A good systematic review highly depends on how the data has been extracted. The incidence of two reviewers recording data from the same study in different ways, or one of the reviewers skipping or ignoring an important element of the study, is not uncommon. These mistakes show up in the final output of your review. Sometimes you may not notice these errors until your review is peer-reviewed. The use of a systematic review data extraction form can help mitigate or eliminate these errors.
With this guide, you will be able to do the following:
- Determine what to include on the form
- Learn how to create, pilot-test, and adjust the form
- Get a sample of what a completed extraction table looks like
- Learn about how and when reviewers extract data on their own and how to resolve conflicts
- Identify the kind of tools used by different types of review teams
Start before you open a spreadsheet. Write down your review question and outcomes you’ll analyze. Each field on the form will connect to one of these. If it doesn’t, it’s probably not needed.
What is a Data Extraction Form in a Systematic Review?
A data extraction form is used to capture data from an included study using the same format for all studies. Most people use a spreadsheet. Some people use an electronic form or paper form in review software.
Don’t get this confused with the data extraction table (or tables) of a published review. The form is your working file. It’s okay to have messy notes, questions you still need to answer, and anything else you want to add. The table in your manuscript is nice, clean, and summarizes the main points.
A well-set-up form can save you hours during analysis. If it’s set up right, your data will already be grouped by study and outcome and then further ordered by time point.
What that manuscript table looks like and which items PRISMA 2020 requires you to report alongside it is covered in the PRISMA 2020 checklist explained.
What Information Should a Systematic Review Data Extraction Form Include?
Your review question decides the exact fields. Still, most forms cover the same four areas.
Study characteristics extraction
- Study ID (one ID per study, even when a study has several publications)
- First author, year, and country
- Study design and setting
- Recruitment period and length of follow-up
Participant, intervention, and comparator details
Which designs you will encounter, and what each one can and cannot establish, is covered in cross sectional study design in medical research and the case-control and cohort guides alongside it.
- What it tells you about
- participants’ baseline characteristics (e.g., age, sex, disease severity)
- the number of participants randomized to each group
- details about the type and dosage of the intervention
- what the comparator was (e.g., placebo, usual care, other interventions)
Types of outcome measures and effect sizes
- For each outcome: how it was defined, measured, and analyzed; when it was measured.
- For continuous outcomes, provide the number of cases for each group and the mean and standard deviation.
- For dichotomous outcomes, the number of cases and the total for each group.
- Effect estimates (e.g., risk ratio, odds ratio, mean difference, etc.) including the 95% CI.
Keep the numbers that the authors provided strictly separate from any numbers that you calculated. It is easy to introduce errors by mixing these numbers.
Risk of bias, funding and additional comments.
The judgments that you made for each domain of the risk of bias assessment, using the risk of bias tool 2, for example.
What those domain judgments mean and how to make them consistently across reviewers is covered in risk of bias assessment in systematic reviews. The domain labels from RoB 2 should appear verbatim in your extraction form.
- Source of funding and disclosed conflicts of interest.
- Page or table number for each piece of information.
- Notes about what is missing, unclear, or contradictory.
Systematic Review
The fields on this form follow from your protocol. If the protocol isn’t settled, neither is the form.
PICO elements, primary and secondary outcomes, risk of bias tool, planned analysis. Every extraction field traces back to a protocol decision made before recruitment started. Learn to build the protocol first and the form almost writes itself.
How to Create a Data Extraction Form, Step by Step
Step 1. State your review question and set your outcomes
Go back to your protocol and break your question into the PICO elements (population, intervention, comparator, outcome) and state your primary and secondary outcomes. Then be strict. If a field won’t provide the answer to your question or go into your planned analysis, remove it. Unnecessary fields give you opportunities to make mistakes and slow everyone down.
How to translate a broad research idea into PICO elements precise enough to generate form fields is covered in the PICO framework for research questions.
Step 2. Select or modify a data extraction template
You don’t have to start from scratch. The Cochrane Handbook, Covidence, and numerous university library guides provide sample forms. Use one and modify it to reflect your review. A form for a diagnostic accuracy review needs fields for sensitivity and specificity. For a review of drug trials, these fields wouldn’t be necessary.
Step 3. Define each field and its coding rules
Write a one-line instruction for every field. This is what keeps two reviewers from filling in the same cell two different ways. A few examples:
- “Sample size” means the number randomized. The number analyzed goes in its own field.
- Type “NR” if something isn’t reported and “UC” if it’s unclear. No blank cells.
- Use drop-down lists for fixed choices like study design or setting.
Put all of these rules on an instructions tab, and give the form a version number.
REDCap enforces these rules automatically and catches inconsistent entries at the point of input rather than during analysis. What is REDCap in medical research covers how to set it up and when a spreadsheet stops being sufficient.
Step 4. Pilot-test the form
Pick two or three included studies that look different from each other. A small two-arm trial and a larger trial with several arms make a good pair. Two reviewers extract data from them separately, then sit down and compare. Check for:
- fields one person filled in and the other left empty
- the same field filled in two different ways
- data that had nowhere to go on the form
- instructions that each reviewer understood differently
Step 5. Revise and finalize the form
Rewrite any confusing definitions. Add the fields you were missing and drop the ones nobody used. Bump the version number and write down what changed. If a change affects studies you’ve already extracted, go back and check them again. Once you’re happy with it, lock the form before full extraction starts. If you change it later anyway, log the change and mention it in your methods section.
