the controlled vocabulary search term "Heart Auscultation".
By clicking on the arrow symbol to the right of Search History, at the top left of the main menu, you can expand/collapse the search history. We see that "Heart Auscultation", exploded, gives us 11,505 references (exp Heart Auscultation/ 11505). We have now identified and searched with the controlled vocabulary search term (MeSH term) "Heart Auscultation". We note this in our box setup:
We then continue by checking if the terms under each of our main concepts exist as controlled vocabulary search terms. We use the same procedure for each of the search terms we have found so far.
Step 3: Build the first search
You have now found and search with controlled vocabulary search terms for all the main concepts in your project.
These can be found under Search History in Medline.
Some of you might think that the term "Heart Auscultation" naturally belongs to the main concept Detecting valve disease. The reason for placing this term under our first main concept (Heart auscultation) is that it will capture articles that address both of these main concepts, in the hope that structuring the search in this way will help us find sources about the effectiveness of one of these main concepts compared to the other.
You are now ready to combine the controlled vocabulary search terms and run the first search for this topic. Controlled vocabulary search terms corresponding to each main concept are combined with OR in the following way. Check each of the controlled vocabulary search terms you have added to the first box.
Then click on the Boolean operator OR, and you will see that these controlled vocabulary search terms are combined in line 7 of the search history.
Repeat this for the controlled vocabulary search terms you have found for the next two boxes. One box at a time! You will then find the combined search terms for the three boxes, in lines 8, 7, and 6 respectively (which only has one controlled vocabulary search term).
You are now ready to combine these three separate searches with AND. Select lines 8, 7 and 6. Then click on the Boolean operator AND. You now see that your first search with controlled vocabulary search terms yields 125 references (line 9).
Step 4: Improve the search
The search we have carried out so far used exclusively controlled vocabulary search terms for each main concept in the project title. You can improve the search by using some synonymous text words in what we call a text word search. Indexing articles with controlled vocabulary search terms, such as MeSH terms, takes time. There may therefore be delays of months, or longer. Minor errors can also occur in the indexing. This means that it is always a good idea to build a literature search that includes one or more synonymous text words, in addition to controlled vocabulary search terms.
Let us return to our example:
Where can you find text words for these controlled vocabulary search terms?
One of the most important tools you have is actually to immerse yourself in the terminology of the specific scientific field you are working in. As a researcher/PhD student, you probably already have a relatively good overview of the terminology, but remember that databases index references from all over the world, and academic communities from different parts of the world may use synonymous terms for the controlled vocabulary search terms we have now found. A very useful tool is therefore to look up the individual controlled vocabulary search terms you have found in the controlled vocabulary. In Medline, this is the MeSH database, which we have previously used.
From the previous example above, you remember that each controlled vocabulary search term has a precise explanation, here exemplified with "Heart Auscultation". Below the heading Used For you will find synonymous text words for the controlled vocabulary search term.
As a researcher/PhD student, we recommend that you use as many text words as possible. This increases the sensitivity of your literature search and will ensure that you find as much of the relevant literature as possible for your project. Even if you are not conducting a systematic literature review, a structured and systematic search will give you a unique overview of current literature within your field.
Our first synonymous text word in our example is "cardiac auscultation*".
Text word searches are done in specific search fields. The most common search fields we use in Medline are: Title, Abstract, and Keyword Heading (authors' own keywords). We note this in our box setup, where we mark the three fields we will search in, with ti (Title), ab (Abstract), and kw (Keyword Heading). The search syntax for the text word, then, is as follows: cardiac auscultation*.ti,ab,kw. (Note the use of the full-stop and comma, and that no spaces are used.)
We use the same method for all the controlled vocabulary search terms. We should also use other text words we have identified from other sources. We use the same syntax as described above.
By working in a structured and systematic manner, as demonstrated, we end up with the following box setup for our search.
