NUTR 369: Systematic Reviews

Do I have to form a question?

It is highly recommended to frame your research in the form of a question. This helps you stay focused in your searching and only gather the information you need to accurately and thoroughly answer the research question.

Creating and searching a question is an iterative and exploratory process. You may have an idea of what you want to research, but it takes some time to determine a precise, searchable question.

1. Do your background research to inform yourself of what's already been published in the literature
2. Formulate a basic question based on your background knowledge
3. Go fishing in a couple of representative databases to see what you find (eg PubMed and Web of Knowledge)
1. Was my topic too broad? Too narrow?
2. What terms and phrases are used in the published literature?
3. What aspects of the research am I truly interested in?
4. Form a specific, searchable question.

It is important to understand that different databases require different search strategies. Some have controlled vocabularies, while others are keyword-based; but all databases are more easily searched if you think of the terms that represent each idea in your question prior to searching.

In a systematic review, you are required to search databases that use both controlled vocabularies and keywords. Setting up a simple chart like the one below can help you stay organized in your searching. Create a column to represent each idea and two rows beneath. One row is to generate the controlled vocabulary terms that describe the topic, while the other are all of the synonyms and phrases to express the idea in a keyword search. Terms within a column will typically be combined with OR, while different columns will be combined with AND.

Multiple charts might be made throughout a Systematic Review; perhaps one for each database searched, or a new iteration if terms were added or removed from a different iteration of the search later in the study.

Framing your search in the form of a precise question allows you to clarify the criteria needed for selecting relevant studies. An example question for research might be: Have laws limiting soda sales decreased rates of obesity in children?

After creating a question, the next step is to break your question into its discrete ideas; which is called "chunking," "diagramming," or "PICO"ing in medicine. Typically, these ideas will be the nouns, because in research and literature searching, verbs are not codified - they tend to be what you are measuring. 

Therefore, for the above question, we have three main ideas:

laws                          soda                    childhood obesity

At this point in the process you can identify if your question is not specific enough. You need to have outcomes that are measurable. Take the question: Is diabetes improved with exercise? Does it include a measurable outcome? (How is "improvement" measured?) Is it specific? (do you want all forms of diabetes or just type II?)

Knowing that you have a specific, searchable question and undertanding what discrete ideas are contained within the question makes setting up your searh much easier.

It is important to keep track of your searches and results every step of the way in a Systematic Review. If you've read the PRISMA Guidelines, you know that there is a flow chart that should be included with your review which maps the results found, excludes vs. includes, reasons for decisions and dates. A spreadsheet in Excel can help you keep track of your searching workflow. Columns include:

• An optional ID to uniquely identify a search
• Date you ran the search
• A copy of the complete search string
• The database searched
• The publication dates searched
• Any filters you may have used
• The number of results retrived from the search
• The number of results immediately discarded
• The number discarded after chaecked against inclusion/exclusion criteria
• The # of results leftover for inclusion

It is important to keep in mind that the length of time required to do a Systematic Review means you will likely have to rerun searches you did a year ago in order to update any information that may have been recently published.