Systematic Review Data Extraction Template: Example
Here’s a filled-in example for a made-up randomized trial testing an exercise program in adults with hypertension.
| Field | Illustrative entry |
| Study ID | STUDY-001 |
| Author and year | Ahmed et al., 2024 (fictional) |
| Country and setting | Pakistan; outpatient clinic |
| Study design | Randomized controlled trial |
| Population | Adults with hypertension |
| Sample size | 120 participants |
| Intervention | Structured exercise program |
| Comparator | Usual care |
| Primary outcome | Systolic blood pressure |
| Follow-up | 12 weeks |
| Intervention group result | Mean change, SD, and n from the report |
| Comparator group result | Mean change, SD, and n from the report |
| Funding source | As reported in the publication |
| Source location | Results section, Table 2 |
| Missing data or notes | Unreported values or details to clarify |
This example is fictional and shows how a form is organized. It does not use real research data.
You will see fields like study ID, design, sample size, funding and source in almost every review. The fields you will edit for your question are intervention and outcome.
A simple trial generally fills out one line of the form. Often, studies are more complex than that. That is why many reviewers use tabs to organize their forms.
Instructions: defines fields and specifies coding rules. Also shows the version of the form.
- Study characteristics: year, setting, design, and funding
- Intervention and comparator: details of the intervention and comparison groups
- Outcomes: definitions and statistics
- Risk of bias: tool used and judgment.
- Discrepancy log: disagreements and how they were resolved
How to Perform Independent Data Extraction and Resolve Discrepancies
A single person reviewing the literature will generally miss things. A second reviewer working independently identifies most of the errors. The general procedure is as follows:
- Each reviewer fills out a copy of the data extraction form for each study.
- They compare their forms and identify any differences.
- If they cannot resolve the differences, they identify the coding rule that governs the decision.
- If they still cannot reach consensus, a third reviewer makes the final decision.
- They log the decision and reasoning for discrepancies.
Be honest and comprehensive in your description of your methods in the manuscript. Extracting data and then having it checked is verification. It is not dual independent extraction. Both methods are acceptable, but using one and calling it the other misleads your readers.
PRISMA 2020 requires you to report the number of reviewers, whether they worked independently, and how disagreements were resolved. Keeping a discrepancy log from day one essentially writes this part of your methods.
The full list of what PRISMA 2020 expects you to report in the methods, results and supplementary files is in the PRISMA 2020 checklist explained.
When a study doesn’t provide the information you need, note this on the form and/or contact the study authors for the missing information. Keep a record of any responses, if provided.
Biostatistics
Continuous or dichotomous. Mean difference or odds ratio. Risk ratio or hazard ratio.
The right field on your extraction form depends on what kind of data the outcome produces. Extract the wrong numbers and the analysis has nothing to work with. Learn which effect measure applies to which outcome type before the form is built.
Best Systematic Review Data Extraction Tools
Excel or Google Sheets: Flexible and cheap. Good for small teams and first reviews, but you have to compare entries manually.
Covidence: Includes extraction templates and a built-in comparison step for two reviewers’ work. It’s a paid service, but many universities have a site-licensed copy.
How the two compare on extraction features, pricing and what each plan actually covers is in Rayyan versus Covidence.
Rayyan: Used mostly for study screening. Check what extraction features your plan includes.
JBI SUMARI: Designed for reviews done in the JBI style (qualitative and mixed methods reviews).
REDCap: Allows building of forms with rules to validate entries and catch errors. Many universities offer it for free to researchers.
A well-organized spreadsheet is adequate for first reviews, or for teams of two. For larger reviews and teams, Covidence is preferred. Check the most recent information before you use a tool.
Common Data Extraction Mistakes to Avoid
- Pulling everything from a paper instead of only what your question asks.
- Using different codes, for example “NR” and a blank in a cell, to represent the same thing.
- Forgetting to put page numbers on a cell, making reviewers manually check each page.
- Considering different papers from the same study as different studies.
- This is easiest to catch during screening rather than extraction. How to write a systematic review and meta-analysis covers how to handle multiple publications from one study from the protocol stage.
- Forgetting to log a change, or not rechecking studies before making an edit.
- Transferring the same data to a new form or question instead of reviewing the new data.
Frequently Asked Questions
Q1. What should a systematic review data extraction form contain?
Ideally you should include Study characteristics, Participant and intervention details, type of Outcome Data, Risk of Bias, and Funding. Include fields your question and analysis uses.
Q2. Can I use Excel for data extraction in a systematic review?
Excel and Google Sheets are great for smaller reviews. Use tabs, drop-down lists and instructions to improve consistency of the entries.
Q3. Should two people extract data independently?
It is recommended. If you implement one extractor and one checker, mention it in your manuscript.
Q4. How do I pilot test a standardized data extraction form?
Ask two people to extract data from two or three studies. Compare their data. Make the form or instructions clear for unclear fields, then only start the full data extraction.
Research Mentorship
Extraction errors that reach a peer reviewer are expensive. The ones a mentor catches before pilot testing are free.
PRISMA 2020 asks how many reviewers extracted independently and how you resolved disagreements. Getting that right from the start, with a discrepancy log and a version-numbered form, is the part most first reviews skip.
The American Academy of Research & Academics works with researchers through systematic review methods, including protocol development, data extraction, risk of bias and PRISMA reporting. If your extraction is underway and something doesn’t look right, bring it to us.
Disclaimer:
Articles published by American Academy of Research & Academics are prepared by our team using information from direct experience, publicly available resources, and educational references. AI tools may be used to assist with drafting, proofreading, and formatting; however, all content undergoes review and approval before publication.
The information provided is intended for educational purposes only. Requirements, policies, and processes may change over time. Readers should consult official sources for the most current information.