In the final search setup for this project, we have used most of the text words we found by looking up the individual controlled vocabulary search terms in the controlled vocabulary. Additionally, we have used further text words that we found in the descriptions of controlled vocabulary search terms that are lower down in the hierarchy of the controlled vocabulary. This is exemplified in the screenshot below, where you see the hierarchical structure of "Echocardiography". Each of the narrower controlled vocabulary search terms will list one or more text words that can be used in title/abstract/keyword searches. Here, it is up to you to assess how many, and which specific text words you want to include. The more synonymous text words in the search, the more sensitive it becomes. To further increase sensitivity, we can also use proximity searches, using the operator ADJ.
What remains is to execute this search in Medline. A good tip here is to proceed in a structured and logical manner. This means that you start with the search terms in the box to the left and search with each search term separately.
We start by searching with "heart auscultation". Then we click on Search Fields (to the left of Advanced Search). Here we see an overview of all the search fields that are used to index references. We also see that All Fields is checked.
We uncheck All Fields and enter our first synonymous text word (heart auscultation*), before selecting Search Fields: Title, Abstract and Keyword Heading. Then we click Search.
We then see that in Medline there are 545 references that have the text word "heart auscultation*", in the title and/or abstract and/or authors' keywords.
A practical tip is to conduct this Search Fields search in the Advanced Search screen. This way, you avoid having to select the various fields in the Search Fields window. You have now learned that the field codes for the three search fields Title, Abstract, and Keyword Heading are respectively: ti, ab and kw. You can enter these directly in the Advanced Search search box, using this format: heart auscultation*.ti,ab,kw. (Note the use of the full-stop and comma, and that no spaces are used.)
When you click on the Search button, you see that the result (line 3 in the search history) is the same as you got by using Search Fields (line 2 in the search history).
We now use the box setup we have developed to enter and search with one search term at a time. This provides a structured and systematic search history. In the screenshot below, you see all the search terms retrieved from our first main concept, or first 'box'.
We are now searching with all the search terms for our three main concepts:
- Box 1: lines 1–23
- Box 2: lines 24–43
- Box 3: lines 44–60
We are now ready to combine our improved search, by doing exactly the same as in step 3. We first select all the search terms from the first 'box' (Heart auscultation). Then we click on the Boolean operator OR.
We repeat this for the search terms for the next two boxes. We now see that we have 3 lines (61, 62, and 63), where search terms for each of the three main concepts are combined with OR. We can now conduct our extended literature search by combining these three lines with AND.
The result of this search now gives us 708 references, compared to the 125 references from our first literature search. In this way, we have conducted a more sensitive search in our chosen database. As a PhD student or researcher, you will need to evaluate the scope and relevance of the references you have found, together with your supervisor or colleagues. If you believe the relevance is too poor, you can conduct a more specific search. In this case, you would use the same setup, but exclude the field code for Abstract in your text word searches. This means that the first text word search that we have used in this example (heart auscultation*.ti,ab,kw) would look like this: heart auscultation*.ti,kw. Here, we have omitted searching in the abstract field, a field that can add some 'noise' in the form of irrelevant literature. When we remove the Abstract search field from our search setup, we end up with 396 references. In this way, we have carried out a more precise search.
Step 5: Adapt the search to other databases
For many PhD projects or systematic literature reviews within health science disciplines, the databases Medline, Embase, PsycINFO (on Ovid's interface), CINAHL, Web of Science and Cochrane will be relevant databases. If you are unsure which databases to choose, you should talk to your supervisor or your colleagues, or contact the University Library. A single database is rarely sufficient to search at this level.
The interface of the mentioned databases can feel very different, but the 5-step method as explained above can be used in any of these reference databases.
We will now convert our example search from Medline to Embase Classic+Embase. From our example, you now know that Medline uses MeSH terms to index articles, so that we can easily find relevant references. Embase, which is also uses Ovid's interface, uses a different controlled vocabulary search vocabulary, called Emtree. When moving from one database to another, therefore, we need to look up the controlled vocabulary search terms used in the first database in the controlled vocabulary of the next database. In our example, we need to find Emtree terms that correspond to the MeSH terms we used in the Medline search.
