Modeling the pandemic risk
How to model the COVID-19 pandemic and forge a path to better risk insurance.
The COVID-19 pandemic disrupted the delivery of medical services in 2020 through lockdowns, canceled or delayed medical visits and procedures, a ramp-up in telehealth utilization in lieu of in-person visits, and hospitals filled to capacity by COVID-19 patients. Disruptions will impact the diagnoses captured for Medicare Advantage members in 2020, resulting in lower 2021 risk scores. These 2021 risk scores form the basis for the risk score projections that Medicare Advantage organizations (MAOs) include in their 2023 bids.
Based on this history, MAOs face unique questions in estimating 2023 risk scores.
MAOs need to be aware of these issues and start planning now for 2023 bids.
The relativity of 2021 risk scores and claims may diverge relative to historical experience. Understated 2020 diagnoses will result in lower risk scores in 2021, while 2021 claim levels for many MAOs are emerging at levels that are equal to or higher than pre-pandemic levels. This one-two punch scenario could result in unfavorable 2021 financial results, which will add to the complexity of 2023 bid projections.
The 2021 risk score disruptions will be MAO-specific. MAOs that were in geographic areas with minimal healthcare service disruptions and active member outreach to ensure that annual wellness visits and specialist visits occurred may have minimal disruptions. Active member outreach could include a smooth transition to telehealth visits when necessary, as long as the telehealth visits meet Centers for Medicare and Medicaid Services (CMS) requirements for diagnosis capture.1 MAOs may see larger impacts if they are in areas with periods of official or unofficial lockdowns, experienced closures, delays, and cancelations of medical services; or provided telehealth visits not eligible for diagnosis capture. Milliman modeled the risk score impact from different disruption scenarios in an October 2020 white paper.2 In the scenarios selected, Part C risk score decreases ranged from 0.8% to 9.0%, though we recognize no scenario will precisely correspond to an individual MAO’s experience. It is also worth noting that the 9.0% scenario represented an extreme case and we expect very few plans to have experienced such a large decrease in risk score.
Impacts may vary by segments of the member population. Access to video telehealth services may be more accessible to younger Medicare members (recent 65-year-olds) than older members and may also vary by income level. In many areas there were also visitor restrictions at long-term care (LTC) and group home facilities during 2020. Members in these settings may have had more difficulty in receiving medical care than other members.
An additional complication is that, just like for the 2022 bids, the 2021 risk scores used in the 2023 bids will not be entirely known by the bid submission deadline because the data submission window for 2021 risk scores has been extended into the summer of 2022. Thus the April 2022 beneficiary file risk scores will not represent the final 2021 risk scores.
There are several analyses that MAOs can perform to estimate the impact of missing diagnoses on 2021 risk scores to support their 2023 bids. The following analyses are ordered from least complex to most complex.
1. Review the pattern of annual incurred medical claims over several years to evaluate at a high level whether claims were materially impacted in 2020. If annual claims were not materially impacted, it could be assumed that 2020 diagnoses (2021 risk scores) may be used to project 2023 risk scores without a pandemic-specific adjustment.
a. Simultaneously, risk scores could also be reviewed over several years to review whether an adjustment is needed.
b. A high-level adjustment could be assumed based on how closely claim disruptions align with scenarios from the Milliman analysis cited above in footnote 2.
2. Project 2023 risk scores starting with 2020 risk scores instead of 2021 risk scores, which assumes the overall plan risk score for 2023 will be consistent with pre-COVID-19 diagnoses. Long-term effects of the pandemic on coding and morbidity should be considered and adjusted for, as needed. Essentially, will 2022 diagnoses return to pre-pandemic levels? Population changes would need to be accounted for separately.
3. Compare January 2022 Monthly Membership Report (MMR) risk scores to January 2021 MMR risk scores to review differences by member. The difference could be used to estimate the impact of COVID-19, which would then be used to project the full year 2021 risk scores to 2023.
a. The January 2021 MMR (as of the original MMR run date, not reflecting the midyear or interim final update) uses July 2019 to June 2020 as the diagnosis period.
b. The January 2022 MMR uses July 2020 to June 2021 as the diagnosis period.
c. This approach assumes similar diagnosis submission patterns for each period, with the only difference between the two periods being actual diagnoses recorded. Depending on the MAO, this assumption may or may not be appropriate, in particular if there were submission issues during 2020 that have since been resolved.
d. The January MMR files are based on diagnoses from July through June, so consideration will need to be given to any impact estimated using this approach, and how that might compare to a risk score impact calculated using a calendar year diagnosis period.
e. You may need to separately account for population turnover.
4. At the member level, compare diagnoses from 2019 to 2020. The impact for 2020 could be estimated by filling in the missing diagnoses from 2019. This assumes missing diagnoses will return by the 2022 diagnosis period, which is the diagnosis period that determines the 2023 risk scores. The analysis will need to consider whether to use the same runout for both years, or to use full 2019 runout. The risk score increase estimated using the full 2019 runout would reflect a combined estimate of remaining 2020 runout (final diagnoses submissions) as well as missing diagnoses. An additional adjustment may be needed to reflect the general aging of the population and conditions that are new over time, as well as to account for population changes.
In any analysis, MAOs need to consider what is expected for 2023. Additional questions for any analysis include:
MAOs continue to face various financial risks because of the COVID-19 pandemic. Deferred or avoided care in 2020 can cause lower risk scores and revenue in 2021, which needs to be considered when developing projections for 2023. MAOs need to start planning for 2023 bids while there is still time to evaluate impacts.
1CMS (April 10, 2020). Applicability of Diagnoses From Telehealth Services for Risk Adjustment. Retrieved October 28, 2021, from https://www.cms.gov/files/document/applicability-diagnoses-telehealth-services-risk-adjustment-4102020.pdf.
2Pipich, R.J., Cross, K., & Bell, D. (October 20, 2020). How far will Medicare Advantage 2021 revenue and risk scores drop if seniors stop going to the doctor? Milliman Insight. Retrieved October 28, 2021, from https://www.milliman.com/en/insight/how-far-will-medicare-advantage-2021-revenue-and-risk-scores-drop.