Why Careful Reporting Matters
A statistical result can be numerically correct and still mislead. “The rate doubled” may be true while concealing that the rate rose from 1% to 2%. “Most patients improved” is difficult to evaluate unless readers know which patients were counted, how improvement was defined, and over what period it was assessed. Clear reporting gives readers enough context to understand both the result and its limits.
A useful report does more than state a number. It identifies the measure, the group to which it applies, the relevant time period, and the units. It also describes the denominator when reporting a count or percentage, makes missing observations visible when they affect interpretation, and avoids wording that claims more than the data support.
Give Every Result Its Context
Readers need to know what a result measures before they can interpret it. For a percentage, specify the event or category in the numerator and the group represented by the denominator. For a measurement such as time or weight, state the unit. In either case, identify the people or units described and the period covered when those details matter.
For example, “18 improved” is incomplete. It does not say what counted as improvement, how many people were assessed, or when the assessment took place. A clearer report might say: “Among the 40 participants who completed the six-week assessment, 18 met the study’s stated improvement criterion.” This wording identifies the outcome, denominator, and time point without implying that all participants improved or that the intervention caused the improvement.
Different denominators can answer different questions. The proportion who responded to an invitation is not the same as the proportion with a particular outcome among respondents. Both may be useful, but they must not be presented as though they describe the same group.
Worked Example: Separating Response Rate From Outcome Rate
In a hypothetical clinic feedback survey, 50 patients are invited to respond. Forty return a survey, and 18 of those respondents say they were satisfied with their visit. Report both percentages and explain their denominators.
Response rate. The numerator is the 40 returned surveys, and the denominator is the 50 invited patients:
Satisfaction among respondents. The numerator is 18 satisfied respondents, and the denominator is the 40 people who returned surveys:
A clear report is: “Of 50 invited patients, 40 returned the survey (80%). Of the 40 respondents, 18 (45%) reported satisfaction.” It would be misleading to say that 45% of all 50 invited patients were satisfied: the satisfaction status of the 10 nonrespondents is not known. The two percentages also answer different questions, so their denominators should appear beside them.
Show What a Comparison Actually Says
A comparison is most informative when it gives the values being compared, not merely a label such as “large increase” or “twice as high.” Percentages can change substantially even when the underlying counts are small. When useful, report both counts and percentages so readers can see the scale of the groups as well as the relative frequency.
The arithmetic provides a simple justification for including denominators: the same count can represent very different percentages in groups of different sizes. For example:
Both groups have six events, but the percentages differ by a factor of ten because their denominators differ by a factor of ten. Reporting only “six events” would hide the group sizes; reporting only a percentage would hide the number of events. Which quantities to emphasize depends on the question, but the denominator must be clear whenever a percentage is reported.
Worked Example: Describing a Difference Without Exaggeration
A hypothetical workplace records whether employees reported a minor injury during two separate months. In Month A, 6 of 30 employees reported an injury. In Month B, 6 of 300 employees reported one. Calculate the percentages and write a balanced comparison.
For Month A:
For Month B:
The count is 6 in each month, while the reported percentage is 20% in Month A and 2% in Month B. A balanced report is: “Six employees reported an injury in each month. This represented 20% of the 30 employees recorded in Month A and 2% of the 300 employees recorded in Month B.” It is also reasonable to state that the reported percentages differed by 18 percentage points. Calling the Month A percentage “ten times higher” may be mathematically defensible as a ratio of 20% to 2%, but without the counts and group sizes it can overemphasize the relative comparison and obscure how few events occurred. These records alone do not establish why the percentages differ.
Use Qualified Language and State Limitations
The strength of a conclusion should match the strength and scope of the evidence. A descriptive result summarizes the observations that were recorded. It does not automatically establish that the same pattern holds in a broader population, that one factor caused another, or that a result will persist in the future.
For instance, if a convenience sample of volunteers reports high satisfaction, the appropriate statement is that satisfaction was high among the responding volunteers. It is not enough to call the volunteers a sample and then claim that all members of a wider population would respond similarly. The way participants were selected and the extent of nonresponse can limit how far the result can reasonably be generalized.
Likewise, a difference between groups does not by itself show that group membership caused the difference. Other characteristics or circumstances may also differ. If the data establish only that two observed groups had different values, describe that difference rather than using causal language such as “led to,” “prevented,” or “resulted in.” A carefully qualified statement can still be useful: it tells readers what was observed while marking what remains uncertain.
Worked Example: Reporting a Measurement With Its Scope
In a hypothetical pilot activity, six volunteers record the number of hours they sleep on the night after the activity: 5, 6, 6, 7, 8, and 8 hours. Calculate the mean and write a report that does not overgeneralize.
