Publish a Results Table That Can Be Reused Without Guessing

Define one result per row

A machine-readable results file should identify the analysis, outcome, population, time point, contrast and effect measure for each estimate. A formatted cell such as 1.25 (0.90–1.73) is convenient for reading but difficult to validate or reuse without parsing. Store estimate, lower bound and upper bound in separate fields.

The W3C tabular-data model describes tables and associated metadata. Use that principle to pair a simple data file with a dictionary defining field types, units, allowed values and missing-value conventions.

Preserve meaning alongside numbers

Include the confidence level, interval method, analyzed sample size and model identifier where relevant. State whether estimates are adjusted, transformed or standardized. Separate a numeric estimate from its display string so rounding for the manuscript does not remove reusable precision.

Use explicit labels for missing, not applicable and not estimable results. Do not place zero in a numeric field merely to avoid an empty cell. If a field can legitimately be absent, document the condition and preserve a reason field.

Design a hypothetical row

Suppose an analysis estimates a risk ratio of 1.25 with 95% interval [0.90, 1.73] at 30 days. A reusable row contains effect_measure = risk_ratio, estimate = 1.25, lower = 0.90, upper = 1.73, confidence_level = 95, time_days = 30 and a comparator definition.

Without the comparator, the same values could be interpreted in reverse. Without the effect measure, a reader cannot know whether one or zero is the null. Without the time horizon, a risk ratio may be mistakenly compared with a result measured at another follow-up.

Validate before release

Check unique result identifiers, endpoint ordering, numeric types and consistency with the published display. Confirm that values from different models have not been joined using an incomplete key. A model-and-outcome key may still need population and time to be unique.

Review disclosure and permissions before sharing data. An aggregate results file can still contain sensitive small-cell information, so use the project’s approved release rules. Provide a version and source-analysis identifier. The final artifact should let another researcher reconstruct the reported table and understand each estimate without inferring hidden units, denominators or contrast directions from a PDF layout.

Source verification: 6 November 2026. Examples are hypothetical.