metadata

license: cc-by-nc-sa-4.0

Primary and Secondary Outcomes Manually Annotated Corpus

Corpus :

86 COVID-19 Randomized Controlled Trials articles. For each article, a part of the full text or abstract supposed to contain information about Clinical Trial Outcomes have been extracted using Entrez API , a XML parser for the Entrez article format, and then multiple regular expression to detect outcome sections. The goal was to annotate entities and relations useful for the detection of discrepancies between clinical trial registration and publication, which includes :

Change in timing of the assessment of the primary outcome
New primary outcome introduced in the paper
Registered primary outcome not reported in the paper
Registered primary endpoint reported as secondary outcome in the paper
Registered secondary outcome reported as a primary outcome in the paper

Annotators :

2 Master students in Epidemiology and Statistics annotated Entities and Relations using brat. A third annotator (PhD Student in NLP, with basic knowledge about clinical trials) did the adjudication.

Associated resources

The code used to process this corpus is available publicly on github