Benchmarks That Mirror the Market
FAIR Health employs recognized statistical methodologies to create our FH® Benchmarks. In addition to our in-house staff of mathematicians and statistical and clinical experts, we consult with independent healthcare economists and statisticians—all leaders in their fields—to gain an external perspective on our methodologies. In fulfillment of our mission of transparency, we make our methodologies available to the public.
Creating Benchmark Products
To create FH Benchmarks, we organize the claims data we receive by procedure code and geographic area. We group our data into 493 geozips—geographic areas typically based on the first three digits of a zip code or group of zip codes. FAIR Health employs a statistical outlier methodology to exclude any extremely low and extremely high values that might otherwise distort the distribution of data. Most of our benchmarks are based on a recent 12-month window of claims. Our FH® Charge Benchmarks and FH® Allowed Medical are refreshed every six months.
FH Benchmarks Methodology
FAIR Health uses two methodologies to create FH Benchmarks, based on how much data are available for a procedure code in a given geographic area. These methodologies are applied to provider charges to create our FH Charge Benchmarks and to the allowed amounts that reflect values negotiated by payors with in-network providers to create FH Allowed Benchmarks.
- Actual methodology. If there is a sufficient number (“frequency”) of actual charges or allowed amounts for a procedure in a geozip, the actual charges/allowed amounts for each procedure code/geozip combination are arrayed from lowest to highest to determine percentiles. A percentile is a position in a distribution of values below which a specified percentage of the values fall. For example, in a distribution of 100 data points, the 70th percentile is the value in the 70th position in the lowest-to-highest array of values. Thus, 70 percent of the values are equal to or lower than the 70th percentile value and 30 percent are equal to or higher than the 70th percentile value.
If the frequency of actual charges or allowed amounts for a procedure in a geozip is insufficient, the benchmarks are derived by
using the values for all procedures in a procedure code group within the geozip. First, the charge/allowed amounts are
“normalized” by dividing each charge or allowed amount by the code’s relative value. Next, the results for all procedure codes in the group are arrayed from lowest to highest and assigned to percentiles,
as described above. In the final step, the relationships between codes are reestablished by multiplying the percentile
values by each code’s relative value. The derived methodology enables the creation of benchmarks for codes for which
there are very few or no data.
FAIR Health is converting our FH Charge Benchmarks product line, which previously offered two lines of modules, one based on the actual methodology with the derived methodology used only for low-frequency codes and the other exclusively using the derived methodology for all codes. As a result of the conversion, which is being implemented in stages, eventually most charge benchmark products will be offered solely based on the actual methodology, with the derived methodology used for low-frequency codes.
Learn more about the methodologies FAIR Health uses to create benchmarks.