Using PubMed – a practical example

For this exercise, the sample research question is ‘What are the views of parents and carers on administering factor replacement therapy?’.

This question contains several present or implied concepts:

  • parents/carers
  • children with haemophilia
  • treatment of haemophilia
  • clotting factors
  • drug administration
  • parental satisfaction with treatment.

Go to PubMed.

In the box top left, click on the arrow and select ‘MeSH’.

In the long box to the right, type ‘parent’. This will give you a list of nine MeSH headings. Of these, the first (’Parents’) seems to be the best. Others, like ‘parent-child relations’, are possible but imply other issues not immediately relevant to the question. It’s worth noting them, just in case.

Now search MeSH headings to find the other most relevant key words:

Key word MeSH heading
parent Parents
carers Caregivers
children Child

Child, preschool
haemophilia Hemophilia A

Hemophilia B
clotting factors Blood coagulation factors
drug administration Drug administration schedule

Drug administration routes
satisfaction Personal satisfaction

Patient satisfaction

Some MeSH headings seem to be exactly what we need but two do not completely cover the concepts in our question. The MeSH headings for drug administration included several in the form of ‘Administration, intra….’ but there was no mention of IV administration, which is what we need for factors.

Search MeSH headings again using the word ‘intravenous’.

There are several that might fit the bill: ‘infusions, intravenous’, ‘administration, intravenous’ and ‘injections, intravenous’. These terms also seem to cover ‘drug administration routes’ so we can now exclude this term.

We might say that factors are given by infusion, not injection, and therefore one heading fits best. This is true but it’s a good strategy to consider the risks of being too literal when interpreting these terms: indexers have guidelines about which terms to use but they may not always be right, or the article may not be specific enough to warrant one heading rather than another.

The easiest solution in this case is to use all three and group them (using brackets) with the Boolean operator ‘OR’. (infusions, intravenous OR administration, intravenous OR injections, intravenous).

The most efficient solution is to check the category trees that contain the headings. Clicking on each of the three headings brings up a record that includes the tree containing the indexed term:

Administration, intravenous:

All MeSH Categories

Analytical, Diagnostic and Therapeutic Techniques and Equipment Category

Therapeutics

Drug Therapy

Drug Administration Routes

Administration, Intravenous

Infusions, Intravenous

Injections, Intravenous

 

Infusions, intravenous:

All MeSH Categories

Analytical, Diagnostic and Therapeutic Techniques and Equipment Category

Therapeutics

Drug Therapy

Drug Administration Routes

Administration, Intravenous

Infusions, Intravenous

 

Injections, intravenous:

All MeSH Categories

Analytical, Diagnostic and Therapeutic Techniques and Equipment Category

Therapeutics

Drug Therapy

Drug Administration Routes

Administration, Intravenous

Injections, Intravenous

From this we can see that the heading ‘Administration, intravenous’ includes the two others below it in the tree. We only need to use the higher term; PubMed automatically searches the lower terms (a process appealingly known as ‘automatic explosion’).

Repeating this exercise for ‘patient satisfaction’ reveals no link with ‘personal satisfaction’ and a closer look at the latter term shows that it comes under the ‘behavior’ subheading within psychiatry and psychology. This can be put aside with the proviso that it can always be recalled if our search does not retrieve many articles.

If we check the tree in which ‘Caregivers’ occurs, we find that this heading includes the caring role of parents. We should therefore be able to omit parents from our search strategy. However, bearing in mind the need to interpret the rules cautiously, we should keep ‘parents’ and be prepared to exclude it if we need to reduce the number of articles retrieved.

Our MeSH headings are therefore:

Parents

Caregivers

Child

Child, preschool

Hemophilia A

Hemophilia B

Blood coagulation factors

Administration, intravenous

Patient satisfaction

Some of these are effectively synonyms and need to be grouped by ‘OR’:

(Parents OR caregivers)

(child OR child, preschool)

(hemophilia A OR hemophilia B)

Now they can all be linked with ‘AND’ in a PubMed search:

(Parents OR caregivers) AND (child OR child, preschool) AND (hemophilia A OR hemophilia B) AND Blood coagulation factors AND Administration, intravenous AND Patient satisfaction

Unfortunately, if you try this you will find that no articles are retrieved!

