257 Background: Hospital Episode Statistics (HES) are routinely collected data describing National Health Service (NHS) hospital visits in England, with procedure & disease codes. This study, embedded in STAMPEDE, aimed to build a model using HES, linked to primary medical records & trial case report forms (CRFs) to identify progressive disease events (PDEs), including skeletal-related events (SREs). Methods: Analysis of 5 STAMPEDE patients (pts) in 2 stages (data to Jul 16). 1: Detailed manual note review of 3 pts’ PDEs were compared to HES & CRFs to build model. 2: Used model to use HES to identify possible PDEs in 2 pts, verified by note review & compared to CRFs. Created algorithm rules to identify PDEs per 8 week interval plus further analysis of HES coding to find SREs. Results: Prostate cancer PDEs coincided with clustering of HES events. HES found 4 PDEs omitted from CRFs but missed 2 (total PDEs: HES 10, CRFs 8). HES found a false positive CRF PDE. Compared with note review HES missed 4 PDEs (false negatives), with 2 missed & 2 upgraded to PDEs post-standard query procedures, plus HES found 3 false positives (1 STAMPEDE treatment & 2 delayed treatments post-PDE). Hence HES found 71% of PDEs in note review (HES 10, note review 14).CRFs found 57% of PDEs compared to note review (CRFs 8, note review 14). Hence HES found 14% more PDEs than were recorded in CRFs compared to note review. HES identified 4 additional SREs not recorded in CRFs but missed 2. Conclusions: Hospital record review revealed site staff may miss reporting major clinical efficacy outcome events on CRFs, especially nearer end-of-life. HES successfully identified most PDEs (often found as a cluster of SREs), plus additional trial events not reported on CRFs compared to note review and as predicted HES & CRFs found less PDEs. PDEs & SREs missed from CRF recording can be identified in HES.This confirmed use of HES to detect PDEs is feasible. HES-identified events have potential as a primary data source when subsequently verified by standard data queries. Future work will test this model prospectively in the forthcoming BladderPath trial. It may offer a superior, cost-effective method of primary data collection compared to traditional CRF recording.