Baseline Characteristics Table¶
Document — instantiates Baseline Covariate Balance Verification
The arm-by-arm 'Table 1' that enumerates a frozen set of pre-treatment covariates and displays their distribution across study groups as the published balance record.
A Baseline Characteristics Table — the familiar "Table 1" of a trial report — is the document that enumerates which pre-treatment covariates were checked and lays out their distribution one column per study arm. Its identity is enumeration and presentation, not metric computation: it fixes a list of variables in advance, certifies that each was measured before exposure, arranges them against the comparison arms, and publishes the result so any reader can see exactly what was and was not examined. The table is the corpus of the balance check made visible; other mechanisms measure and interpret it, but this is the artifact that pins down the covariate set and hands it to the reader.
Example¶
An education trial testing an after-school tutoring program randomizes 600 students to tutoring or control and, in its pre-analysis plan, freezes the baseline covariate list: prior standardized test score, attendance rate, grade level, free-lunch eligibility, and English-language-learner status. Table 1 shows each variable as a row, with the Treatment and Control columns giving means (or percentages) and dispersion, plus the count in each cell. Because the list was frozen before any outcome data arrived, the team cannot quietly drop the awkward "prior test score" row when it later looks slightly uneven — the registry is on the record. When a reviewer asks "did you check whether the groups started at the same achievement level?", the answer is one row in a published table rather than a defensive reconstruction.
How it works¶
- Freeze the covariate registry in the analysis plan before outcomes are seen, so the list of checked variables cannot be edited to flatter the result.
- Enforce the measurement window — a variable earns a row only if it was recorded before exposure began; nothing measured after randomization qualifies.
- Lay out the comparison frame as columns, one per arm, so the groups that must be comparable are named explicitly.
- Report a summary and dispersion per cell (mean ± SD, or count and percent) plus n, and publish. Deliberately, no baseline significance tests.
Tuning parameters¶
- Covariate breadth — a lean prognostic core vs. an exhaustive descriptive list. More rows document more, but bury the variables that matter in clutter.
- Summary statistic — mean ± SD vs. median (IQR) per cell; the latter is more honest for skewed variables.
- Missingness column — whether to report per-arm missing counts. Including it exposes differential missingness that a clean mean would mask.
- Overall column — whether to add a pooled column alongside the arms; useful for description, irrelevant to the balance comparison itself.
When it helps, and when it misleads¶
Its strength is transparency and pre-commitment: a frozen, published registry of pre-treatment covariates prevents selective reporting and gives reviewers a fixed reference for what the study claims to have checked. It is the artifact that makes the whole diagnostic auditable.
Its signature failure is the decorative baseline table — values are printed, but no threshold, metric, or interpretation says what counts as imbalance or what happens if it appears. A close cousin is the Table 1 fallacy: appending significance tests to each baseline row in a randomized trial, which is logically incoherent because any difference there arose by the chance the randomization is known to produce, so the test answers a question no one is asking.[n1] The guarding discipline is to treat this table as the registry-and-record only, and to pair it with a metric-and-threshold mechanism that actually adjudicates imbalance rather than letting the raw numbers imply a verdict.
How it implements the components¶
pre_treatment_covariate_registry— the table's rows are the frozen registry of variables the study commits to checking.baseline_measurement_window— the inclusion rule that every row must have been measured before exposure enforces the window.causal_comparison_frame— the columns name the arms that must be comparable, fixing the contrast the balance check is about.transparent_balance_record— the published table is the human-readable record any reader can inspect.
It computes no standardized metric or tolerance judgment — that is Standardized Mean Difference Table — and renders no visual scan of the balance picture, which is Covariate Balance Plot.
Related¶
- Instantiates: Baseline Covariate Balance Verification — it is the registry-and-record artifact the rest of the diagnostic reads from.
- Sibling mechanisms: Standardized Mean Difference Table · Covariate Balance Plot · Randomization Integrity Audit · Stratified Balance Check · Automated A/B Balance Dashboard · Prespecified Adjusted Estimation Plan · Balance Exception Report
Editorial Notes¶
Form Classification¶
Form family: Representation, Specification & Plan
Rationale: The arm-by-arm 'Table 1' that enumerates a frozen set of pre-treatment covariates and displays their distribution across study groups as the published balance record, making its operative form a non-executable information artifact that externalizes static or prospective structure.
Independent corroboration: The frozen evidence defines Baseline Characteristics Table as 'The arm-by-arm 'Table 1' that enumerates a frozen set of pre-treatment covariates and displays their distribution across study groups as the published balance record', so its operative form is Representation, Specification & Plan.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Medicine & Healthcare
Origin pattern: Single lineage
Present-day reach: Specialized
Rationale: Clinical-trial reporting institutionalized the arm-by-arm baseline characteristics 'Table 1.'
Related originating lineages:
- Statistics & Experimental Design — Experimental design supplies covariate balance and pretreatment measurement principles.
Review resolution: The artifact is explicitly the published clinical-trial 'Table 1.' CONSORT's trial-reporting guideline requires a table of baseline demographic and clinical characteristics for each study group and rejects baseline significance testing, so medicine and clinical research are primary while experimental statistics supplies the design and summaries.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
Notes¶
The table's value is that it displays and enumerates; it does not adjudicate. Keeping enumeration separate from judgment is what lets a study improve how it measures imbalance — switch to standardized differences, add a plot — without re-negotiating which covariates were supposed to be checked in the first place.
[n1] The Table 1 fallacy is the practice of significance-testing baseline covariates in a randomized trial; Stephen Senn and others have long noted it is illogical, since under valid randomization any baseline difference is by construction a chance difference, and the test's null hypothesis is known in advance to be true. Standardized differences, judged against practical tolerances, are the recommended alternative. ↩