Go back to the PubMed page and reselect the PubMed database within the box, screen top left. Then type in the search terms sequentially, linking them with AND (remember to include brackets). The number of articles retrieved decreases with each additional term:

Search term + AND Number of articles
Parents OR caregivers 248,486
child OR child, preschool 137,520
hemophilia A OR hemophilia B 226
Blood coagulation factors 64
Administration, intravenous 5
Patient satisfaction 0

The search strategy now needs to be reconsidered to decide which of the headings are least important, or might be implied by the other headings. Try making the following changes to the search strategy.

  • ‘administration, intravenous’ is redundant: factors can’t be given any other way so we can afford to omit it. Doing so retrieves 2 articles.
  • ‘child’ and ‘child, preschool’ are implicit when we look for articles connected with parents and carers. Removing these terms increases our list to 6.
  • Could we be less specific about treatment? We can substitute ‘blood coagulation factors’ with the subheading ‘drug therapy’ for haemophilia. The syntax for this is ‘hemophilia A/drug therapy’. This only retrieves 7 articles; two of them were not in our previous list of 6, though they seem less relevant.
  • Can we exclude treatment altogether and look at patient satisfaction with haemophilia generally? Removing the subheadings increases the list to 16, including 9 not previously found. However, the articles retrieved are becoming less specific to our research question.

This isn’t what we expected to find on a topic as popular as this, so the search strategy needs to be reconsidered completely. Time for a shortcut. If you made the above changes you’ll now have a list of 16 articles, one of which is

Patient and parent preferences for haemophilia A treatments. Mohamed AF, Epstein JD, Li-McLeod JM. Haemophilia. 2011 Mar;17(2):209-14.

Click on this to open the abstract, then go to the bottom of the page and click on ‘MeSH Terms’:

This brings up a list of the MeSH headings under which this article was indexed.

Publication type, MeSH terms, Substance

Publication type

MeSH terms

Substance

The asterisk next to three of the headings denotes a ‘Major Topic’ – that is, what the indexer considers to be a focus of the article. There are several headings here that were not in our search strategy.

Go back to the list of 7 articles and repeat these steps for the article

Treatment-related knowledge and skills of patients with haemophilia and their informal caregivers. Novais T, Duclos A, Varin R, Lopez I, Chamouard V. Int J Clin Pharm. 2016 Feb;38(1):61-9.

This provides an extensive list of MeSH headings:

Publication types, MeSH terms, Substance

Publication types

MeSH terms

Substance

Again, there are several here that we did not originally include. They include age groups (adolescent, young adult, infant), factor therapy (Factor VIII, coagulants, medication therapy management), outcomes (health knowledge, attitudes and practice) and subheadings (psychology). This shows that, although we originally selected headings that closely matched our question, we should take into account the flexibility and ambiguity in the way PubMed indexes articles. These terms can now be used to develop a new search strategy. Try several options and explore how these changes affect the number and nature of articles retrieved.

Another option – one that is additional to, not a substitute for amending our search strategy – is to use the ‘Similar articles’ tag. Provided there is at least one article that closely corresponds to what we’re looking for, we can use this facility to find others that are indexed with many of the same headings.

Find the article ‘Patient and parent preferences for haemophilia A treatments. Mohamed AF, Epstein JD, Li-McLeod JM. Haemophilia. 2011 Mar;17(2):209-14’ from our earlier list of two. Underneath, you will see the link to similar articles. Click on it.

This creates a list of 93 articles, with those with closest resemblance of index terms coming first. Read through the titles on this list, clicking in the square box on the left of each one that appears to be relevant (the clicks remain when you change pages). When you’ve finished, go to the Format box in the top left of the screen and select abstract. You will now have a new list of potentially relevant articles and their abstracts which you can read through to determine which will be useful.

So what has this exercise illustrated?

  • A search strategy using MeSH headings needs to take account of how terms are indexed
  • Too many headings in a search excludes an excessive number of articles
  • Some MeSH headings can be used to cover the concepts of others
  • A systematic approach seems efficient but risks missing some articles
  • Using the ‘similar articles’ tag can identify new articles but is time-consuming
  • Using one strategy and one database is unlikely to find all the articles relevant to the search question.

These shortcomings are not unique to PubMed but are common to all large databases. They show that multiple databases must be accessed if a literature search is to be truly comprehensive. It is therefore very useful to work with an information specialist to formulate and carry out the search.

Incidentally, a free text search with the phrase ‘parents AND concern AND hemophilia AND treatment’ retrieves only 7 articles, none of which is relevant.