SRDR 5. Tips for Creating a Data Extraction Template

Now it is time to start building out the extraction template. You’ll find this tool under the Extraction Form Builder tab at the top.

 

But, before we talk about how to build out the different sections let’s talk first about some more general good practices for building an extraction template.

Extraction Template Good Practices

First, the types of information (data) that you’ll want to pull from the research articles should be guided by your PICO question and analytic framework. Take a look back at your PICO or your analytic framework and think carefully about the pieces of information you’ll need to answer your question. For instance, looking at this example analytic framework, we can already see a number of fields we will need to add to our extraction template:

The Brown University Evidence Synthesis Academy provides some very helpful direction (What Should You Extract?) on how to use your PICO (D) to begin planning for what fields you’ll need in your extraction template.

Second, know how to organize your questions within your template. Happily, SRDR+ makes this easier by organizing different types of information you’ll need into a series of tabs.

Again, if you haven’t already, take a look at the Brown University Evidence Synthesis Academy What Should You Extract? page for guidance on which aspects of your PICO (D) question go within which SRDR+ tab.

Third, remember that there is a balance between too little information and too much information.The principle of data extraction for systematic or scoping reviews is that you are extracting only the information you need and in a format you can use it. You don’t want to extract everything (otherwise, you could just work from the full text articles), but you want enough information to allow for your planned analyses.

• Too little information extracted will mean either a very limited analysis or may mean going back later to update your template to capture information you missed.
• Too much information, and you will find yourself wasting time extracting information that you will not need for your analysis.

For instance, you will want to plan ahead of time for sub-analyses if you will be carrying out a meta-analysis. So you’ll need to extract information about differences between the samples, interventions, measurement methods, etc. used in the different research articles.  Taking these study characteristics into account can provide strong evidence for why interventions or exposures or tests work differently (i.e., have different outcomes) in different situations. But you have to plan ahead for capturing this information (without trying to extract every piece of information from a study).

Fourth, you can use the Dependencies functionality to help guide the data extractors (and yourself) in knowing which questions to answer, and which they can safely ignore.

Let’s look at a simple example. You’ll notice below that there are two questions that I’ve set up to capture the information on the sex distribution in the sample. What I want to avoid is a free-text field that captures the proportion of both males and females since this may mean I’d have to recode this later (if I wanted to carry out a sub-analysis on single sex studies) and may introduce typos.

This simple structure allows the data extractor to simply click on the option that best describes the sample. If the “Both Male and Female” option is selected then the second question (currently grayed out) will turn on and the analyst only need enter the proportion of females in the sample. So, a click and a simple number entry rather than unnecessary typing.

Why not ask the data extractor to enter the proportion of males in the study as well? For binary options (e.g., male/female), if you know the proportion of one alternative, you automatically know the proportion of the other alternative (e.g., if I know the sample consists of 60% female, I can easily figure out that 40% of the sample is male).

How do you set up a dependency between questions?

When you click on the Dependencies link, you’ll see:

The next series of pages will show you how to actually set up different types of questions in SRDR+