The All of Us research program has a data gap that looks like a consent problem but is actually a conversion problem. According to MedCity News, more than 300,000 patient records are missing not because people refused to participate, but because the data could not be successfully converted into a usable format. That distinction matters: it shifts the focus from policy to engineering.
Consent has often been seen as the main barrier to building diverse biomedical datasets. But in this case, the source argues, the real gap is in the pipeline that transforms raw records into research-ready data. The missing records represent a technical failure rather than a lack of willingness to join the program.
The good news, the article suggests, is that engineering problems are more tractable than trust problems. Fixing the conversion process could recover a large share of those records and strengthen the program's representativeness. The source does not offer a specific technical solution, but it frames the issue as one that can be solved with better systems rather than more outreach.