Hacker News Reader

Comment by rscho | Hacker News Reader

The funniest part is that this data is actually mostly (and often completely) useless for the stated purposes of statistical analysis. Routine clinical data collection is of abysmal quality, but many buyers don't see the full extent of the catastrophe.

It'll be much more useful for legal and not-so-legal purposes. And last but not least: insurance. As an aside, privacy enforcement for medical data is much easier through fingerprinting tethered to a NDA.

View on HN

I completely disagree. A lot of the data is standardized already, such as lab values and other so-called "structured" data, and can be used almost immediately without much cleaning. This is most of the routine data collection. The largest remaining part is patient notes, which are free form text but still contain plenty of useful information searchable by keyword or extractable by human curators.

You are correct that if doctors could be more strategic about how they collect info it would make the job of the data scientist much easier. This is especially true in radiology and pathology.

Yeah yeah, I've heard the same story many times. So where are the fantastic new discoveries made on mass lab statistics, then? Because epidemiologist have had access to this same data since a long time already, and they are not any less capable than industry "data scientists". The saying goes: "90% of diagnoses are made on patient history". Most of labs require context for interpretation, so the vast majority of what is collected is statistical noise.

What does more strategic mean? Doctors aren’t data collectors. Their job is to provide patient care. If you want data collected, design a system that doesn't add to documentation burden.

Reply on news.ycombinator.com