[Skip to Navigation]
Sign In

Featured Clinical Reviews

November 19, 2020

How to Leverage the Medicare Program for a COVID-19 Vaccination Campaign

Author Affiliations
  • 1Coalition for Epidemic Preparedness Innovations (CEPI), Harvard Medical School, Boston, Massachusetts
  • 2Humetrix, Del Mar, California
JAMA. 2021;325(1):21-22. doi:10.1001/jama.2020.22720

Although the Medicare program is primarily an insurance program, it serves additional functions by virtue of its population-wide coverage (individuals older than 65 years, as well as people who are blind or disabled or who have end-stage kidney disease) and its vast and comprehensive data on beneficiary health care utilization and costs, clinical conditions, and medication use.

Medicare data have been used to examine process and outcomes of care and prevalence of chronic conditions of public health importance, and for infectious disease surveillance, such as influenza, and disease-related outcomes, such as hospitalization and death. Data on Medicare beneficiaries have also been used in disaster preparedness and response to identify individuals who require access to electricity (such as those receiving ventilator support) and, while maintaining privacy protections, to notify public health agencies of their names and addresses in anticipation of and during disasters such as hurricanes, wildfires, and other disaster-related power outages.1

Near real-time Medicare data supported the response to the 2009 novel influenza A (H1N1) pandemic. For example, analysts from the Centers for Medicare & Medicaid Services (CMS) generated weekly reports of hospitalizations and deaths due to pneumonia and influenza and were able to monitor disease trends by age, sex, beneficiary geographic location, and race/ethnicity.2 Once a vaccine became available for the H1N1 influenza virus, analysts were able to link special billing codes for vaccination to beneficiary claim records and to monitor vaccine uptake, including by region of the country and by race and ethnicity. These data supported efforts to enhance vaccine uptake in African American and Latinx populations, whose vaccination rates were lower than White populations. Nonetheless, in general, CMS data have been underused when it comes to public health interventions and surveillance.

Once a safe and effective coronavirus disease 2019 (COVID-19) vaccine is developed, Medicare data could be used to support a COVID-19 vaccination campaign. This support may include operationalizing the prioritization of vaccine distribution, allocating vaccine to states and local jurisdictions based on the number of individuals at high risk of severe disease, monitoring vaccine uptake, supporting pharmacovigilance once licensed vaccines start to be administered, and identifying waning immunity and potential severe acute respiratory syndrome coronavirus 2 infection and adverse events following vaccination. The analytic support could come from CMS or from external entities with access to comprehensive, up-to-date Medicare data.

Because most clinicians and health care entities bill CMS electronically, CMS receives an initial claim for most services within several days of service delivery or discharge. Those claims are then adjudicated to ensure their accuracy, a process that may take up to 6 weeks. Over time, CMS has determined the range of differences between preadjudicated and final claims and is able to adjust preadjudicated claims accordingly for analytic purposes. This enables near real-time surveillance for events. In 2014, the accuracy of these data was assessed for those with claims for electronically dependent medical equipment and oxygen and found to exceed 93%.3

On October 2, 2020, the National Academy of Medicine released a prioritization scheme it had developed, at the request of the National Institutes of Health and the Centers for Disease Control and Prevention (CDC), for allocation of COVID-19 vaccine.4 The framework recommends that health care workers receive vaccine first, followed by older individuals and those with certain chronic medical conditions. It further recommends priority be given to those living in geographic areas identified by a Social Vulnerability Index.

Using CMS Medicare fee-for-service claim data on more than 15 million beneficiaries, including 770 000 confirmed COVID-19 cases, as well as CDC Social Vulnerability Index data, a predictive model was developed for COVID-19 hospitalization and death at the individual and local population levels.5 Although the initial goal of this model, developed by the Department of Defense Joint Artificial Intelligence Center, was to help the National Guard and other military personnel in their support for regions and hospitals affected by COVID-19, such a model could also be used to classify individuals at risk of severe COVID-19 and their priority ranking to receive the initial available doses of COVID-19 vaccine (after health care workers). This may be necessary because after initial release of the vaccine, sufficient doses will not be available to vaccinate everyone, and risk stratification is likely to be necessary.

