This checklist defines the minimum requirements for datasets used in patient narrative generation. All items must be met for successful ingestion and processing, including draft or interim data.
1. File Format & Structure
☐ SAS (.sas7bdat), CSV (.csv), or Excel (.xlsx) format
☐ Single header row; no extra headers, footers, or notes
☐ Single rectangular table; file opens without warnings
2. Identifiers & Keys
☐ Patient identifier present and populated (e.g., USUBJID)
☐ Event/record identifier present where applicable (e.g., MHSEQ, AESEQ)
☐ Identifiers are non-missing, stable, and consistently formatted
☐ No unintended duplicate keys
3. Column Names
☐ No duplicate column names
☐ No leading/trailing spaces or special characters
☐ CDISC variable names used consistently
4. Data Type Consistency (Critical)
☐ One consistent data type per column
☐ No mixing of text and numeric values
☐ Numeric columns contain numeric values only
☐ Text columns remain text (including IDs)
5. Date & Datetime Fields (High Risk Area)
☐ Single consistent format per column
☐ Preferred ISO 8601 format (YYYY-MM-DD or YYYY-MM-DDThh:mm:ss)
☐ Partial/unknown dates follow a consistent convention
☐ No mixed date formats or free text
6. Missing Values
☐ Missing values represented consistently
☐ No mixing of blank, NA/N/A, UNK, or '.' within a column
7. Controlled Terminology
☐ Expected controlled values used (e.g., Y/N, M/F/U)
☐ No unexpected free-text values in coded fields
8. CSV / Excel-Specific Checks (If Applicable)
☐ CSV: UTF-8 encoded; consistent delimiter
☐ Excel: No merged cells; values only (no formulas)
☐ One dataset per worksheet
9. Draft Data Confirmation
☐ Draft/interim data permitted
☐ Structure, identifiers, and data types are stable
Datasets that do not meet these minimum requirements may result in ingestion failures, processing errors, or delays in patient narrative generation.