Missingness Audit — instructions
StatMedAnalysis | Revised 6 November 2026

Open the matching CSV in a spreadsheet or text editor. Keep an untouched copy.
The first data row is a clearly labeled hypothetical example, not study data.
Delete it before entering working records. Add rows without changing headers.
Use public, hypothetical or aggregate information only; no patient identifiers.

Use one row per variable, time point and group. Define eligible denominator first and distinguish missing from structurally not applicable. Counts are nonnegative integers.

Field guide
- variable: Name the variable and retain its operational definition in the data dictionary.
- timepoint: State the visit or assessment window, including permitted timing deviations.
- group: Name the comparison stratum and its inclusion rule.
- eligible n: Enter the integer number for whom this measurement should exist.
- observed n: Enter the integer count with a usable observed value among eligible records.
- missing n: Enter eligible_n minus observed_n; reconcile disagreements before analysis.
- known reasons: Summarize documented causes of missingness; distinguish unknown reasons.
- analysis assumption: Describe the proposed missingness assumption and unresolved evidence; it is not a proven mechanism.

Decision rule
Require eligible_n = observed_n + missing_n for the stated eligible group. Observed patterns cannot establish MAR or MNAR.

Write UNKNOWN, NOT REPORTED or NOT APPLICABLE with a reason in text fields.
Leave unresolved numeric fields empty and explain them in separate project notes;
never encode unknown values as zero. Store the source record and reviewer/date
in your project record. This planning aid is not a validated instrument.

Official guidance
https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-10