Either the Medicare program or state and local health departments could use this model and the resultant outputs to notify individual beneficiaries of their risk and recommendations for vaccination. CMS could also notify individuals electronically or by mail to encourage both first vaccination and a second dose, as they do for seasonal influenza vaccination. In addition, it should be possible to link such a reminder to appointment scheduling at a vaccination clinic that can provide the immunizations, or to notify beneficiaries or their physicians about where vaccine is readily available in their community, based on their address. There is precedent for sharing information that identifies beneficiaries with state and local health departments when it is used to save lives; methods to protect the confidentiality of beneficiary status via the Health Insurance Portability and Accountability Act (HIPAA) are regularly used in the emPOWER Program,1 which is updated monthly and identifies people with electricity-dependent medical equipment or home oxygen supplementation and can notify health authorities, provided they can protect confidential information, about vulnerable individuals.

Current Procedural Technology codes for COVID-19 vaccination as well as vaccine administration are being developed for COVID-19 vaccines. Thus, as was the case for H1N1 influenza, CMS should be able to monitor vaccine uptake in close to real time, as well as to monitor immunization rates by region, age group, risk category, and race/ethnicity. This could enable additional efforts to identify geographic areas with high proportions of minority beneficiaries, which have historically lower vaccine uptake rates than White or Asian American beneficiaries, and to strengthen local efforts to increase vaccine understanding, awareness, and uptake.

Monitoring the safety of vaccines is of paramount importance to a vaccination campaign. During the H1N1 influenza vaccination campaign, the US Food and Drug Administration developed systems to link immunization registry data to data from health plans and was able to conduct a real-time safety surveillance program for up to 38 million health plan enrollees (now 60 million enrollees), supplementing the CDC-based Vaccine Adverse Event Reporting System (VAERS) and Vaccine Safety Datalink system.6 Because older populations will be prioritized for receipt of COVID-19 vaccine, CMS is well positioned to act similarly, linking near real-time data on receipt of vaccination to beneficiary claims. Predetermined potential safety signals could be monitored using specific International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes in submitted claims for individuals who did and did not receive a vaccine. Such data could also be used to address concerns about clinical events that are temporally linked to vaccination. In this way, CMS could quickly examine and report on the frequency of such occurrences in the vaccinated and unvaccinated Medicare population. Through Medicare.gov, CMS could also attach a VAERS concerning event questionnaire to its emails for self-reporting of potential vaccine adverse events in immunized beneficiaries, in addition to the clinician-based VAERS system.

Moreover, because COVID-19 vaccines are new, their effectiveness in the community in preventing breakthrough infections that result in hospitalization and death from COVID-19 will be unknown, as will be the effect of waning immunity, especially for older individuals who have decreased immune responses to vaccines. CMS is well positioned to use its data for such rapid effectiveness and waning studies, comparing COVID-19 infection rates, benign vs severe disease outcomes in immunized and nonimmunized Medicare beneficiaries. Hence, Medicare data could be used to create a “virtuous vaccination circle” for Medicare beneficiaries, beginning by identifying high-risk individuals, inviting them to receive a vaccine, recording what vaccination they received, and monitoring the vaccine’s safety and long-term effectiveness.

CMS data are limited to Medicare beneficiaries. Obtaining similar data for all population groups is critical, including if and when vaccines are recommended for children and pregnant women. Complementing this approach with other non-Medicare health care data will be important. However, Medicare beneficiaries are at the highest risk of severe COVID-19 and have accounted for 80% of COVID-19 deaths,7 and the program can generate data on COVID-19–related illness more rapidly than most other systems. With regard to privacy concerns, in emPOWER, HIPAA-protected personal information is limited to those who may be affected by the public health emergency; while most Medicare beneficiaries are asymptomatic, those at highest risk but not yet infected are also potentially affected during the pandemic, as they would be by natural disasters. Similar privacy provisions could apply.

The Medicare program is more than an insurance program and a lifeline for many older US residents. Supporting the public health and vaccination needs in the COVID-19 pandemic should be an important program imperative and could serve the US population as a whole.

