Investigate missing data as a pattern

Analyse [DATASET DESCRIPTION] focusing on missing values. Break down missingness by field, time, source and relevant group only where privacy allows. Propose tests for entry errors, workflow changes and systematic exclusions; distinguish unknown, not applicable and genuinely zero. Provide a missingness matrix, two sensitivity analyses and a decision rule for when imputation would be inappropriate. Do not silently drop records or invent values. Finish with a plain-language note describing how missingness might bias the headline result and what extra data would be most useful.