4

Years ago, I seen an interesting article that made reference to database search results. The author run a few queries through PubMed and reported on the number of records. This was a neat introduction to the wider discussion on the state of the scholarship in the given field. I'm interested in doing something similar. I would appreciate if someone could point me to out to papers or notes on how such findings should be reported. For example let's say that there is x amount of articles on Term A and y amount of articles on Term B. Consequently, the one could assert something on the lines:

Contrary to common expectations, it seems that relatively little is published on Term A in relation to the Term B. For example, both terms appear only on n occasions in the Database D.

I'm interested in how to state such results in a non-confusing manner, in particular I would like to know:

  • Where and how would it be appropriate to search the nature of the query
  • Presumably, the date when the query was applied should be stated as well?
  • Should the reference be introduced to the bibliography? If so I would be grateful for sample entry in APA or any other common style
4

This sort of thing is pretty standard when it comes to meta-analyses, since each begins with a search for relevant studies. Generally, descriptions of the search query and its results are included in the methodology section of a paper; in addition to including the keywords used in the search (and, preferably, the reasoning for their inclusion), the start and end dates of the search should, as you mentioned, be noted. A diagram may also be helpful for readers to indicate the way in which you selected studies for analysis from the initial pool of results, and the languages in which the studies you assessed were written.

There are many papers that demonstrate this; as an example, "Psychosocial factors and chronic spontaneous urticaria: a systematic review" published in Allergy states the following:

We first conducted a search of the PubMed and OVID/Medline databases using the keywords ‘urticaria’, ‘chronic urticaria’, ‘chronic spontaneous urticaria’ – given that the definition of CSU was not clearly established in early studies, we also included more general terminology such as urticaria and chronic urticaria, but excluded articles clearly assessing physical or acute urticaria – ‘psychopathology’, ‘stress’, ‘depression’, ‘anxiety’ ‘life events’ and ‘axis I’ and ‘axis II’, including full-text accessible articles in English, French and Spanish. We then performed a meta-analysis that included all these studies (see Fig. 1) from 1 January 1935 to 1 January 2012. After two reviewers (M.B.S. and I.B.) independently evaluated all potentially relevant studies, we conducted statistical analyses using Stata® version 12 (StataCorp LP, College Station, TX, USA). Methodological quality of included studies In order to assess the quality of the aforementioned studies, we employed a standardized measure specifically tailored to this systematic review, based on the Newcastle-Ottawa Scale (NOS) (8). This approach included the appraisal of external and internal validity, as well as biases common to observational studies specific to CSU and psychosocial factors. The independent reviewers mentioned above evaluated the study quality separately and resolved the differences in opinion by consulting a third reviewer (A.R.) (Table 1).

To sum up:

  1. Databases searched
  2. Search terms used (potentially alongside their justifications)
  3. Dates of publication included in search
  4. A diagram which explains how you winnowed these studies down to the relevant ones (and a table for the ones you decided to include in your paper)
  5. As a bonus, if multiple authors are responsible for assessing studies for inclusion in your paper, include how these decisions were made/conflicts resolved.

I've bolded some of these ideas in the paragraph above to give yo a good idea of how this works.

  • Thanks, is it in a good practice to report the number of returned results? I'm particularly interested in highlighting that certain terms occur more/less often in publications than other. – Konrad Apr 27 '15 at 6:31
  • Yes, in that case you would certainly want to report the number of returned results. – mhwombat Apr 27 '15 at 17:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.