Structure of Search Strategies for Systematic Reviews: Line by Line versus Block by Block versus Single-Line
Note: This post has NO advertisement/marketing component.
Over the past 18 years, I have had this discussion several times with information specialists, researchers, students, and colleagues. Some seem to be defending a single approach at the beginning of the conversation, but as we go forward, they change their minds.
Unsurprisingly, all three approaches of structuring the search strategies are useful at the right time and place. There is no advice in PRESS, PRISMA, MECIR, or PRISMA-S to push you to follow one of the approaches.
Since there is no best practice guide, many get confused about what structure to choose for each purpose. Before I start explaining and avoiding wasting your time, this post is not about structuring the search based on PICO families (PECO, PIPOH, PICOS, etc.).
Line by Line Search Strategy
Borrowed from computer programming, we put the single term in each line in this structure, and there is no guide on how much information you can put in each line. Since each line is concise — usually one term or one set of terms followed by field tags — the entirety of the structure becomes long, and you have to scroll from top to down to see all the lines. In some cases, there are 300–400 lines! At the same time, since each line carries only a little information, it is easy to follow each line and line numbers to figure out the structure by looking at Boolean logic such as AND, OR, and NOT. Line by line searching makes it very easy to find the errors. The main trouble is that if you want to run this search, you have to copy the line, paste it into the search box, and press Enter/Search 300–400 times, depending on the number of lines. If this repetitive task does not turn you into a mad hatter, it will waste time. On the other hand, it is easy to check and peer-review such strategies and shows the contribution of each term/line to the search, so it can be beneficial when you develop/validate a search filter. I use this structure for training purposes as students get it well and find the errors on the spot. See a single example from Ovid Embase below: