How to Report Quantitative Results Without Turning Your Results Section Into a P-Value List

Before model results, orient the reader.

Depending on the design, report:

• number screened or approached;
• number eligible;
• number enrolled;
• number analysed;
• missing observations;
• loss to follow-up;
• group sizes;
• relevant exclusions.

If the denominator changes between outcomes, make that visible.

A result such as:

63% used contraception

is ambiguous if the reader cannot tell whether that means:

• 63/100;
• 630/1,000;
• 63% of participants with complete data;
• weighted population estimate.

Numbers need denominators.

Use descriptive statistics that help readers understand who was studied.

For continuous variables, this might include:

• mean and standard deviation;
• median and interquartile range;
• range where relevant.

For categorical variables:

• number and percentage.

Do not automatically report every questionnaire item.

Description should support the study question.

Suppose Objective 1 is:

To estimate modern contraceptive use.

The results should answer it directly:

Modern contraceptive use was reported by 56% of participants (95% CI 52%-60%).

Suppose Objective 2 is:

To compare uptake between urban and rural participants.

Report:

Uptake was 63% among urban participants and 48% among rural participants, an absolute difference of 15 percentage points.

Add the appropriate confidence interval and effect measure.

The objective should be visible in the results.

There was a significant relationship between residence and contraceptive uptake (p = 0.002).

What was the relationship?

Which group had higher uptake?

How large was the difference?

Modern contraceptive use was reported by 63% of urban participants and 48% of rural participants, a difference of 15 percentage points (95% CI [x-y]).

If an adjusted model is central:

After adjustment for [prespecified variables], urban residence was associated with [effect estimate] (95% CI [x-y]; p = [value]).

Now the reader sees:

• the raw pattern;
• the adjusted estimate;
• uncertainty.

Major reporting guidance such as STROBE and CONSORT emphasizes effect estimates and measures of precision.

A useful reporting order is:

Estimate → confidence interval → p-value if relevant

For example:

RR 0.72 (95% CI 0.60-0.87; p = 0.002)

rather than:

p = 0.002, therefore RR 0.72.

The p-value does not deserve top billing simply because software prints it prominently.

If you report:

• crude/unadjusted estimate;
• adjusted estimate;

label them clearly.

For example:

Then explain what the adjusted model includes and why.

Do not make readers guess whether “OR” is crude or adjusted.

A common results table contains an “adjusted model” built from variables that were significant in bivariate analysis.

That may be what you did.

Report it honestly.

But do not confuse statistical screening with causal justification.

If your aim is causal or explanatory, the methods should explain why covariates were included.

Our guide on covariate selection explains this in more detail.

Where p-values are useful:

• report exact values where practical;
• avoid writing only “p < 0.05” if the exact result is available;
• use appropriate thresholds for extremely small values, such as p < 0.001, depending on journal guidance;
• report them alongside the estimate and interval;
• do not turn p-values into a ranking of “important” variables.

Avoid:

p = 0.000

A p-value is not literally zero.

Software output can round.

Use an appropriate inequality.

Report them too.

Do not hide a result because p > 0.05.

For example:

The estimated difference in attendance was 1.6 percentage points (95% CI -0.5 to 3.7; p = 0.14).

That tells the reader what the data remain compatible with.

“Not significant” alone does not.

The text should guide interpretation.

If Table 2 contains:

• 15 exposures;
• estimates;
• CIs;
• p-values;

the narrative does not need to recite all 15 rows.

Use text for:

• primary findings;
• important patterns;
• unexpected results;
• key adjusted estimates;
• results that require contextual explanation.

Let the table hold detail.

Let the prose tell the story.

If analysis denominators vary, say so.

For example:

Of 1,200 enrolled participants, 1,164 had complete outcome data and were included in the primary analysis.

If you used:

• complete-case analysis;
• multiple imputation;
• inverse probability weighting;
• another missing-data approach;

describe it in Methods and report enough in Results for readers to understand what happened.

Missing data should not disappear between the flow diagram and regression table.

Do not call:

• odds ratio a risk ratio;
• prevalence odds a prevalence ratio;
• relative effect an absolute effect.

And do not write:

OR 2.0 means participants were “twice as likely”

without considering whether odds and risk are being conflated.

The result should match the model.

If the effect measure is difficult for the audience, explain it.

Residence was statistically significant (p = 0.02).

Modern contraceptive use was more common among urban than rural participants (63% vs 48%; p = 0.02).

Modern contraceptive use was reported by 63% of urban and 48% of rural participants. After adjustment for [prespecified confounders], prevalence remained 18% higher among urban participants (adjusted PR 1.18, 95% CI 1.03-1.35; p = 0.02).

Now we can see the actual finding.

Avoid:

The higher uptake among urban women may be because they are more educated and have better access.

That may belong in the Discussion unless it is directly tested and appropriately presented.

The Results section should primarily report what the analysis found.

Interpretation comes later.

The exact boundary varies by journal format, but do not quietly introduce causal stories as though the analysis demonstrated them.

Fifty-six percent of participants reported current modern contraceptive use (95% CI 52%-60%).

Uptake was 15 percentage points higher among urban than rural participants (63% vs 48%; 95% CI for the difference [x-y]).

After adjustment for [variables], urban residence remained associated with higher contraceptive use (adjusted PR 1.18, 95% CI 1.03-1.35).

Attendance was 1.6 percentage points higher after the intervention, although the confidence interval included both little change and a modest improvement (95% CI -0.5 to 3.7).

Use the effect measure appropriate to the design.

1. Who was analysed?
2. What did the sample look like?
3. What is the answer to Objective 1?
4. What is the answer to Objective 2?
5. What are the primary adjusted estimates?
6. How precise are they?
7. What missing data or deviations matter?

If a paragraph exists only because SPSS produced another table, reconsider it.