How AI & Data Science Can Foster More Equitable Distribution of Health Resources During COVID-19

• December 9, 2020
• COVID-19

The COVID-19 pandemic has both highlighted and exacerbated racial inequities. Research by McKinsey’s Center for Societal Benefit through Healthcare and others has found that COVID-19 is disproportionately affecting communities of color. Compared to white Americans, the estimated age-adjusted COVID-19 mortality rate for other American racial/ethnic groups rises sharply. It ranges from 1.5 times greater for Asian Americans to 3.8 times greater for Black Americans.

The concentration of both COVID-19 and communities of color in urban areas provides one explanation. Though it has continued its march into towns and rural areas, 90 percent of COVID-19 cases to date are concentrated in urban areas. Such urban settings also are the home of 84 percent of Black Americans and 88 percent of Hispanic/Latinx Americans.

These populations also have had greater difficulty in getting tested for the virus. Though Black and Hispanic/Latinx Americans are more likely to attempt to get ­­­­­­tested, they are less likely to be tested – just one example of persistent racial and ethnic disparities in healthcare delivery.

In addition, the rapid pandemic-driven shift to telehealth threatens to leave communities of color even further behind in their access to healthcare. A recent study by the Mayo Clinic projects that by the end of 2020 almost all large employers will be using telemedicine as they expand access to virtual care among employees. However, community health centers, which serve a disproportionate share of low-income and racial/ethnic minority populations, have experienced difficulties with rolling out telehealth, and as a result, those populations are less likely to have access to virtual healthcare. This is further exacerbated by disparities in access to stable internet and lower rates of smartphone adoption as a result of the digital divide within and among U.S. communities.

Artificial intelligence can play a role in eliminating such biases and making sure resources are distributed equitably. In cities, the issue will be of growing importance as vaccines could be rolled out as early as the end of this year in the U.S., after Pfizer recently announced that its vaccine demonstrated 90 percent efficacy in clinical trials.

Too many algorithms have built in bias that discriminates against minorities and other vulnerable populations. However, ethical AI, and in particular predictive data models, can help governments creating vaccine distribution plans address these many disparities to ensure that access to COVID-19 testing, telehealth, and ultimately to a vaccine is fair and equitable, provided that public accountability for algorithms is safeguarded and potential bias addressed. In order to plan for and deploy immunization, governments have to be able to distribute vaccinations and testing kits on a large scale, prioritizing those communities that are disproportionately affected by the pandemic.

This is where geographic information systems (GIS) and AI can help to identify and prioritize critical populations, identify gaps in access and alternative distribution models, and manage resources.

GIS company Esri designed examples of COVID-19 vaccine distribution tools that can be used to determine population phases, allocate resources, and select new sites in a sample region. The Vulnerable Population Dashboard can help officials understand regional populations and allocate supplies in line with the demand by identifying geographic areas with relatively higher concentrations of people who may be vulnerable to COVID-19 and plan COVID-19 related public health measures based on their age, physical or behavioral health conditions, access to healthcare services, and social factors.

And, the U.S. Digital Response Team’s online dashboard, created in partnership with the Pennsylvania Department of Health to track the availability of hospital beds and ventilators on a county-by-county basis, could serve the purpose of tracing the supplies of vaccines and testing kits, and use AI and predictive analytics to help predict a population’s behavior and the risk of transmissions on a local level for different sub-populations and adjust supply to address changing hotspots.

Finally, the company Bluedot, which spotted early signals of the pandemic before anyone else, developed an AI-driven infectious disease surveillance system. Their system uses a proprietary algorithm to provide insights on the spread of infectious disease based on the analysis of data collected outside official healthcare sources, such as the worldwide movements of travelers on commercial flights; human, animal and insect population data; and local information from journalists and healthcare workers extracted from online articles. A similar model could be used by cities with the addition of data on vulnerable populations as a vector in AI-driven supply predictions.

Public officials are interested in using these and other tools to ensure an equitable distribution of vaccines. The deputy chief information officer of Los Angeles County, Jagjit Dhaliwal, is partnering with a team of MIT and Harvard students at the MIT Media Lab Global Ventures course that I mentored to brainstorm a solution to ensure equitable vaccine distribution across local pharmacies by leveraging operational and analytical models. Dhaliwal said, “…while the federal/state governments are laying out macro-strategies in partnership with global organizations for effective distribution of vaccines, we need to look at the bottom-top approach too to strengthen the on-ground operational needs of local vaccine provider agencies.” The goal is to create a AI- driven model that leverages patterns from previous epidemic situations in order to prioritize and effectively manage vaccine distribution to high-risk populations.

In fact, a data-mining company, Palantir Technologies Inc., is already working with the U.S. federal government to design the Tiberius vaccine management system. Tiberius would enable the U.S. to allocate vaccine shots to high-priority populations and identify those of the highest risk of infection.

Ultimately, state and local agencies need to efficiently manage the distribution and allocation of resources, including demand forecasting, staffing, service time, patient tracking, follow up, and tracing, and eventually, equitable vaccine access for all. AI and data analytics can help in developing a coordinated and equitable distribution response by public authorities.