MRF Processing Notes - October Edition

Matthew Robben
October 20, 2023

Boring title, hopefully not boring content. 

It’s been busy at Serif Health the past few months. Now that we’re all back from HLTH and SIIA, time for another processing notes post! 

The payer data still isn’t perfect, and we’re usually the first to admit that to anyone who asks. We’re seeing more deep dive analytics and broad studies calling out inconsistencies and issues with the data we’ve highlighted in previous posts, and Andrew Wilson and Francois de Brantes’ article in Health Affairs takes a new (highly insightful) angle documenting the difficulty in real-world matching to claims and identifying fundamental gaps in some MRF data given these were adjudicated, actually paid out claims from those networks.

While we’ve heard from payers there are fundamental methodology issues in generating one price for each code for MRFs vs. how claims actually get priced, the issue isn’t missing prices, it’s missing providers. There’s no good excuse for 15% of providers missing from in-network rates files. 

Thankfully, there’s is some good news on macro posting and processing trends to report out this quarter. Many payers we’d highlighted for failing to post have returned to posting monthly, and we’ve seen meaningful improvements in posting size and EIN/NPI count increases came from payers like Aetna, Ambetter, Multiplan, and Priority Health. 

Of course, it wouldn’t be a processing notes post without some lowlights and public calling-outing - Optum and Florida Blue take the stage this month. Read on for details.

Vast majority of plans have returned to posting MRFs monthly

From our ingest pipeline in October, we were able to automatically pull in 159 out of 186 attempted indexes, and manually fetch 18 out of 29 non-compliant (missing a TOC) plan data files. Excellus and Univera are posting on the regular again, along with BCBS Kansas and some other large plans we’ve tagged in prior posts. Out of the October error cases, we were able to shim or correct all but a dozen plans. Meaning, about 90% of plans are posting usable files each month. 

We attempt to report outages and gaps to each plan as we see them, and generally we get a response and a fix in 30-60 days. 

All of the missing postings lately have come from small regional plans, with one big exception. Optum. Despite numerous attempts to contact and inform members of Optum and UHG’s compliance and PR teams, for several months now their postings all return a “PublicAccessNotPermitted” exception. Try it yourself: 

We'll assume its an unintentional misconfiguration...three months running.

This is an obvious misconfiguration of the serving bucket behind their posting. It’s basically a checkbox one IT team member needs to click in their Azure account, but we can’t seem to get through the wall of lawyers and executives at this company to get that button clicked. 

Serif Health maintains a complete library of MRF files from prior months, which we can use as a stopgap to serve data to interested parties. Given only a small percentage of contracts roll over or change membership on a monthly basis, we still have reasonable insights dating back to July. But we’d love to see this get corrected soon, as we’re all flying blind on the current behavioral health payment rates from the largest health insurance company in the country.

Perhaps the employer leadership at General Electric, PGA Tour, State of Ohio, State of TN, Nielsen, HCA, and Hewlett Packard (to name a few of Optum’s plan customers) care and can reach out to their plan compliance officers? Price transparency SME & consultant Brian Cotter has noted publicly that If those plans are self funded, those employers are likely at more risk than Optum for the distribution problem. Solid leads for his new employer-focused MRF verification service!

Wrapping this with a positive - rate limiting, blocking, TiC portal errors, and url masking of TiC postings have reduced overall since July 2022, and seem to have significantly reduced since last quarter. ‘urldefense’ has been removed from many blues’ table of contents files, and the datestamps in the HealthSparq portal we’d been reporting seems to be resolved. 

Only a small handful of plans continue to use portals (cough, BCBS VT -> Change Healthcare, cough) that don’t actually allow for machine readable access. 

I can finally put away the “non-machine-accessible machine-readable-file” joke. Which is good, as it was never that funny. 


CMS Schema violations are common, but not particularly meaningful

A substantial number of payers continue to make schema mistakes in their published MRFs, mistakes which make processing and using the data more difficult than it should be. 

