How to Extract Data for a Systematic Review

Data extraction means systematically recording the characteristics, methods and results of each included study. A well-planned, piloted form, extraction in duplicate or with independent verification, and a record of contacts and decisions reduce errors.

Plan what to extract

• Study identification: authors, year, registration number, funding, conflicts.
• Methods: design, setting, dates, sample size.
• Participants: eligibility, age, sex, baseline characteristics.
• Interventions and comparators: details and duration.
• Outcomes: definitions, measurement tools, time points.
• Results: effect estimates, confidence intervals, event counts or means and SDs for each group.
• Risk-of-bias information.
• Notes: where in the paper each datum came from.

Process

1. Draft the form from your protocol.
2. Pilot it on 3–5 studies and revise.
3. Extract independently (or one extracts and another checks all or a sample) and resolve differences.
4. Contact authors for missing data, noting dates and replies.
5. Convert and calculate with care, documenting formulas and assumptions.
6. Lock the dataset before analysis.

MRN provides a free data extraction template as a spreadsheet starting point.

Illustrative example

Invented for teaching. A study reports medians and interquartile ranges, but the planned meta-analysis uses means. The team records the original numbers, applies a documented conversion method from the Cochrane Handbook and flags the study for a sensitivity analysis.

Common extraction traps

Effect sizes may be reported as odds ratios, risk ratios or hazard ratios; do not mix them without thought. Outcomes at several times need a rule for choosing which to use. Intention-to-treat and per-protocol results can differ, so record which one you extracted. Subgroup data may appear in supplements. Keep a list of data you calculated and the formula, since reviewers may ask.

Common pitfalls

• Extracting without piloting.
• Overwriting entries with no change log.
• Mixing up different time points.

Key takeaways

• Plan, pilot, double-check and document.
• Keep source locations for each data point.

This page is general educational information about research methods. It is not medical, legal or regulatory advice; follow your institution's ethics board and your target journal's instructions.