Alec Stein wanted to organize independent software developers to pour through health insurance data to determine what companies were paying to specific providers for particular services. Stein, a data bounty administrator at software company Dolthub, quickly ran into problems.
On July 1, health insurance companies published public, machine-readable files including the negotiated prices they pay to in-network providers and the allowed rates for out-of-network providers. These trillions of prices require sophisticated software to parse and insurers aren't using standardized file formats, making comparisons of their reimbursement rates nearly impossible to make. Moreover, the requirement that insurance companies update their data each month to keep them current has the side effect of making it difficult for data analysts to evaluate information across the industry.
“No one appreciates the scope of the data,” Stein said. Uncompressed data from the five largest health insurance companies dwarf the amount of information held by the Library of Congress, English-language Wikipedia and the entire Netflix catalog combined, he estimated.
Other software developers have encountered similar challenges working with these data, which health insurance companies recently disclosed as part of a broader federal push for price transparency. Regulators have asked for input on additional upcoming transparency requirements. But before officials begin issuing final guidance regarding the advanced explanations of benefits and other transparency requirements, insurers, developers and researchers are pressing the Centers for Medicare and Medicaid Services for clarification on the existing regulations.
The Transparency in Coverage rule, which emerged from the Affordable Care Act, is meant to shed light on the long-secret rates health insurance companies negotiate with individual providers, which can vary greatly. Policymakers intended for providers, patients, researchers—and health insurers themselves—to use this information for their own purposes. Hospitals and other healthcare providers could determine how they are paid compared to their peers, patients could maximize their insurance benefits by shopping around for lower-cost care, and researchers could analyze healthcare spending at a high level.
But these early difficulties with accessing and reviewing these data are hampering the potential for transparency to promote a more efficient healthcare system.
“There was a lot of hope that this data would really shine a spotlight on these payer-provider contract negotiations, but we’re just not there yet,” said Sabrina Corlette, co-director of Georgetown University's Center on Health Insurance Reforms. “There’s just an incredible amount of frustration. CMS really needs to rewrite the requirements here, otherwise it’s never going to achieve the policy goals that were laid out by the administration.”
Big implications for big data
The health insurance price transparency requirement complements the mandate for hospitals. But insurers are widely complying, in contrast to health systems, which have been slow to adhere to the rule. Within 100 days after the regulation took effect, health insurers covering 90% of commercial policyholders had made their negotiated rates publicly available, according to Turquoise Health, a startup aggregating the data to sell to insurers, providers and researchers.
“We've really seen the largest carriers publish significant data. The implications of how much new pricing data is out there are pretty big,” Turquoise Health CEO Chris Severn said. However, insurers have produced so much information that it will take five years for it to be useful to patients, he said.
Too much of a good thing
The files health insurance companies have posted are so large that a typical personal computer can't handle them, said Michael Chernew, a health economist at Harvard Medical School who also chairs the Medicare Payment Advisory Commission. Chernew is leading a Harvard team that aims to use the data to analyze price variations among insurers.
“We're talking terabytes of data, not even gigabytes, we’re at a level above normal claims databases, and they refresh the data every month,” Chernew said. “Even if you thought you had a process for running through it, the way it’s posted can change.”
Humana’s price information, in particular, has posed problems for developers, Stein said. The insurer posted its information in a different file format than what CMS requires. The company also doesn't have sufficient server capacity to allow developers to download more than eight files at once, he said.
“If you wanted someone to do the worst possible job that was technically compliant and contained all the information, Humana did that,” Stein said. “Absolutely legal, but completely annoying.”
Serif Health had to lease multiple servers to accommodate all the available insurance data, said Matthew Robben, co-founder and chief technology officer at the startup, which assists small digital health companies negotiating with large insurers.
Serif Health spent about two weeks working to download Humana's complete dataset, compared with the few days it took to retrieve the equivalent information from other insurers, Robben said. Humana was also the only carrier that did not include required information about how rates differ for inpatient versus outpatient services, he said.
Humana offers support via its website, where outside developers can submit questions and receive responses in a few days, a spokesperson wrote in an email. When Stein tried to use this function to email Humana's developers, his request bounced back.
Across the industry, developers are struggling to work with insurers' information.
CVS Health’s Aetna lists multiple prices, with significant spread, for the exact same services and sites with no explanation why, or when different rates should be applied, Robben said. "I'd love to see CMS offer clarifying guidance to payers for cases like this one," he said. "If there is some kind of tiered fee schedule in place, can we get more clarity on whether it's a credential, location or some other distinction that drives the differences in posted rates?"
Aetna's files follow the format required by CMS and were not designed to serve as a member cost estimator, a spokesperson wrote in an email.
Elevance Health, formerly Anthem, has published several repetitive and redundant files without specifying if the networks listed are local or represent the national reciprocity agreement between Blue Cross and Blue Shield insurers, Robben said. Elevance Health did not respond to interview requests.
The majority of insurers’ files listing out-of-network rates are empty, Robben said.
“We, as consumers of the data, have had to do a lot more engineering and adaptation than we initially thought to work with it. Probably more than CMS hoped with the regulation,” Robben said. “But that's also the reality of the complexities of contracting.” The differences among insurers underscore the need for CMS to host a single directory with rates listed in a standard format, he said.
That would be the ideal, but is unlikely, Corlette said. If CMS stays with the current fragmented approach, it should standardize the index insurers use to explain where information is posted and how to search for specific services. Requiring insurers to adopt common file-naming conventions, standardize the codes associated with individual procedures and organize different services into separate subfolders would help researchers, she said. And mandating that carriers publish smaller files would widen public access to the data, she said.
Making these changes would not require new regulations because it could simply rewrite the technical specifications required of insurers, but the agency would benefit from public input, Corlette said.
Insurers invested a significant amount of time and money into complying with the price transparency policy. CMS should ensure patients can use this information before it adds more rules, said Ceci Connolly, president and CEO of the Alliance of Community Health Plans, a trade group for nonprofit insurers.
Next year, insurers will be required to disclose out-of-pocket costs for 500 common covered services via online, self-service tools. The following year, insurance companies will need to include personalized information for all medical services. Eventually, insurers may also need to disclose what they pay for prescription drugs, although the government has indefinitely delayed that requirement. Many of these No Surprises Act and Transparency in Coverage provisions overlap, so CMS should work to align them and focus on how they will benefit consumers, Connolly said.
“You're sort of piling on requirements here, and it's not clear to us that that is going to be very consumer-friendly,” Connolly said. “It could be very burdensome, and it could be duplicative.”