When you now conduct this search in Embase, you must check that each controlled vocabulary search term has a corresponding controlled vocabulary search term in Embase (that is, if there is a corresponding Emtree term). If you have already conducted your search in Medline and saved it, you can now click on the Change button to the right of 1 resource selected.
You will then get a new window where you select Embase Classic+Embase. Then you get the choice of clicking on Continue or Run Search. We recommend that you click on Continue. The reason for this is that Run Search will conduct your Medline search in Embase, without giving you the chance to check whether the controlled vocabulary search terms you used in Medline exist as controlled vocabulary search terms in Embase. By clicking on Continue, you start a new search in Embase, and you will then have full control over each step in the search process, as described above.
To illustrate that there can be differences between the naming (and level in the hierarchical structure) of controlled vocabulary search terms, let us look at the controlled vocabulary search term "Echocardiography", which we found as a MeSH term in Medline. Below you see where this controlled vocabulary search term is positioned in the MeSH hierarchical structure.
When we look up the controlled vocabulary search term in Embase Classic+Embase, we find a somewhat more complex structure, with more sublevels than we did in Medline.
To ensure that the definition of the controlled vocabulary search term "Echocardiography" in Embase matches the definition we found for the MeSH term in Medline, we can click on the information icon under Scope.
Here we read that the Emtree term "Echocardiography" is defined in exactly the same way as the MeSH term "Echocardiography" (remember to consider whether to select Explode or not). Additionally, this Emtree term has a longer list of synonymous text words, which you find under Used For. Consider including these in your search as text words. Just as in Medline, you can search with text words in the Title, Abstract, and Keyword Heading fields in Embase, using the field codes: ti,ab,kw.
If you find further text words that you have not used in previous searches when you, as in this example, move from Medline to Embase, you should consider including these in your original search setup, and conduct a new search in Medline. We do not demonstrate this in our example, but this is something you need to be aware of. The point is that when transferring a search from one database to another, the searches should be as similar as possible, notwithstanding differences in terms of controlled vocabulary and search syntax.
Example from PsycINFO, ERIC & Web of Science
We base this on the following example of a project:
Interventions targeting school belonging in secondary education: A systematic meta-analytic review
Step 1: From research question to searchable terms
Before you start identifying search terms, it is important to break down your research question into its main concepts. Ask yourself the following question: Which concepts in the research question must be mentioned in a source for that source to be considered relevant?
In this example, we can identify the following main elements:
- School belonging
- Secondary education
- Interventions/Intervention studies
Note that sometimes, based on what we learn throughout the process, we may need to take a step back and reconsider which concepts are the most suitable to build a literature search around. In this example, we could proceed with the concepts we initially identified.
To systematize these main concepts and get ready to find search terms, we recommend setting the main concepts as headings in separate ‘boxes’. This provides a good overview of the individual search terms and helps you conduct the search in a structured way.
Step 2: Find search terms for each main concept
We can now begin the work of finding suitable search terms. Since the search tools and research literature will primarily be in English, we need to have search keywords in English.
There are many methods for identifying good search terms. An obvious first step would be to reflect on this yourself and consult your supervisor and knowledgeable colleagues, or perhaps a chatbot. We can also use dictionaries (e.g. ordnett.no) or online translation services to find good translations of terms from one language to another.
In our example, we will initially use this most obvious method: noting down the keywords that come to mind based on our somewhat limited prior exposure to the relevant literature. We end up with...
The second method for finding search terms mentioned in the general description of the 5-step method is to use the controlled vocabulary interface of a particular database. This is important and useful for at least two reasons: First, we can identify good controlled vocabulary search terms, if they exist. We should always use controlled vocabulary search terms whenever possible. Second, we can also become aware of additional synonyms (often listed as Entry Terms or Used For in the explanations of the controlled vocabulary terms) for the terms we already have.
