When vaccine access is prioritized for the most disadvantaged communities, it improves both social utility and equity — even when such populations have strong vaccine hesitancy. An international research team co-led by Prof. Pan HUI from the Hong Kong University of Science and Technology (HKUST) and HKUST(Guangzhou), Prof. James EVANS from the University of Chicago, and Prof. Yong LI from Tsinghua University, reveals the key to breaking the dilemma of multiple ethical values from a data-intelligent epidemic model which can predict COVID-19 curves in US metro areas accurately.
Classical epidemic models make strong assumptions about population mixing, i.e., people in an area mix homogeneously and thus have equal infection/fatality rates and equal chances to spread the virus. This is clearly not the case in the COVID-19 pandemic. The research team instead designed a model that explicitly incorporates mobility behavior and demographic differences to capture the diverse COVID-19 risks associated with different communities. The joint incorporation of human mobility data and demographic structures, both on a neighborhood level, allowed the team to depict more realistically how different subpopulations mix. For example, it’s crucial to notice that low-income families are worse off in this COVID scenario because they have to sustain their original level of mobility for the sake of livelihood, which exposes them to greater risk. Consequentially, they’re more likely to spread the virus, making them a key group to vaccinate (compared to many white-collar workers who can work from home).
The study produced two key findings. First, it underscores the importance of jointly considering mobility behavior and demographics when designing vaccine prioritization policies. Most existing vaccination schemes are designed based solely on age or a combination of age and occupation. The US Centers for Disease Control and Prevention has a social vulnerability index that they use in some regions to prioritize vaccines. Still, it fails to capture behavioral data and the differential likelihood of spreading and being exposed to COVID. By contrast, the proposed model dramatically improves the ability to target those disadvantaged persons that make the most out of limited vaccine resources to achieve the greatest benefit for everybody. The researchers also note that their smart model used coarse behavioral data — dispelling concerns about privacy leakage. In fact, many aggregate data sources can be used to fuel epidemic models without much worry about privacy or other issues.
The second major takeaway is that the authors argue for vastly boosting vaccination campaign budgets for disadvantaged, vulnerable populations. That includes both for outreach and for the risk that vaccinations could go to waste as the uptake may not be as swift among populations with greater vaccine-hesitancy, some with good historical reason. But more funding among these populations — who move about and mix with others in the community — goes a very long way to keep everyone safe. The advantage is persistent even if the vaccine hesitancy of the most disadvantaged populations is five times that of the better-off populations.
Prof. Hui, Chair Professor of Emerging Interdisciplinary Areas at HKUST and Chair Professor of Computational Media and Arts at HKUST(Guangzhou), said, “Epidemics not only threaten our society as a whole but also exacerbate inequalities that may tear society apart, as disadvantaged communities face more obstacles in reducing high-risk contacts and seeking healthcare. In situations of scarce medical resources, it is crucial to smartly distribute them so society can make the most out of them. To this end, our study presents a possible pathway to improve vaccine distribution decisions with data intelligence. Hopefully, data accumulated and lessons learned from the COVID-19 pandemic can help us better prepare for future challenges.”
This work was recently published in the scientific journal Nature Human Behaviour.