I've worked extensively with credit bureau data, and there's a fair bit of misinformation both in the article and the comments. First, bureaus get very little

I've worked extensively with credit bureau data, and there's a fair bit of misinformation both in the article and the comments. First, bureaus get very little detailed information about the episode of care that led to the medical bill: essentially, they'll just get info on when the bill was due and what amount on it is still overdue, plus enough info on the patient to match it to the right person (name, address, phone, SSN if available). HIPAA and similar legislation allows the sharing of billing data like this, and no other PII (or any PHI) gets shared in the process. Second, medical debts have been a part of bureau data for a while now. What's new here is that Experian is trying to build a machine learning model on it to try to sell the model as a new product. I've never worked for any of the bureaus before but know many people who do, and in general, these specialized scores seldom sell well, but since the marginal cost to Experian is low (let a few data scientists at it for a few weeks), they still put out new ones anyway. Even things like new FICO/Vantage scores like the new FICO that's been in the news lately tend to take a while (on the order of years) to make it out into the wild since lenders like to have consistency in their processes as long as possible. Third, it's actually beneficial to most consumers to have medical debts split out of the bureau data explicitly. Most big lenders use raw bureau data much more than the aggregated credit scores, and since the US medical system has so many distortions that correspond more to broken processes rather than a person's creditworthiness, many exclude medical debts from consideration explicitly, and others implicitly as models they build don't often find those attributes as useful as others. All in all, it's good to be thoughtful about how information like this is disseminated and used, but the dire warnings of the article are really not warranted in this case.