“It’s all fun and games until you’re revising your Literature Search Strategy”
You’ve done literature reviews before and you’re confident you’ll nail this one too. Your research objectives are clear, with multiple search strategies in mind. You know which databases you’re targeting and are positive that your chosen date range will return enough relevant articles for you to write your report.
The first stop is PubMed. You run your search parameters, and satisfied with what you see, you extract the data into an excel sheet so that you can screen all your literature together later on. You run your next PubMed string(s), extract and save the results again.
Move to your next database. Rinse, and repeat.
You’ve now extracted data from all your search strings, the excel sheet looks proper, column headings are where you want them to be. Today has been productive.
However, midway through your screening you notice that you haven’t identified many relevant articles yet. And as you progress through the 2nd half of the list, the status remains unchanged.
Your determination doesn’t falter, but reality soon sets in that you won’t have enough relevant articles for your report!
You must revisit your search strategy (and perform the laborious task of data extraction for ALL your strings, again). As your colleagues are wrapping up their workday, you stare hopelessly at Microsoft Excel, and wish for a miracle.
Typical symptoms of the I don’t want to do this again syndrome commonly seen in medical writers. Where junior writers and interns are the most vulnerable.
Assuming your original date range was from 01-Jan-2018 to 31-Dec-2021, this is what you will probably do:
- Extend your search range by a couple of years: it now is 01-Jan-2015 to 31-Dec-2021
- Search for the additional date range 01-Jan-2015 to 31-Dec-2017 for all the search strings, maintaining the same filters
- Stitch the new results to your existing data set individually for each string
This should work, right? Unfortunately, not quite. In the pursuit of maintaining a continuous date range of 01-Jan-2015 to 31-Dec-2021 within your report, you will be faced with challenges when extending the search date range manually. Let’s take an example:
|String Type and Date Range
|Original 01-Jan-2018 to 31-Dec-2021
(primary focus of the research)
|“left atrial appendage” AND (closure OR occlusion) AND imaging AND 2018/01/01[PDAT] : 2021/12/31[PDAT] AND (eng[Language])
||631 – PubMed Results
|Additional 2 years 01-Jan-2015 to 31-Dec-2017
||“left atrial appendage” AND (closure OR occlusion) AND imaging AND 2015/01/01[PDAT] : 2017/12/31[PDAT] AND (eng[Language])
||241 – PubMed Results
|Total upon stitching results (631 +241)
|Continuous 01-Jan-2015 to 31-Dec-2021
||“left atrial appendage” AND (closure OR occlusion) AND imaging AND 2015/01/01[PDAT] : 2021/12/31[PDAT] AND (eng[Language])
||849 – PubMed Results
Looking at the example above, stitching individual date ranges together does not always give you the same results as a continuous date range. Presenting results from a stitched dataset as results from a continuous date range string is incorrect as you will not be able to replicate the same number of results when you run the continuous string. Repeatability is the core of a search strategy; hence, mismatched results would deem your search strategy inaccurate.