CDC’s v-safe text mining is laughable
CDC’s V-safe Analysis: Key Findings
This article criticizes CDC’s analysis of v-safe free-text responses from COVID-19 vaccine recipients. Key points:
- CDC only analyzed free-text responses by mapping them to already known, benign symptoms from their checkbox fields
- The analysis ignored potential serious adverse events that weren’t part of the original checkbox options
- There were approximately 2,500 unique serious VAERS reports from v-safe users, averaging 10+ symptoms per report
- The CDC’s methodology used outdated text mining approaches (simple string matching) rather than modern NLP techniques
- Multiple serious adverse events that were in clinical trial documents were not included in the analysis