Start with the number, then find its label: who was counted, what was measured, what it was compared with, and when. Keep those details together when you summarize a sleep study. A percentage without that context can answer a different question from the one you thought you were reading.
This guide helps you take notes on published research. Every numerical example below is invented for practice; none describes a Hostage Tape study or a treatment recommendation.
1. Find the result behind the headline
Open the paper linked to the claim and locate the relevant results table or paragraph. Save the title, publication date, and exact table or section alongside your notes. If you have only an abstract or a promotional summary, record that limitation rather than filling in the missing details.
For randomized trials, the CONSORT 2025 reporting checklist provides useful signposts. Item 14 covers the outcome, calculation, summary measure, and time point; items 21b–c cover analysis populations and missing data; item 26 covers results and precision. These are reporting requirements for randomized trials, not a universal quality score for every kind of research.
2. Write the denominator next to the percentage
Imagine that 25 people enter a study and 20 complete its final questionnaire. Of those 20, 16 report improvement.
- Among questionnaire finishers: 16 ÷ 20 × 100 = 80%.
- As a share of everyone enrolled: 16 ÷ 25 × 100 = 64%.
Both calculations use the same 16 people, but they answer different questions. The second calculation does not establish that the five missing participants failed to improve. Their final responses are unknown in this example.
A clear summary would be: “Sixteen of the 20 final questionnaire respondents reported improvement; 25 people originally enrolled.” It preserves the observed result and makes the missing responses visible.
For an actual paper, use its stated analysis population and missing-data method. Cochrane’s methods handbook, section 6.4.2, explains why the enrolled sample and the number with observed outcomes can differ. It also cautions that rounded percentages may fit more than one numerator. Do not reverse-engineer an exact participant count from a rounded headline percentage.
3. Separate people, scores, and percentage points
In the invented example, 80% describes the proportion of respondents reporting improvement. It does not tell you how much their scores changed.
Try a separate arithmetic exercise: a count falls from 10 to 7. That is a decrease of 3, or 30% of the starting count: (10 − 7) ÷ 10 × 100. If a proportion falls from 10% to 7%, the decrease is 3 percentage points, also a 30% relative decrease. Writing “3% lower” would leave the reader guessing which calculation you mean.
Before copying a number, add its unit in your notes: people, minutes, score points, percentage points, or relative percentage change. Cochrane distinguishes outcome types and ratio versus difference measures in sections 6.1.1–6.1.2. Retain the measure the authors actually used.
4. Keep the measurement method in the sentence
A participant’s report is a defined type of outcome. FDA’s explanation of clinical outcome assessments describes patient-reported outcomes as coming directly from the patient without another person interpreting the response. It also distinguishes clinician, observer, and performance assessments.
When taking notes, write down the questionnaire or instrument named in the paper. Preserve wording such as “reported better sleep” if that is what was assessed. Replacing it with “spent more minutes asleep” would introduce a different measurement. Likewise, retain whether a result is a final score or a change from the starting score.
5. Keep the comparison and time window attached
Complete this sentence in your own notes: “The authors compared ___ with ___ at ___.” If the comparison is with the same participants’ starting values, say so. If it is between study groups, name the groups. If the time point is three weeks, keep three weeks in the summary.
For randomized trials, CONSORT items 13 and 26 call for the comparator, group results, and effect precision, such as a confidence interval. Copy any reported interval beside the estimate, and note where the paper discusses limitations. The checklist helps you locate information; it does not establish that an intervention works for you. CONSORT 2025
A reusable reading note
Keep one note for each result you want to quote:
- Source: paper title, link, and exact table or paragraph.
- People: eligibility, enrollment, and the group included in this result.
- Measurement: outcome, instrument, and unit.
- Comparison: starting value or comparison group.
- Timing: visit or follow-up window.
- Result: estimate, denominator where relevant, and reported uncertainty.
- Open questions: missing responses or details you could not locate.
Then write one sentence using only what the note supports. In our invented questionnaire example: “At the final questionnaire, 16 of 20 respondents reported improvement; five of the 25 enrolled participants had no final response.” That sentence is ready to check against its source without turning a participant count into a score change or a promise.