Systematic reviews and meta-analyses are widely acknowledged as the highest level of evidence for evidence-based medicine (EBM). They are comprehensive reviews of the literature carried out using systematic methods that aim to obtain all appropriate studies that answer the research question. A review is often the first step in most PhD projects as it provides a thorough understanding of what is already mentioned in the literature and a robust summary of the results. This then enables you to find the gaps in the research and base your research on it. Through this 5 part blog series, I want to share my reflections on conducting a systematic review and meta-analysis. This first part will go through formulating a research question, inclusion & exclusion criteria and search terms.
Formulating your research question
So before anything, you need a research question to answer through your review. Here are some things I had to keep in mind while forming mine:
- Not too broad or too specific: finding the perfect balance can be tricky but it’s so important. Having a question that is too broad is going to retrieve too many papers, whereas if it’s too specific you won’t find any studies. For example, if you are researching mental health this can be quite broad because there are so many mental health conditions. Try to be more specific with which conditions you are interested in e.g. depression, bipolar etc. On the other hand, if you focus your research question on a single region of the world e.g Bangladesh, although it might be an interesting question on its own, you may have to expand out the location to something like South Asia to retrieve a sufficient number of papers. This was the case for my review so I expanded the location out to include India and Pakistan which share similar cultural and social norms as Bangladesh.
- PICO: structure your question into population, intervention, comparison and outcome. This ensures you address all the key elements of a research question. This works well for RCT’s, but if you are doing a review of observational studies, like I did, you might consider using something like SPIDER: Sample, Phenomenon of interest, Study design e.g. cohort, cross-sectional etc, evaluation of outcome, research type (quantitative, qualitative or mixed-methods). This will then help with the inclusion & exclusion criteria.
Inclusion & Exclusion criteria
Now that you have your research question, you want to form a criterion for what studies you will include and what studies are not appropriate- your inclusion & exclusion criteria. I had to refine these a few times to ensure they capture exactly what studies I want and do not want.
- Inclusion criteria: these are the criteria which all your included studies must have. Go through your PICO or SPIDER tool for this. State what population type must be in the study e.g. pregnant women, what intervention type or phenomenon of interest e.g. cognitive behavioural therapy or depression, outcome e.g. mental health score or risk factors and what study designs you will include e.g. RCT’s, cross-sectional, cohort etc. Try to be specific with these criteria as they will really aid your study screening phase.
- Exclusion criteria: these are criteria which you specifically do not want in your included studies. These commonly include language restrictions, date study was published, study design, or if there is anything more specific with regards to your research that you do not want to include. For me this was excluding RCT’s, any studies not in English, published before 2000 and excluding all other psychiatric disorders other than depression and anxiety.
- Make sure you can explain each of your choices. For example, if you are restricting the date of the study, why? I personally did it for my review because from an initial scoping search of a database I found no studies before 2000 so this would focus my study search.
Finally, now that you have your criteria for your study inclusions you can form search terms. These are the words/phrases you will use in the databases to search for articles. The most common database platform in medicine is OVID (Medline & Embase) with others including Web of Science, Scopus, CINAHL and many more. These databases have advance search options where you can input search terms using Boolean search which allows you combine terms using AND, NOT and OR.
- Go through your PICO or SPIDER terms and use as many appropriate synonyms for each term. For example, if your population is perinatal women you can also use: maternal, antenatal, pregnant etc.
- Form your search string for each synonym using OR
- Use AND to add your PICO/SPIDER terms together
- Use NOT if you there is a specific category you do not want you search to pick up
- Truncation (*) is a tool that allows you to pick up variations in the word e.g pregnan* will pick up all articles that pregnant or pregnancy because you put the * at the end.
- Tip: look at other studies you have found that answer your research question and see what terms they have used in their title and abstract which might help you to come up with other important terms!
Example search strategy
|Search strategy for the research question- Social determinants and views of maternal depression and anxiety in South Asia|
|1. Perspective* OR views OR opinions OR attitude* OR perception* OR factors OR “risk factors” OR determinants OR dimensions OR correlates|
|2. Maternal OR perinatal OR antenatal OR postpartum OR postnatal OR pregnan* OR women|
|3. Depress* OR stress OR mood OR Anxi*|
|4. Bangladesh* OR “south Asia*” OR Pakistan* OR India*|
|1 AND 2 AND 3 AND 4|
- Always keep a diary of your decisions throughout, it is very easy to forget a search term or inclusion criteria if you don’t make a note of it
- Clarity is key! Make your question, criteria and search terms as clear as possible with no wishy washy statements
- Check Prospero to make sure a similar review protocol has not been published already!
- Always ask yourself why have I made this decision, if you have a solid answer backed with evidence then go ahead with it
So that concludes the first part of the systematic review series! These first few steps make up the backbone of your review protocol so it is really key you spend time on this to make it robust. The next part will go through reference managers & screening!