Our research question lies at the intersection of educational research and psychology. We could choose a database like ERIC (Educational Resources Information Centre) to start with, but in our specific example, PsycINFO may provide equally good coverage. The choice therefore falls on PsycINFO, as it offers a much better search interface and functionality (on the Ovid platform) than the alternatives.
If you are unsure which databases are most relevant for your own project, you can contact the University Library or discuss it with your supervisor or colleagues.
We start with the concept that is the most distinctive thematic part of this research question, namely School belonging. We first try to see if the controlled vocabulary in PsycINFO (whose official name is Thesaurus of Psychological Index Terms) contains controlled vocabulary search terms corresponding to this concept.
We navigate to PsycINFO from the database overview in Oria. Here, we select the Advanced Search tab and ensure that Map Term to Subject Heading is checked. We test an obvious candidate for our first concept (School belonging) in the search box:
When we now click Search, we are not directly searching the database of publications but are asking to 'map' or link what we have entered in the search box to potentially useful terms from the controlled vocabulary. The mapping algorithm now provides us with a list of suggestions:
Sometimes the suggestions are few and relevant, while at other times there are no reasonable suggestions. In this case, there are quite a few suggestions, but only a few are relevant. "Belonging" is the only one we consider worth exploring further here.
To learn more about this controlled vocabulary search term, we click on the hyperlinked word "Belonging" in the list of suggestions. This gives us the following view:
We see from this that the controlled term "Belonging" is used for, among other things, "Sense of Belonging", which was also one of the candidates in our own, initial search term suggestions. On this screen, we can also check whether the search term we are investigating has broader terms and/or narrower terms in the subject heading hierarchy. It is good practice to always check this. Furthermore, we see that there is a column on the right labeled Scope Note. If we click on the blue circle with the i, we get a bit more information about the keyword. Again, it is good practice to always do this for terms we are considering using in our search. If we do this here, we see the following:
Here, we often find a definition and, sometimes, other important information about how the search term is used to index articles in the database. The definition here tells us that the controlled vocabulary search term "Belonging" (and its synonyms "Sense of Belonging" and "Belongingness") is defined relatively broadly and without reference to anything school-related. Nevertheless, we decide, with some hesitation, to use this term in our search. We navigate back in the browser to the previous screen, check the box for the search term to select it, and then click Continue.
We are then taken back to the Advanced Search screen. We have now retrieved the records for all 5,219 documents that are tagged with the controlled vocabulary search term "Belonging/" in PsycINFO. That list begins immediately below the search box (though we do not see it in the screenshots here). We can now also expand the search history by clicking on the small arrow next to Search History above the search box. We should always have the search history visible when working with searches in reference databases. It is a crucial tool for maintaining an overview and building a clear and well-structured search.
We have now conducted an initial search using a single controlled vocabulary search term that we think might fit our first main concept, School belonging. We have also noted—from the overview within the thesaurus—that this term has the synonyms "Sense of Belonging" and "Belongingness". At the same time, we have discovered that there is no controlled vocabulary search term specifically for school-related feelings of belonging.
We now repeat this process for the other two main concept. We will not show every single step here, but will skip to the stage where we have noted both relevant controlled vocabulary search terms and some additional possible synonyms. At this point, our notes look like this:
The third method for finding good search terms is a very useful one. It involves using articles we know are relevant, if we have them or can access them. Perhaps you already have a couple from previous, incidental contact with the literature. Perhaps your supervisor has recommended some articles. Or maybe you have done some simple, intuitive searches yourself using simpler tools like Google Scholar, Oria, Keenious, or the Basic Search tab in the Ovid databases. We can then look up these articles in the database we are searching in and see which controlled vocabulary search terms they have been tagged with.
This method can be used in several steps of our 5-step method. In this example, we will save it for Step 4: Improving the search.
