Coding differences in Medicare Advantage plans led to $33 billion in excess revenue to insurers, study finds

An analysis of differential coding patterns between Medicare Advantage (MA) and Traditional Medicare (TM) plans estimated how much coding differs between insurers and how much extra revenue insurers receive as a result.

The analysis found that because of coding differences, the average MA risk score in 2021 was 0.19 higher than the average TM risk score and MA plans received an estimated $33 billion in additional revenue, with $13.9 billion, or 42% of the total, going to UnitedHealth Group. The study is published in Annals of Internal Medicine.

MA plans are paid more for sicker members and less for healthier members, providing an incentive for MA plans to report as many diagnoses as legitimately possible. Prior reports have shown that MA plans report diagnoses more intensely than TM, and past research has found large differences in coding between MA and TM.

According to the authors, however, no research to date has estimated the extent to which each MA insurer codes differentially or the amount of extra revenue each insurer receives.

Researchers from University of California San Diego and colleagues studied data from the Centers for Medicare and Medicaid Services (CMS) Chronic Conditions Data Warehouse (CCW) from 2015 to 2020 and the Master Beneficiary Summary Files from 2015 to 2021 to provide insurer-specific estimates of the effects of differential coding on risk scores and revenues. The core analytic sample included 697 contracts that were active in 2021.

To measure contract-level differential coding, the researchers analyzed the effects of “persistence” and “new incidence” on risk scores. Persistence is defined as the percentage of members coded with a diagnosis in year 1 that persisted in year 2, and new incidence refers to the percentage of members with a diagnosis in year 2 that was not recorded in year 1.

The researchers calculated persistence and new incidence statistics for members who were continuously enrolled in a single MA contract or continuously enrolled in TM for 24 months for five cohorts of beneficiaries: 2016–2017, 2017–2018, 2018–2019, 2019–2020, and 2020–2021.

They used the average persistence and cumulative new incidence statistics for each contract to estimate the effect of differential persistence and new incidence rates on the 2021 risk score. The researchers identified the top 10 diagnostic groups that account for virtually all of the difference between MA and TM risk scores and calculated the average persistence and cumulative new incidence separately for them.

To estimate the effects of differential coding and payment received by MA plans, the researchers assumed MA plans do not adjust their bids in response to differential coding.

With this assumption, differential coding results in additional payment to plans because the rebate that the plan receives is larger if the plan codes intensely.

The researchers found that the average MA risk score was 18.5% higher than the average TM risk score. For the top 10 diagnostic groups, persistence in MA averaged 78.1% compared to 72% in TM, and cumulative new incidence was 46% in MA compared to 33% in TM.

The average MA risk score was 0.19 higher than it would have been if MA and TM had identical persistence and new incidence rates. Persistence and new incidence rates varied across insurers, with UnitedHealth Group’s average 2021 risk score 0.28 higher than it would have been if persistence and new incidence had been at TM levels, substantially larger than the MA industry average of 0.19.

Differential coding resulted in an estimated $33 billion in additional payments to MA plans in 2021, and UnitedHealth accounted for $13.9 billion of that total.

Differential coding resulted in a $1,863 increase in revenue per UnitedHealth member, substantially greater than the industry average of $1,220. Because the effects of differential coding vary across insurers, any MA payment policy reform targeting differential coding would have disparate effects across insurers.

An analysis of differential coding patterns between Medicare Advantage (MA) and Traditional Medicare (TM) plans estimated how much coding differs between insurers and how much extra revenue insurers receive as a result.

The analysis found that because of coding differences, the average MA risk score in 2021 was 0.19 higher than the average TM risk score and MA plans received an estimated $33 billion in additional revenue, with $13.9 billion, or 42% of the total, going to UnitedHealth Group. The study is published in Annals of Internal Medicine.

MA plans are paid more for sicker members and less for healthier members, providing an incentive for MA plans to report as many diagnoses as legitimately possible. Prior reports have shown that MA plans report diagnoses more intensely than TM, and past research has found large differences in coding between MA and TM.

According to the authors, however, no research to date has estimated the extent to which each MA insurer codes differentially or the amount of extra revenue each insurer receives.

Researchers from University of California San Diego and colleagues studied data from the Centers for Medicare and Medicaid Services (CMS) Chronic Conditions Data Warehouse (CCW) from 2015 to 2020 and the Master Beneficiary Summary Files from 2015 to 2021 to provide insurer-specific estimates of the effects of differential coding on risk scores and revenues. The core analytic sample included 697 contracts that were active in 2021.

To measure contract-level differential coding, the researchers analyzed the effects of “persistence” and “new incidence” on risk scores. Persistence is defined as the percentage of members coded with a diagnosis in year 1 that persisted in year 2, and new incidence refers to the percentage of members with a diagnosis in year 2 that was not recorded in year 1.

The researchers calculated persistence and new incidence statistics for members who were continuously enrolled in a single MA contract or continuously enrolled in TM for 24 months for five cohorts of beneficiaries: 2016–2017, 2017–2018, 2018–2019, 2019–2020, and 2020–2021.

They used the average persistence and cumulative new incidence statistics for each contract to estimate the effect of differential persistence and new incidence rates on the 2021 risk score. The researchers identified the top 10 diagnostic groups that account for virtually all of the difference between MA and TM risk scores and calculated the average persistence and cumulative new incidence separately for them.

To estimate the effects of differential coding and payment received by MA plans, the researchers assumed MA plans do not adjust their bids in response to differential coding.

With this assumption, differential coding results in additional payment to plans because the rebate that the plan receives is larger if the plan codes intensely.

The researchers found that the average MA risk score was 18.5% higher than the average TM risk score. For the top 10 diagnostic groups, persistence in MA averaged 78.1% compared to 72% in TM, and cumulative new incidence was 46% in MA compared to 33% in TM.

The average MA risk score was 0.19 higher than it would have been if MA and TM had identical persistence and new incidence rates. Persistence and new incidence rates varied across insurers, with UnitedHealth Group’s average 2021 risk score 0.28 higher than it would have been if persistence and new incidence had been at TM levels, substantially larger than the MA industry average of 0.19.

Differential coding resulted in an estimated $33 billion in additional payments to MA plans in 2021, and UnitedHealth accounted for $13.9 billion of that total.

Differential coding resulted in a $1,863 increase in revenue per UnitedHealth member, substantially greater than the industry average of $1,220. Because the effects of differential coding vary across insurers, any MA payment policy reform targeting differential coding would have disparate effects across insurers.