4 ms·
> Though few patients appeal coverage denials generally, when UnitedHealth members appeal denials based on nH Predict estimates—through internal appeals process
by tekacs 3y ago
> Though few patients appeal coverage denials generally, when UnitedHealth members appeal denials based on nH Predict estimates—through internal appeals processes or through the federal Administrative Law Judge proceedings—over 90 percent of the denials are reversed, the lawsuit claims. This makes it obvious that the algorithm is wrongly denying coverage, it argues.
The '90% error rate' in the title sounded like a pretty serious twisting of some numbers so I scanned the article for where it's quoted from ^, which... makes a mockery of the title.
In case it's not obvious and to save parsing that, 'few people appeal' (which makes sense given the internal appeals process or Administrative Law Judge proceedings). Then, /of those cases/ 90% are overturned.
- If 50 / 10,000 cases and 45 were overturned or 500 / 100,000 cases appealed and 450 were overturned... would all meet this '90% error' mark.
- Similarly, 'overturned != error', because it may be as simple as a policy on United's side to not fight super hard and allow overturns (for any number of reasons, including just saving costs on the proceedings, it needn't be altruistic).
Did Ars used to be better than this, or were they always like this on some issues?
The /business/ side of medical insurance isn't really a topic that I know enough to weigh in on, but hopefully people can read the above sentence and parse what it's actually saying for themselves before reading the article.
Irrespective of how I/we feel about the outcomes, this headline is awful.
- michaelmrose 3y agoOf the only claims that have been adjudicated 90% were found to have been erroneous. Without analysis there is no particular reason to believe that other denials were more justified. Few doctors on average make nonsensical requests for patient care while insurers issue spurious denials as a matter of course. You dont need to know the total number of denials instead you need a neutral party to analyze a random sample of denials. We could do this as matter of course and tie failing an adequate benchmark to ruinous penalties to company and CEO.
- hayksaakian 3y agoI think the OP is suggesting that hypothetically speaking, people would only go through the hassle of appealing if they were pretty sure they would win to begin with
- kelnos 3y agoI think that logic is just as faulty as the assumption that 90% of the un-appealed claims would also be overturned. I suspect many people just don't know that they can appeal. Those that do might think it's too difficult to do so, or believe it requires some specialized knowledge to do properly.
- vikramkr 3y agoAnd this is a perfect example of the type of conversation that happens when the correct answer is that we don't know the answer but everyone keeps talking in circles pretending there's a way to know with any certainty other than testing all (or a carefully chosen random sample) of the other denials and getting the actual data. Whenever there's a disagreement where both sides seem reasonable it means that both sides are wrong because the correct answer is that the information present is inadequate to distinguish. All the potential reasons for things going one way or another are also just hypothesis to test since the gut feeling could be right and the reason wrong, and just getting a percentage on the rest isn't enough to figure out why that percentage is what it is
- blitzar 3y ago> testing all or a carefully chosen random sample of the other denials and getting the actual data We should build an Ai to test this.
- michaelmrose 3y agoFunny.
- miki123211 3y agoFor a model whose only output is allow/deny, 90% error rate is pretty good, you can just flip the answer and then you have a 10% error rate. If you know the first thing about AI (or even statistics in general), that headline sounds very suspicious.