We now have enough search terms noted down to make a first attempt at building a decent search.
Step 3: Build the first search
We now have a collection of search terms, primarily based on our own somewhat naive intuition and our exploration of suggestions from the mapping algorithm of the controlled vocabulary in PsycINFO. This has resulted in the following setup:
We will now search with these terms in PsycINFO, one main concept at a time and one search term at a time. Note! A common mistake is to create overly complex combinations of search terms too early. To maintain a clear overview, it is important to proceed step by step.
We enter the controlled vocabulary terms as we demonstrated for "Belonging/" in Step 2. There are several ways to enter the non-controlled terms (text words). Here, we show one common method using the text word "school connectedness".
In the advanced tab, we type the search term into the search box. But before we click Search, we add what are called field codes, like this: .ti,ab,id. These stand for title, abstract, and the authors' own keywords (which in PsycINFO are misleadingly called 'Key concepts' and, inexplicably, have the field code id). These three fields are the typical choice for free (non-controlled) search terms, or text words. Note that the codes for the same metadata fields may differ in other databases.
(By selecting the Search Fields tab, you can see an overview of all the searchable fields in the database and their field codes. This can be useful for a variety of slightly more specialized searches.)
Note that the search history already has four lines, one for each of the previous keywords in the School belonging box from our notes.
When we now click Search, we skip the mapping phase we used in Step 2. (The interface does this automatically when the search box contains a field code or an operator.)
The search history is now extended with a fifth line. (The first line then disappears upward, and we need to click Expand in the lower right corner of the search history to view it in its entirety.) Now we can combine the search terms we currently have for our first main concept by checking the boxes next to each line and clicking the OR button to the right of Combine with:.
We repeat all of this for the other two main concepts. For each main concept, we first enter controlled vocabulary search terms using Map term to subject heading, then text words using field codes. We end up with the following search history:
Notice how we have grouped the search terms for each main concept into 'clusters' or blocks, which are first combined with OR individually (lines 6, 14, and 19), corresponding to the box structure we outlined in Step 1. Finally, we combined the result of the OR-seaches for the three main concepts with AND to find what lies at their intersection (line 20).
This is a decent first attempt, put together with relatively little effort and research. The number of hits is relatively modest. However, when we skim through the results list to get an impression of how well the search performs, relevant results are few and far between. This indicates that there is significant room for improvement.
Step 4: Improve the search
In Step 2 of this example, we used two different methods to find suitable search terms. We relied on our own somewhat naive intuition, and we used the database's tools to find, understand, and use controlled vocabulary search terms.
The third method for finding search terms is perhaps the smartest of all: using articles we already have, and that we know are exactly the kind of articles we want to capture with our search. This method, of course, requires that we actually have such articles.
Most people embarking on a systematic literature review have at least some familiarity with the literature they aim to summarize and therefore usually have a small handful of thematically relevant articles. Alternatively, we can take advantage of more 'helpful' search tools, where machine learning, natural language processing, and relevance ranking boost naive searches and provide us with thematically relevant results. Examples of such tools include Google Scholar and Keenious. The point is that we should have a small handful of relevant articles that we know are the kind we want to capture, as we can then use them to improve our search.
In this example, we have a review article with roughly the same research question as our own, from a few years back. (Our ambition is to create a better literature review that includes even newer research studies and supplements the review with a meta-analysis.) In the reference list of the slightly older review article, we find several articles that could potentially be included in our own study. We can now use these to see how well our search performs.
We do this by searching for some of these articles individually in the database where we built our initial search, and then checking whether our search captures each of them. If it does, then we have built a search that is sensitive enough. If not, we can check the indexing of the specific article to try to understand why our search did not capture it and what might reasonably be changed in our search to ensure it does capture that specific article.
To search for a single article, we can select a segment from the title that we think constitutes a fairly unique combination of words and match this against the database's title field. We will now demonstrate how to do this for two articles we consider to be the kind we want to find with our search.
