===== Real-world validation =====

**Real-world [[validation]]** refers to the process of evaluating a [[model]], [[intervention]], [[device]], or [[strategy]] under actual usage conditions, outside of controlled or [[experimental]] [[setting]]s.

==== Common contexts ====

  * **Medicine**: Validation of clinical scores (e.g., [[HERMES-24 Score]]) using real-world data from hospital records or patient registries.
  * **Artificial Intelligence**: Testing models on noisy, unstructured, or biased data typically encountered in operational environments.
  * **Pharmaceuticals**: Post-marketing studies to confirm drug efficacy and safety in broader, more diverse populations.
  * **Engineering**: Field testing of devices, software, or systems under variable and realistic conditions.

==== Comparison with controlled validation ====

^ Feature                ^ Controlled Validation           ^ Real-world Validation                        ^
| Environment            | Laboratory or clinical trial    | Routine clinical practice or real conditions |
| Data                   | Clean, selected                 | Noisy, heterogeneous, and complex            |
| Objective              | Internal validity               | External validity (generalizability)        |