The sum is \(5+6+6+7+8+8=40\) hours, and there are \(n=6\) recorded values. The mean is:
A suitable report is: “The six volunteers recorded a mean of 6.67 hours of sleep on the night after the pilot activity.” The report identifies the statistic, unit, group, and time point. It should not say that the activity caused the volunteers to sleep for this amount, or that all people who take part in the activity would average 6.67 hours. The six observations describe these volunteers; the small volunteer group may not represent a wider population, and no comparison is provided to show what their sleep would have been otherwise.
Make Missingness and Definitions Visible
A denominator is meaningful only when the included observations are clear. Some people may not respond; some records may lack the measurement of interest; or a value may be excluded under a documented rule. If the number included differs across outcomes, report the denominator for each outcome rather than giving one overall sample size that could be mistaken for the base of every calculation.
Definitions matter just as much as counts. Terms such as “completed,” “improved,” or “on time” can have more than one interpretation. A report should state the rule used when readers could reasonably interpret the term differently. If an outcome is defined by a threshold, name the threshold and its units. This lets readers understand what was counted and makes the result easier to compare with another report using a different definition.
Limitations should be relevant and specific, not a vague disclaimer added at the end. If only 40 of 50 invited people responded, say so when that nonresponse could affect interpretation. If measurements cover one site or one short period, name that scope. Such details do not make the result useless; they help readers judge what it can support.
Worked Example: Reporting Different Denominators for Two Outcomes
In a hypothetical assessment, 60 participants enroll. At week four, 52 provide a symptom score, and 39 of those 52 meet a stated improvement threshold. At week eight, 45 provide a symptom score, and 31 of those 45 meet the same threshold. Report the results without silently treating missing scores as non-improvement.
At week four:
At week eight:
A transparent report is: “Among participants with an available symptom score, 39 of 52 (75.0%) met the improvement threshold at week four; at week eight, 31 of 45 (68.9%) met it. The assessment began with 60 participants, so scores were unavailable for 8 at week four and 15 at week eight.” This wording reports each denominator and makes the changing availability visible. It does not count participants without a score as unimproved, because their outcome is unknown. Nor does the difference between the two percentages by itself explain why fewer scores were available later or establish a change in the outcomes of all 60 participants.
A Practical Reporting Sequence
Before sharing a result, check that a reader can answer four questions: What was measured? Which people or units are described? What is the denominator or unit? What limits the conclusion? A concise reporting process helps make those details consistent.
Define the outcome or variable, including any threshold or classification rule needed to understand it.
Say who or what was observed and when. Distinguish the people invited, the people who responded, and the people with usable measurements when these groups differ.
Report the count and denominator for a percentage when appropriate, or give the units for a measured quantity. Make different denominators explicit.
Describe observed differences without implying causation or broader generalization unless the study design and evidence support those conclusions.
Note missing observations, restricted coverage, or other features that materially affect what readers can infer.
Common Mistakes
- Giving a percentage without its denominator. “45% were satisfied” is incomplete if readers cannot tell whether the denominator is respondents, invited patients, or another group. Name the base explicitly.
- Confusing the group with the denominator. The group might be the 50 invited patients; its denominator is the number 50. Keeping the group and its numerical size distinct makes percentage statements precise.
- Using a dramatic comparison without the underlying values. “Ten times higher” can magnify a relative difference while hiding small counts. Provide the counts, denominators, and percentages when they clarify the comparison.
- Treating missing outcomes as known outcomes. A participant with no recorded score cannot automatically be classified as either improved or unimproved. Report the available denominator and the number without a measurement.
- Turning an observation into a causal claim. A difference observed between groups does not, on its own, show that the group condition caused the difference. Use descriptive language unless causal evidence is warranted.
- Generalizing beyond the observed group. Results for a small or selectively recruited group describe that group directly. Broader claims require evidence that supports applying the results elsewhere.
Check Your Understanding
For each question, focus on what a reader would need in order to interpret the result correctly.
- A survey invite is sent to 80 people; 60 respond, and 42 respondents choose “satisfied.” What are the response rate and the satisfaction percentage among respondents? State each denominator.
- Explain why a report of “12 cases” is difficult to interpret without the group size and the period covered.
- A report says that 70% improved but provides no definition of “improved.” What additional information would make the result clearer?
- At a later assessment, fewer people have recorded outcomes than at an earlier assessment. What should the report say about the denominators, and why should missing outcomes not automatically be counted as failures?
- Two groups have different observed outcome percentages. What wording describes the difference without claiming that group membership caused it?