Here are the references for the two relevant articles:
- Chapman, R. L., Buckley, L., Sheehan, M., & Shochet, I. M. (2013). Pilot evaluation of an adolescent risk and injury prevention programme incorporating curriculum and school connectedness components. Health Education Research, 28(4), 612-625.
- Frank, J. L., Kohler, K., Peal, A., & Bose, B. (2017). Effectiveness of a school-based yoga program on adolescent mental health and school performance: Findings from a randomized controlled trial. Mindfulness, 8(3), 544-553. https://doi.org/10.1007/s12671-016-0628-3
Below our search, we now enter a part of the title from each of them and match it against the title field. In the screenshot below, we have already performed the title search for the Chapman article (it now appears in line 21 of the search history). In the search box, a title string for the Frank article is ready.
Note that for the first article, we only include the first part of the title. This is because if we include "and" (which is the next word in the title), the interface interprets it as an operator, which can cause some issues. Also, note that for the second article, we use the publication year to isolate it, as there were two articles with the same title segment in the database. The lower part of the search history now looks like this:
Note that if we get 0 results from such a title segment search, it means that the article we searched for is not indexed in the relevant database—assuming, of course, that we searched correctly.
Now we can use the AND operator to check whether the two articles are included in our keyword search. Like this:
From this screenshot, we can see that the first of the two articles is captured (line 23), while the second is not (line 24). Since we believe that this article should also have been captured by our search, we now need to figure out why it is not. We can do this by clicking Display Results on line 22, so the article appears in the results list:
Then, click on Complete Reference for this article in the results list to view the full record and all the metadata registered for it.
This gives us the following view:
As we can see, there is a lot of metadata here. And yet, we are not seeing everything! A characteristic of this type of database is that they have extensive metadata schemas and very high-quality assurance for the values in all the various fields.
By looking through the record, we see that none of the search terms we used to capture the School belonging concept are represented in the metadata for this article - neither the controlled vocabulary search term ("Belonging/"), nor any of the others. However, we do see that a near-synonym, "school engagement", appears both in the Abstract field and in a field called Tests & Measures.
This is important information. If we still believe that this article is the kind of article we want to capture with our search, this may mean we need to be more generous with synonyms and near-synonyms to capture articles like this one. By checking a few more articles, it also becomes clear to us that the general search terms for belonging—"Belonging/", "sense of belonging", and "belongingness"—rarely appear in the literature on school belonging.
For the other two concepts, there is a wide variety of search terms associated with the metadata of the articles we check. We note all of them and use them to improve the search.
We cannot show the checking process for all the articles here, but must settle for illustrating the technique and thought process as we just did. After more research along these lines, we eventually arrive at an improved search. It looks like this:
Note that we have removed the general belonging search terms from the first part of the search and are now focusing exclusively on synonyms for "school belonging". We have gathered all of them into a slightly open proximity search string in line 1.
Also, note that we are taking advantage of a unique strength of the PsycINFO database, namely the Tests & Measures field, by adding the field code tm to the other three.
Additionally, we have added quite a few search terms associated with the other two main concepts of our research question.
This results in a search that is both more precise (less noise from belonging that is not school-related) and much more sensitive (more synonyms for School belonging, many more synonyms for educational levels, and many more search terms associated with intervention studies). Our number of results becomes four to five times larger than in the first attempt.
Step 5: Adapt the search to other databases
We are now ready to adapt the search to other databases. Because other databases will have different controlled vocabularies, different searchable metadata fields, interfaces, and operator syntax, we cannot simply copy and paste. In this example, we will briefly demonstrate how our search has been adapted to the educational research database ERIC and the interdisciplinary database Web of Science. We begin with ERIC.
ERIC (Education Resources Information Center) is available at UiT through the EBSCO interface. In the following, we will briefly show how to find and search with a controlled vocabulary search term and a non-controlled text word. We will then look at the fully adapted search and point out some important differences.