As a validation experiment, we tested in-network files from 110 different plans with the CMS schema validator, by grabbing the first URL out of their table of contents, attempting to download and feed it into the schema validator tool.

Depending on your level of incoming cynicism, the results may surprise you - 61% of these plans had an in-network file that passed schema validation. 

It’s worth pointing out many of the failures are somewhat trivial violations, and some of the payers who pass (Aetna, United) have been directly called out in articles as having substantial amounts of missing data, which IMO is more important than strict schema compliance at a ‘JSON key capitalization’ level. 

Still, schema compliance is not a small or regional payer issue; here’s the failing plan list:

  • AllWays Health Partners
  • AmeriHealth Insurance Company of New Jersey
  • Aspirus Health Plan
  • Avera Health Plans Inc
  • AvMed
  • Baylor Scott and White Health Plan
  • Blue Cross and Blue Shield of North Carolina
  • Blue Cross and Blue Shield of Vermont
  • Blue Cross Blue Shield of Minnesota
  • Capital District Physicians Health Plan
  • CareFirst BlueCross BlueShield
  • Caresource
  • Cigna Corporation
  • CommunityCare
  • Florida Blue
  • HAP
  • Harvard Pilgrim Healthcare
  • Hawaii Medical Assurance Association
  • Hawaii Medical Service Association
  • Health Payment Systems
  • Health Plan of Nevada
  • Healthcare Highways
  • Henry Ford Health System
  • Humana
  • Independence BCBS
  • Kaiser Foundation Health Plan
  • Molina Healthcare
  • Optima Health
  • Optum
  • PacificSource Health Plans
  • Physicians Health Plan of Northern Indiana
  • Piedmont Health Plan
  • Point32Health
  • Priority Health
  • Prominence Health Plan
  • SelectHealth
  • Trilogy
  • Tufts Health Plan
  • UCare
  • UHA
  • University of Utah Health Plans
  • WellSense Health Plan
  • WPS Health Plan

While we’ve shimmed and found ways to work with these payers, basic schema compliance feels like an obvious enforcement step for CMS, and would prevent schema abuse and high processing costs due to the engineering required to clean and normalize each payer.

Individual Network Stats Highs and Lows

Quick reminder that file size is not a good indicator for data quality.  Extremely high volumes actually indicate the data is loaded with ‘zombie rates’, and this volume masks posting errors, missing data points, and gaps in the data, requiring expensive scans and external data sets like claims to find issues. 

As a case study, checkout Florida Blue’s Bluecare network in August:

And what happened in September:

If you’re purely looking at file size, you’ll miss the important takeaway, which is that EIN and NPI counts dropped _substantially_ month over month. This feels like another UHC level ‘oops we lost 92% of our data’. We’ve reached out to Florida Blue to see if this is deliberate, or a technical issue, and are waiting to hear back.

But it’s certainly not all gloom and doom with network sizes. 

Aetna’s HMO in Florida seems to be growing quite nicely and coverage in the MRF is growing with it. Perhaps they’re taking all the providers from Florida Blue?

Rental network operators like MultiPlan are correcting their unreasonably small early postings with realistic, national-scale MRF files. Check out these percentage changes month over month for their MPI product:

And some exchange / regionals are improving their files. Here’s the Priority Health PPO in Michigan, which looks far more rich across the code range:

Here’s Ambetter OK - a small file, but on a good trajectory.


The takeaway from this processing notes post for price transparency stakeholders is a positive one. The increased concerns about utility and accuracy of the data as a whole is increasing transparency (pun intended) about payer data gaps, but it seems like the trendline is positive. 

As backpressure continues to build, we expect payers will invest further in cleaning up and filling gaps in the MRFs.  Here’s hoping for a continued positive trendline throughout Q4!

Image credit: 

As always, if Serif Health can be helpful with accessing or processing price transparency data in part or in bulk, or if you need a backfill for your Optum files, don’t be shy, reach out!