Raw payer rate files are not reliable for contract benchmarking, because 91.8% of the negotiated rates in them belong to providers who do not perform the service. A November 2025 study in Health Affairs Scholar by David Muhlestein checked 3,433,560,471 negotiated rates from 61 insurers and found 3,153,469,476 of them were ghost rates.
The files still hold usable numbers for a small set of codes. MGMA's December 2025 poll of 207 medical groups found that only 18.0% use them in payer negotiations.
What is a ghost rate?
A ghost rate is a negotiated rate attached to a provider or group that does not deliver that service. Muhlestein's example is a psychiatrist who has a listed rate for a heart transplant. Federal Transparency in Coverage rules require insurers to post negotiated rates in machine-readable files, so every posted rate counts in the totals.
The study covered 61 insurers: three national commercial insurers and 58 Blue Cross Blue Shield insurers. The median insurer had 84.3% ghost rates. The range ran from 12.6% at the cleanest insurer to 97.3% at the most affected.
By subtraction, 280,090,995 of the 3,433,560,471 rates, or 8.2%, were not flagged as ghost rates. Fifteen insurers fell in the 80% to 85% band, the most common result.
How the share changes by level and by code
The ghost share depends on how the data is cut. Muhlestein measured it three ways, and the table shows the results.
| Measure | Ghost share | Insurer range |
|---|---|---|
| All negotiated rates (3,433,560,471 rates) | 91.8% | 12.6% to 97.3% |
| Provider-to-billing-code pairs (117,682,739,557 pairs) | 95.4% | 16.8% to 98.6% |
| Top 100 codes, group level | 70.3% | 6.3% to 89.8% |
| Top 100 codes, provider level | 75.5% | 5.3% to 89.2% |
Source: Muhlestein, Health Affairs Scholar, November 2025, 61 insurers.
The top 100 billing codes make up 56.5% of claims but only 1.1% of negotiated rates, according to the same study. For those codes, the group-level ghost share is 70.3%, against 91.8% across all rates. A practice that limits its review to its own top 100 codes works with a cleaner slice of the data, though 7 in 10 rates there still need checking.
The median insurer had 95.7% ghost provider-to-code pairs. At the provider level, 112,297,416,974 of 117,682,739,557 pairs were ghost pairs.
What else makes the files hard to read
The Peterson-KFF Health System Tracker described the data problems in a February 25, 2025 brief by Gary Claxton, Lynne Cotter and Shameek Rakshit. Turquoise Health, a firm that processes the filings, handles about 300 TB of price transparency data each month. Turquoise Health also estimated that as much as 60 percent of reported rates may be implausible.
The brief tested four insurers in New York City with 26 inpatient diagnosis groups, 40 institutional procedure codes and 3 professional procedure codes. For one knee replacement code, the files listed rates under 79 practitioner classifications. One diagnostic radiologist showed 11 unique negotiated rates across 22 tax IDs at 21 addresses.
An AJMC analysis from December 2025 tested completeness for 3 national insurers. It looked at 3 physician specialties (cardiology, family practice and orthopedic surgery), the 100 most common codes in each, and all 771 inpatient diagnosis-related groups. It found physician and hospital outpatient data more complete than inpatient data.
How much do real prices differ?
Real variation is also large, so a single rate from a file can mislead. A February 2025 Health Affairs Scholar study of transparency data covering about 270 million insured people found that mean insurer prices for CPT 99213 ranged from $82 to $115. Mean diagnostic colonoscopy prices ran from $1,481 to $2,686.
The ratio of the 75th to the 25th percentile price was 3.1 for diagnostic colonoscopy and 2.1 for emergency department visits. Across states, outpatient prices varied by 688% and inpatient prices by 174%. The correlation between an insurer's inpatient and outpatient facility price level was only 22%.
A rate that looks low for one insurer in one state tells a practice little about what that insurer pays elsewhere. Local, code-level comparisons carry the most weight. The Muhlestein study also shows why: the cleanest insurer had a 12.6% ghost share and the most affected had 97.3%, so the same method gives very different results from one insurer to the next.
How many groups use the data today?
MGMA polled medical groups on December 30, 2025 and published the results on December 31, 2025. Of 207 applicable responses, 18.0% use the rate data in payer contract talks, 46.0% do not, and 36.0% are unsure.
The groups that use it set and defend target reimbursement rates, and benchmark at the CPT code level. MGMA listed the barriers for the rest: lack of awareness, an unclear payoff, too few tools, time or skills, third-party contract management, and the technical weight of machine-readable files.
A practice with five to ten high-volume CPT codes can start small. Pull the group-level rates for those codes, set them next to the practice's own paid amounts from remittances, and mark each code where the two differ by a wide margin. Codes with a wide gap go to the top of the contract review list.
That work does not need the full 3.4 billion rate dataset. It needs a short code list and the practice's own paid-claims data. The 36.0% of groups that were unsure about the data could run this test on one insurer and one code first.
Sources
- Health Affairs Scholar, High prevalence of ghost rates in transparency in coverage data (Muhlestein), November 2025
- MGMA Stat, Using TiC negotiated-rate data to negotiate smarter payer contracts, December 31, 2025
- Peterson-KFF Health System Tracker, Challenges with effective price transparency analyses, February 25, 2025
- Health Affairs Scholar, Understanding health care price variation: evidence from Transparency-in-Coverage data, February 2025
- AJMC, Price Transparency With Gaps: Assessing the Completeness of Payer Transparency in Coverage Data, December 2025
Talk to us
Fifteen minutes about your practice and the systems you run.