In a new Health Affairs Scholar article, Haeder and Zhu conduct a follow-up study on Pennsylvania insurance directories on the entries that they had identified as being wrong in their first study. The short version of the results is that bad data sticks around for a long time:
We re-surveyed inaccurately listed Pennsylvania providers (n = 5170) between 117 to 280 days after a previous secret-shopper survey. Overall, 19.0% (n = 983) of provider directory listings that had been identified as inaccurate were subsequently removed, 44.8% (n = 2316) of provider listings continued to show at least 1 inaccuracy, and 11.6% (n = 600) were accurate at follow-up. We were unable to reach 24.6% (n = 1271) of providers.
Why does this matter?
Currently, we regulate networks on the basis of their directories. We assume that the data is good and that it accurately-ish represents reality in the very recent pass. However, we know that the data is seldom good. The types of errors range widely from office hours being wrong (an error might be the directory listing hours from 9:00-6:00pm when it is really 9:00 to 5:00) or a doctor not working at an address any more as they completed their fellowship and are now three states over to the office not accepting a particular insurance at all or accepting new patients with that insurance.
Insurers have an obligation to update and correct their directories. Speaking as someone who was responsible for the UPMC Health Plan provider file from 2011-2014, that is an extremely difficult task. When I was working that task, directories were neither a profit center or a heavily regulated activity. We did our best but there was minimal competitive advantage to having a pristine directory. Newer regulations have put some teeth and cash incentives to having better directories but those are not huge.
Haeder and Zhu are showing that these incentives aren’t strong. Less than 1 in 8 identified errors were confirmed to be updated after they were identified. I would bet that if they resurveyed the good locations from their previous study, there would be a decent number of new errors. Data only stays good for short periods of time unless very substantial and expensive efforts are made to keep the data good.
As long as we rely on directory based network adequacy standards, insurers will have incentives to be optimistic about their reported networks. The Centers for Medicare and Medicaid Services is proposing to move towards wait time network adequacy standards. That approach will change insurer network directory maintenance incentives to more accurately reflect who is actually able to take patients. I think that might help a little bit.
* Why yes, I am in that weird liminal space between submitting my dissertation and defending it. More details tomorrow!




