Data Entry, Validation, and Discrepancy Management
Data entry moves source observations into the database; validation catches problems; discrepancy management resolves them. Know the entry modes: single entry, double entry (with comparison for discrepancies), and direct eCRF entry at the site. Validation includes automated edit checks (run at entry or on demand) and manual review. A discrepancy or query is a formal request to clarify or correct data; it has a lifecycle: raised → sent to site → response → resolution/closure, all tracked and time-stamped.
Understand query types: data clarification (value inconsistent), missing data, and coding queries. Self-evident corrections are minor, unambiguous fixes (e.g., obvious spelling) that may be made by the data manager per predefined rules without a site query - but only if the DMP authorizes it. Never alter source data; corrections flow from the site.
Traps: assuming a query can be closed by the data manager without a site response (only self-evident corrections per policy); confusing validation (checking data against rules) with verification (comparing transcription to source); forgetting that unresolved queries block database lock; and treating a monitor's source data review as a substitute for edit checks. Another trap: entering a value that fails a hard check and assuming it saved - hard checks prevent saving.
Rule of thumb: every query needs a documented resolution and a reason for any change. Worked example: a subject's weight is entered as 750 kg. A range check flags it; the data manager raises a query; the site confirms 75.0 kg and corrects the eCRF; the audit trail records the original, the new value, the user, and the timestamp.