In ERIC, we need to select Thesaurus in the blue main menu at the top to access the controlled vocabulary.
Here, we need to make sure that we use the lower search box, as this is the one that provides suggestions from the controlled vocabulary. In this case, we have entered "intervention" because we want to see if we can find a controlled vocabulary search term that corresponds to this main concept. As we can see, several suggestions appear, with the first one being the most relevant. We click on that.
Similar to the Ovid interface, we then get a bit more information about how this term is used:
After ensuring that this is a controlled vocabulary search term we want to use in our search, we check the box to the left of it and click the Add button. The search term is now transferred to the top search box with syntax indicating that we are searching the metadata field where the controlled vocabulary search terms are listed (field code DE in ERIC). We can now click the green Search button next to the top search box to actually perform the search.
To search for text words (non-controlled search terms), it is easiest to use the search boxes on the main page. (We can always return there by clicking the Advanced Search link or the EBSCOhost logo.) In ERIC, we can also choose different metadata fields, but we cannot search multiple field codes simultaneously without repeating the keyword or string multiple times. Fortunately, the field code TX (All Text) serves roughly the same purpose as searching in title, abstract, and keywords. Therefore, we often choose this when searching for free-text keywords in ERIC. For example, it might look like this for our School belonging string:
The string is the same as the one we used in PsycINFO, but we have adapted it by choosing the corresponding syntax for the proximity operator ("N3" instead of "adj3") and a field code as explained above. We need to click on the Search History link just below the search boxes to expand the search history.
The fully adapted search looks like this:
| Line | Keywords/String | Hits |
| S1 | DE "Student School Relationship" | 5535 |
| S2 | TX (school N3 (belonging* or engagement or bond* or connect* or membership)) | 9414 |
| S3 | S1 OR S2 | 14194 |
| S4 | DE "Intervention" | 58001 |
| S5 | DE "Program Effectiveness" | 74350 |
| S6 | DE "Program Evaluation" | 56981 |
| S7 | TX ("school based" or intervention or program*) | 718724 |
| S8 | TX (trial or assigne* or random* or cluster or crossover or experimental or quasi or comparison or control or matched or "propensity score") | 300951 |
| S9 | AB ("wait list" or treatment) | 48748 |
| S10 | S4 OR S5 OR S6 OR S7 OR S8 OR S9 | 921027 |
| S11 | EL (grade N1 (7 or 8 or 9 or 10)) | 22890 |
| S12 | EL (middle or high or secondary) | 193747 |
| S13 | S11 OR S12 | 197225 |
| S14 | S3 AND S10 AND S13 | 2293 |
Here it is presented in a formatted table, as the search histories in the EBSCO interface become incredibly long and unmanageable as screenshots. We can note the following important adaptations:
Unlike in PsycINFO, we have found a controlled vocabulary search term that corresponds to school belonging, namely DE "Student School Relationship".
We have also found controlled vocabulary search terms for the other main concepts, but since the controlled vocabulary is not the same, they are not identical to those used in the PsycINFO search.
The section for educational level is more compact here. This is because we are using a unique strength of the ERIC database, namely that it has its own metadata field for educational level (field code EL).
In Web of Science, the differences become even greater because this database does not have a controlled vocabulary. We are therefore limited to using text words exclusively, rather than other metadata fields. For most purposes, selecting Topic in the dropdown menu for the search field (field code TS= if using the Advanced search interface) works well. This includes title, abstract, authors' keywords, and Web of Science's Keywords Plus. For our School belonging string, it might look like this:
Again, the string is the same, but now with yet another new syntax for the proximity operator.
To access the search history and the ability to combine lines within it, you need to select Advanced Search. On that screen, you will also find an overview of all metadata fields and field codes.
The original search, in its Web of Science-adapted version, looks like this:
Adapting a search in this manner often requires us to consult the databases' search help to figure out field codes, syntax, and other details.