Public Health Matters

Research: Crafting a Search Strategy and Compiling Seed Lists

Shiying He -

Crafting a search strategy

In any research project, you need to justify every step of your methodology. This applies to the literature review stage as well. Hence, it is important to document your search strategy and justify it.

A search strategy requires the following: 1

  • Databases searched. MEDLINE, PubMed, Google Scholar, grey literature, bibliographies of retrieved articles?
  • The review question. what question are you trying to answer?
  • Timeline of publication. From year 20XX to 20XX? Will articles published before 20XX be excluded?
  • Languages included. English? French?
  • Types of studies included. Randomised controlled trials? Observational studies? Qualitative studies?
  • Search terms. More applicable for systematic or scoping reviews.

Here is an an example taken from PROSPERO:

We will search MEDLINE, PubMed (articles not indexed in MEDLINE), EMBASE, Cochrane Central Register of Controlled Trials, Web of Science, Google Scholar, and grey literature for studies relating to antibiotic prescribing and ambulatory care. We will also examine the bibliographies of retrieved articles. Only articles which were published in English between 2001-2019 will be included in the review. Articles published before the year 2001 will be excluded.

Here was my final search strategy for a current project:

We will search MEDLINE for studies pertaining to the nature of antimicrobial resistance research networks or collaboration in Asia. Only articles published in English between 2011-2020 will be included in the review.The bibliographies of retrieved articles will also be examined to retrieve relevant articles. Due to broad spectrum of topics to be explored with regards to AMR in human, animal and environment health, all types of studies will be considered, including observational, qualitative, quantitative, mixed methods and reviews. Articles published before 2011 will be excluded. Following search terms will be used,  adapted to the requirements of databases. Additionally, we will also search Google Scholar to identify grey literature.

(("antimicrobial resistance" or "AMR" or "antibiotic resistance") and (network* or partner* or collab* or coordin* or joint* or relation* or fund*) and (research* or activit*)).ti 
(("antimicrobial resistance" or "AMR" or "antibiotic resistance") and (research priorities).ti

 

My team excluded PubMed and Scopus. PubMed was excluded because we wanted to use MeSH-controlled vocabulary only, which are official categories ascribed to all papers in the database.2 Next, although Scopus has titles not included in the MEDLINE database, many variations of our search terms did not yield relevant results. Hence, we decided to exclude it altogether.

Selecting relevant sources 

Here was how I did it. I had downloaded the literature derived from my search and went through them one by one. For each paper, I first read the abstract, introduction and conclusion. Doing so provided me with a quick insight into whether the paper would be relevant or not. Next, I dove into the content proper, making highlights and notes wherever relevant. Finally, if I deemed the paper useful to my research question, I added the paper to a separate folder and started making literature notes.

Literature notes are key information from the paper that I foresee being useful later in my research report, only explained in my own words. My literature notes also include notes on how I intend to use the information. For example, I may wish to refute a particular argument stated in the paper, or use data collected in the paper to make my case. Whenever necessary, I would add useful quotes or highlights from the paper too. When making literature notes, I use RoamResearch, which serves as a powerful archive for report writing later. 

I also save the relevant papers in a specific folder shared with my team on OneDrive.

I will dive into the second part of how to do a literature review next week. It will be about consolidating your sources and situating your research project in the current literature.

Compiling a seed list for a social network analysis

Social network analyses has been used to map and understand the nature of relationships between individuals, organisations, communities and societies. It has also been used to map and understand the flow of goods, services, information between and within those same networks.

More concretely, social network analysis is a research method used to:

  1. Identify how individual actors are situated/ located/ embedded in an overall network
  2. Identify how individual actors (choices, behaviours, relationships) contribute to patterns observed on a macro, network-wide scale.

When conducting social network analysis, your end goal is to produce a network map. To get there, you need to sample people in the network. One way of starting is to compile a seed list.

A seed list is a focused list of people you want to contact first. After a seed list is done, you can then ask each person in the list to name some or all of their ties to other actors you are interested in mapping. The process then repeats from there, generating names that you add to your network map.

In my project, we are interesting in creating a network map of researchers involved in antimicrobial resistance (AMR) in Asia. I learned from a team mate that there three primary ways of putting a seed list together.

  1. Recruitment through conferences.
  2. Identifying authors of key reports.
  3. Through personal contacts.

We decided to compile our seed list in the following way:

  1. Recruitment through the PMAC conference, which is the most important conference on global health in Asia. Invited participants would be key experts in their fields.
  2. The key report used would be a landmark Lancet paper. The list of authors would be filtered through to identify those who meet our project’s inclusion criteria.
  3. Personal contacts from a wide variety of research institutions around Asia.

And that is how a seed list is compiled.

The seed list would amount to about 30-40 individuals. Snowball sampling would be done from then on in order to map the rest of the network.

In my next research post, I will share more about how to organise a seed list based on seed attributes.

  1. I used this as reference: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=170504 ↩
  2. PubMed citations come from 1) MEDLINE indexed journals, 2) journals/manuscripts deposited in PMC, and 3) NCBI Bookshelf. Both MEDLINE and other PubMed citations may have links to full-text articles or manuscripts in PMC, NCBI Bookshelf, and publishers’ Web sites. If you limit your PubMed search to MeSH controlled vocabulary or the MEDLINE subset, you will see only MEDLINE citations in your results. Read more at https://www-nlm-nih-gov.libproxy1.nus.edu.sg/bsd/difference.html ↩