In a relational clinical database, the most efficient and normalized structure for data collected repeatedly over time-such as vital signs-is one record per patient per visit.
Each patient will have multiple records, one for each visit when vital signs are assessed. This structure supports:
Time-based analysis (e.g., trends across visits),
Accurate data linkage with visit-level metadata, and
Efficient querying for longitudinal data.
According to the GCDMP (Chapter: Database Design and Build), the relational design principle dictates that data should be stored at the lowest unique level of observation. Since vital signs vary by both patient and visit, the combination of patient ID + visit ID forms a unique key for each record.
Option A (per visit) lacks patient identification, while options B and D aggregate data too broadly, losing temporal detail.
Thus, option C (One record per patient per visit) correctly represents the normalized design structure.
Reference (CCDM-Verified Sources):
SCDM GCDMP, Chapter: Database Design and Build, Section 4.2 - Normalization and Table Structure CDISC SDTM Implementation Guide, Section 5.3 - Visit-Level and Observation-Level Data Structures ICH E6(R2) GCP, Section 5.5.3 - Data Handling Principles