"a Walmart executive admitted that the company has been looking at anonymized customer data and finding that shoppers taking the diabetes drug and its weight loss management counterpart, Wegovy, are buying less food."
Yeah, through insurance. Walmart sells most consumer goods - food, but also clothing, medicine and toiletries, sporting goods (including guns, in some places). Lots of people use their pharmacies, so presumably they're able to look at correlated sales of drugs and food purchases.
That seems ... bizarre. Does something like HIPAA not prevent companies from using pharmacy patient data against their regular sales info on the same person? I get It's supposed to be anonymized data, but how can they figure out who's buying Ozempic vs not for this specific comparison?
Welcome to the world of "clean rooms", often used with tricky consumer data like this [1].
I recently was a Tech PM for a large ad/marketing agency and we utilized them for effectiveness of movie goers for a large studio. Essentially, we wanted to see who saw our ads on social media _AND_ set top boxes _AND_ searched for the movie title in particular Zip codes.
Obviously highly specific data that fingerprints a single user wouldn't be given to us by Meta, Comcast, and Google (first-party data), but we can ship that data to a "clean room" who will venn-diagram it together to get us our ultimate numbers, per Zip code, to find effectiveness/reach.
Wal-mart being a first party with both of their doors (retail and pharmaceutical) presumably can do this all themselves with their own data scientists looking at register receipts.
I’ve spent a bunch of time with HIPAA. It’s a lot less protective than most people think.
Practically (though it’s more nuanced), if data is truly anonymous, they can use it however they want. The challenge is getting data to be truly anonymous. It’s very difficult once you want to identify geographic or regional data. Age, gender, and zip code creates plausible identity in many places.
For HIPAA purposes, there is no such thing as anonymized info if someone can coax out zip code, gender, and birth date. That's the current SotA standard in terms of k-deanonymization.
And yes. BAA's apparently allow companies to throw that shit around like hot cakes. Remember, it isn't actually illegal until someone has been sued for it, and they lose (no settlement).
Not being obese and more healthy brings you much more life quality.
It might be hard to comprehend for some Americans, but not everything in Life revolves around $
I first saw this article in 'The Daily Mail / UK' [0] a few days ago. I thought it was dark pattern advertising for a weight loss drug...
The DM article had pictures of a couple of attractive older women (skinny, the AFTER shot). Perhaps older women who feel overweight being the main target audience. Plus a picture of a (surprisingly) thin Elon Musk (ie; you too can be successful!)
Many of the comments on that page where people saying that buying less was the result of inflation. Something that makes far more sense. Maybe I'm just being cynical because a search just now [1] shows a few other sites parroting the same.
Or maybe Walmart has a vested interest in this drug because it feels like an odd thing to announce (a baseless claim?).
So they are surveilling people they know are on the drug, and they are seeing those people buy less food. Wow. The surveillance bothers me a lot more than the fact Walmart might experience a slight downturn in the bottom line. Here's a non-paywalled recap of the story: https://futurism.com/neoscope/walmart-ozempic-food
It's just data analysis, the same way that other retailers do it for product recommendations. The only difference with more anodyne correlations (eg. people who buy cereal are also more likely to buy milk) is that they're using anonymized pharmacy data.
It's not being shared. Walmart is both the pharmacy and the grocery store. If you use the same account to buy drugs and food from them, they can see that. Nobody needed to give them the data and they aren't giving it to anyone else.
It could be transaction level data (eg. credit card ending in 1234 bought X at store Y).
It's anonymized in the sense that one can't know what medication Joe Blow takes just by looking at the dataset, even if their transaction data is part of it.
If such analysis is concerning, then one can choose to buy their medication elsewhere or pay cash without any rewards card.
