As the COVID-19 pandemic continues to spread across the US there are emerging patterns for the disease burden coming to light. African-American and Latinx populations are suffering a disproportionate amount of the disease burden and deaths. It is often said that pandemics are a window into society and amplify health disparities and social suffering patterns that have been present prior to the outbreak. In this blog post we take a look at some of the analytical tools and challenges that are emerging to help evaluate the impact of the pandemic as well as shed light on the policy responses that will be required going forward. The stimulus efforts to date have not done enough to directly impact communities bearing a disproportionate amount of suffering. Any future healthcare stimulus funding, or HITECH Act 2.0, efforts need to focus on tools and resources that can build a bridge from health IT to public health and precision public health efforts targeted at the communities hardest hit.
African-American and LatinX populations are bearing a disproportionate COVID-19 disease burden. Health disparities, environmental inequalities and lack of insurance are exacerbating health inequalities through risk profiles, access to care and treatment.
Missing Data. CDC has not released disaggregated data that can be used in data analytics efforts. Much more work needs to be done to acquire better data and fill in the holes in datasets so that accurate analytics of the burden of disease and overall epidemiology are understood better.
Environmental factors are rising in importance. Preliminary research is revealing that air quality can dramatically impact risk of death from COVID-19. Exposures to PM 2.5 is cumulative over the lifespan and many low-income communities are made more vulnerable due to the damage already done.
Future stimulus funding for health IT, a HITECH 2.0, needs to take into account social determinants and broader approaches to strengthen health systems. The pandemic exposes and exacerbates existing inequalities. Our ability to contain outbreaks and resume economically active lives means that health equity needs to be front and center. Gaps in the system impact the whole.
In recent days an increasing amount of data on the racial disparities in mortality rates has come to light as the pandemic proceeds across the country. Preliminary data from Florida indicate that minority populations are getting hit harder than white populations, in Michigan where African-Americans make up 15% of the population they constitute 33% of cases and 41% of deaths. In Louisiana the initial data are indicating that 70% of deaths are African-Americans who comprise only 32.4% of the overall population.
Many states and cities are not releasing data on the ethnic and racial outcomes, however, we know that there is great variation in hospital quality across the country and low-income communities and minorities will assume the burden of poorer quality of care. A large number of social determinants ranging from employment and housing to underinsurance and closer living quarters for extended families come into play as well. Furthermore, there is still a lot of missing data on the disaggregated population impact of the pandemic.
Machine learning tools are rapidly being configured and retooled to serve the needs of hospitals, public health and social services as the pandemic proceeds. We will see even more reliance on some of these tools should we head into a second, and possible third, COVID wave later this year through winter. The pandemic is putting more pressure on the system to innovate around precision public health, or the ability to customize interventions at the community level given their distinct risk profiles, vulnerabilities and case-loads.
Open Data For AI-Driven Pandemic Tools
One of the first major tasks in developing tools that can help us understand the contribution of race and socio-economic status to the disease burden of coronavirus will be access to data. The White House Office of Science and Technology Policy has created the COVID-19 Open Research Data Set for data scientists to analyze with the goal of creating tools useful for clinicians and public health professionals. Kaggle, MIT SOLVE and the Allen Institute for AI are all offering either data sets or data challenge platforms to contribute to the pandemic response.
One of the biggest challenges is obtaining race and ethnicity disaggregated data on coronavirus cases. To date, the CDC has not provided this data so it has been left to states and local public authorities to provide these data. Even the New York Times is contributing to the cause through their county-level data tools. Black Demographics is providing race disaggregated data where available. It is a very urgent issue that better data are collected on who is impacted and where so that better tools can be developed to address this long-term issue. Massachusetts has recently announced they will be publishing disaggregated data very soon as part of their daily case count reports and we expect others to follow soon.
Figure 1: Racial Disparities in COVID-19 Pandemic (Source: BlackDemographics.com)
Most of the drivers of the racial disparities in health outcomes are the product of decades of social policies and systemic racism. However, as stimulus packages are passed by Congress it is imperative that the health-centric AI and the data analytics tools we have are used to inform policies and direct resources to communities in the most effective way to drive better outcomes.
Air Quality as Risk Factor and AI Solutions
Where you live has a huge impact on exposure to air pollution and we can see clear race and income gradients linked to such exposures. However, Air pollution has been a somewhat neglected aspect in public health despite efforts such as the Clean Air Act. Unfortunately, early data from the COVID19 pandemic are beginning to shed light on the likely effect of air pollution in creating a pool of immuno-compromised individuals and damage to the respiratory and circulatory systems that can increase the risk of acquiring or dying from the virus.
Researchers at Aarhus University (Denmark) and the University of Siena (Italy) have found an association of higher air pollution exposures to higher rates of mortality in Lombardy and Emilio Romagna where death rates have reached 12%. Even economic historians have found similar associations between proximity to coal powered electrical plants during the 1918 Spanish Flu Epidemic and differentials in mortality rates across cities. Researchers at Harvard just published a study demonstrating that even very small increases in PM 2.5 exposures as low as one microgram per cubic liter can contribute to a 15% increase in death rates.
Jvion’s COVID Community Vulnerability Map has become quite useful for public health researchers in identifying risk factors at the community level. Air quality in areas such as Louisiana’s “Cancer Alley” where racial disparities in air pollution exposures and COVID risks currently appear to be quite pronounced. The Jvion tool is one example of a tool that could be used for creating the ground truth that community organizations and data analysts need for resource allocation efforts and programs.
Other Social Determinants and Steps Toward Re-thinking A “HITECH ACT 2.0”
Public health practitioners, data scientists and social service organizations will need to bring localized analytics of social determinants and COVID-19 interactions to drive policy changes going forward. For example, density of population, mobility and lack of insurance are key drivers of who gets infected by COVID-19. These factors come together when many people lack jobs that allow them to stay at home for mitigation efforts. Reuters has been publishing maps showing mobility rates over time as the pandemic proceeded that offer another window into WHO can adopt social distancing measures (Figure 2). Lower income individuals were more likely to continue working outside of the home after mitigation efforts began because many do not have jobs that can be accomplished remotely. This means additional social welfare policies may need to be targeted to enable these households to survive under lockdown conditions.
Figure 2: Mobility and Income (Source: https://graphics.reuters.com/HEALTH-CORONAVIRUS/USA/qmypmkmwpra/ )
Chronic Disease Burden and Race, Gender and Income
The American Heart Association Center for Health Metrics and Evaluation offer a large number of visualization tools that illustrate the associations between heart disease and COVID-19 outcomes as well as data on hospitals and health system capacity. Bringing together environmental data, mobility and co-morbidity data may provide more granular understandings of COVID-19 outcomes and resources that enable a more precision public health approach to the pandemic.
Access to diagnostics and population screening will need to have a social justice component (eg. insured vs. uninsured) to ensure that we can equitably identify hotspots and allocate resources. This will enable all citizens to return to work more quickly. AI developers will need to be extremely careful in rooting out bias that has resulted in algorithms that adversely impact women and minorities as we have seen in the recent past.
At Chilmark Research we are increasingly discussing the need for a new HITECT ACT 2.0 that moves beyond EHRs for providers and build a health technology infrastructure that includes post-acute care, assisted living, public health and other health centric organizations. Surveillance systems, investment in community and social services, databases on community resources, geographic information systems, and AI can all play a role. We also need the human element of cooperation that moves beyond health IT systems in isolation to urban and regional planning, transportation, local economic development, the Environmental Protection Agency, and beyond. Intersectoral interoperability, albeit challenging, will be necessary for future pandemics.
Yes, there will be more.