Using Publicly Available Health Plan Pricing Data For Research And App Development

“We used Elevance Health’s in-network pricing data and medical claims data. We focused on in-network prices because these are the prices negotiated by the payer, and out-of-network providers represent a small percentage of use for planned procedures. The formatting of the released data files follows the specifications provided by CMS. We limited our analysis to one state, Colorado, and one month, August 2022, to illustrate the concepts. [..]

We split the prices into two components, the professional price and the facility price. Notably, these two components may not be the only costs associated with a procedure as there are often other costs from associated items such as prescription of painkillers or sedatives, other related services that occur on the same day, and costs associated with downstream complications. We did not assess pricing data for these “additional” components.

[..] we present highlights of our findings and recommendations that follow from each point.

The Majority Of Listed Prices Do Not Have Associated Use

In the pricing data, providers have listed prices for procedures with zero use during the claims data period. Only about 1 percent of unique NPIs [National Provider Identifiers] in the Colorado pricing file have a claim for a colonoscopy (current procedural terminology [CPT] code 45378) during the study period. In the pricing file, limiting to the state of Colorado, there are 34,659 unique NPIs with a listed price. In the claims data, there are 360 unique NPIs with at least one claim for diagnostic colonoscopy. [..]

Recommendation #1—If possible, researchers and data scientists should confirm actual use via claims. If not possible, use provider specialty field to limit the prices to only those who perform a procedure of interest.

Not All Providers Listed In A State Operate In That State

[..] there were more than 34,000 NPIs (representing nearly 6.0 million price observations) that listed a colonoscopy price, but only 9,667 that had a Colorado ZIP code (representing 3.1 million price observations).

There are a number of possible reasons that many of the providers listed are not located in Colorado; for example, a hospital group may have locations in multiple states and therefore have prices for all providers that are part of that hospital group in all states in which the hospital group operates. About 28 percent of the NPIs in the Colorado state pricing data are associated with a Colorado ZIP code. This feature of the data is likely not unique to Colorado, although the percent likely differs from state to state.

Recommendation #2—Use ZIP code associated with NPI when looking at prices by geography to accurately represent prices in a given region.

The Distribution Of Use Affects The Mean Price Observed

In the pricing file, every price is observed once; however, use does not follow that pattern. In claims, we see a higher percentage of colonoscopies with higher-cost providers. For example, the mean outpatient hospital professional price in the pricing file is $408.10, while the mean price in claims is $454.03.

Recommendation #3—Supplement the price files with usage data if possible and weight the prices by useIf no usage data are available, caution should be taken when reporting summary statistics from the data, and full distributions should be presented.

The Total Cost Of A Procedure Often Consists Of Professional And Facility Components

Facility prices do not exhibit the same characteristics as professional prices. In both data sources, there are fewer unique facilities and fewer prices per facility, compared to the number of professionals and prices per professional. One reason professional providers may have more prices is that a provider may practice in multiple locations or multiple places of service. The total cost to consumers for the procedure usually is the sum of the professional and facility component.

Recommendation #4—Differentiate between professional and facility prices when making comparisons. Don’t use price data alone for episode-level inferences as just adding together the mean professional and facility prices may not equal the mean episode-level cost because the unique combination of facility and professional is important.

Place Of Service Can Have A Large Impact On Prices

The Elevance Health pricing files contain prices for all place-of-service codes, independent of whether they are observed in claims. Looking specifically at professional services, we see that all claims occur with one of four place-of-service codes (11, 21, 22, 24), although there are prices listed for many more place-of-service codes that do not have any use during the study period. For diagnostic colonoscopy, 98.5 percent of claims occur in two places of service—on campus-outpatient hospital (22) and ambulatory surgical center (24). In both the claims and price files, prices are lower at ambulatory surgical centers compared to outpatient hospitals, a fact that has been reported in previous research. Furthermore, we observe a higher number of prices negotiated in the outpatient hospital setting yet more procedures occur in the ambulatory surgical center setting, another example of how use patterns cannot be inferred from the pricing data.

Recommendation #5—When working with price data, limit analysis to places of service that are procedure-specific. Stratify results by place of service where possible.

Stratification By Place Of Service Is Difficult Using Facility Prices

Looking at the facility prices and claims, four service codes (11, 21, 22, 24) are observed in claims with 99.6 percent occurring in service codes for outpatient hospital and ambulatory surgical center (22, 24). [..] Looking at the distribution of costs from the claims file, we can see that claims with service code 22 are higher cost, and claims with service code 24 are lower cost. The pricing file contains prices distributed across both service codes (22, 24) as would be expected but is only associated with service code 22 in the pricing file.

Recommendation #6—When working with price data, limit analysis to places of service that are procedure specific. Stratify results by place of service where possible. Use publicly available NPI data to determine place of service when working with facility data.

A Single Provider Can Have Multiple Prices For The Same Procedure

About three-fourths of professional NPIs have more than one price for a colonoscopy (CPT code 45378) in the on campus-outpatient hospital setting (service code 22) in the pricing file (exhibit 7). These price differences within NPI can arise from several sources: such as differences in plan, contracts, fee schedules, and network.

Recommendation #7—Don’t use price data alone for episode-level inferences, as patient-level insurance characteristics can also impact the price of a procedure.


Descriptive statistics derived from the pricing data do not fully capture the intricacies of the out-of-pocket price faced by a consumer. As academic research organizations begin to study and analyze price transparency data, it is important to use additional data such as claims, from partner teaching institutions, professional providers, or other third-party claim sources to supplement payer machine-readable price files. Use data from claims give the best representation of which NPIs are performing a given procedure. Without data related to use, limiting data to NPIs in a given geography and specialty can give a better representation of the prices of procedures being performed.”

Full article, D Cullen, J Romine, R Cobb et al., Health Affairs Forefront, 2023.5.19