Back to top
Article Information

Corresponding Author: Nicole Lurie, MD, MSPH, Coalition for Epidemic Preparedness Innovations (CEPI), 1900 Pennsylvania Ave NW, Washington, DC 20006 (nicole.lurie@cepi.net).

Published Online: November 19, 2020. doi:10.1001/jama.2020.22720

Conflict of Interest Disclosures: Dr Lurie reported formerly serving as assistant secretary for preparedness and response at the Department of Health and Human Services, where some of the ideas discussed in this article were initially developed. Dr Experton reported that funding for the development of the predictive model for severe COVID-19 cited in the article was provided by the Johns Hopkins University Applied Physics Laboratory under a prime contract with the Department of Defense Joint Artificial Intelligence Center.

Disclaimer: The views expressed in this article do not necessarily represent those of the Coalition for Epidemic Preparedness Innovations or Harvard Medical School.

Department of Health and Human Services. HHS emPOWER Program fact sheet. Updated March 30, 2020. Accessed November 12, 2020. https://empowermap.hhs.gov/HHS%20emPOWER%20Program_Fact%20Sheet_FINAL_v9_508.pdf
Salmon  DA, Akhtar  A, Mergler  MJ,  et al; H1N1 Working Group of Federal Immunization Safety Task Force.  Immunization-safety monitoring systems for the 2009 H1N1 monovalent influenza vaccination program.   Pediatrics. 2011;127(suppl 1):S78-S86. doi:10.1542/peds.2010-1722LPubMedGoogle ScholarCrossref
DeSalvo  K, Lurie  N, Finne  K,  et al.  Using Medicare data to identify individuals who are electricity dependent to improve disaster preparedness and response.   Am J Public Health. 2014;104(7):1160-1164. doi:10.2105/AJPH.2014.302009PubMedGoogle ScholarCrossref
National Academies of Sciences, Engineering, and Medicine. National Academies release framework for equitable allocation of a COVID-19 vaccine for adoption by HHS, state, tribal, local, and territorial authorities. Published October 2, 2020. Accessed November 12, 2020. https://www.nationalacademies.org/news/2020/10/national-academies-release-framework-for-equitable-allocation-of-a-covid-19-vaccine-for-adoption-by-hhs-state-tribal-local-and-territorial-authorities
Experton  B, Tetteh  H, Lurie  N,  et al. A multi-factor risk model for severe Covid-19 for vaccine prioritization and monitoring based on a 15 million Medicare cohort. medRxiv. Preprint posted November 3, 2020. doi:10.1101/2020.10.28.20219816
Salmon  D, Yih  WK, Lee  G,  et al; PRISM Program Group.  Success of program linking data sources to monitor H1N1 vaccine safety points to potential for even broader safety surveillance.   Health Aff (Millwood). 2012;31(11):2518-2527. doi:10.1377/hlthaff.2012.0104PubMedGoogle ScholarCrossref
Centers for Disease Control and Prevention. Coronavirus disease 2019 (COVID-19): older adults. Updated September 11, 2020. Accessed November 12, 2020. https://www.cdc.gov/coronavirus/2019-ncov/need-extra-precautions/older-adults.html
1 Comment for this article
Ensure Prioritization Algorithms Don't Disadvantage Underserved Populations
Michael Steinman, MD | UCSF and San Francisco VAMC
The authors propose a superb use of claims data to support fair and appropriately prioritized deployment of vaccine. As these models to identify priority groups are developed, we must guard against biases that are baked into claims data carrying forward to create risk models that end up disadvantaging historically underserved populations.

Recent scholarship has shown numerous examples of this. For example, a recent study found that risk prediction models ascribed incorrectly low risk to minority populations because outcomes were based on health care costs. Lower access to care among minority populations meant that equally sick people from these
populations tended to generate less costs, and thus perversely were judged at lower risk and thus less in need of services.

There is increasing awareness of these issues among researchers. But I fear that even if models are developed with close attention to these issues, they may be misapplied by policymakers in ways that perpetuate the cycle of disadvantage in underserved populations. Thus we must pay special attention to making these tools "idiot-proof" - building and disseminating them in such a way that facilitates their appropriate use and makes their misapplication more difficult.