The coronavirus pandemic has fallen upon us and this new pathogen has us scrambling to explore, assemble, and analyze data for quick and yet critical decision-making. The fact that the officially announced estimates keep changing in a wide range is an indicator of data quality and interpretation issues. In the meantime, we continue to neglect some of the important data to address critical needs. For example, healthcare workers, who are assigned to work with coronavirus patients or suspected cases are being sent home after their shifts. Are they transmitting the virus to their family members? Should they be isolated from their families while they are working on the assignments? It is clear that we need to quickly form a strategic data framework to address all the critical needs.
Current coronavirus tests focus on existence of virus’ nucleic acid. The data serve better for diagnostic purpose, rather than epidemiological purpose since virus is removed from the body after the patients have recovered. Due to limited capacity, tests are mostly done among those who are suffering from severe symptoms, leaving out a great deal of mild cases as well as asymptomatic virus carriers and resulting in biased estimation. As testing capacities expand and increasing portion of mild cases and asymptomatic carriers are tested, the characteristics of the tested samples continue to evolve. Modeling using biased data with ever-evolving characteristics is very likely one of the important reasons that official estimates have to be adjusted so frequently and abruptly.
To improve estimation accuracy, regular random sample testing should be done at various levels. In addition to measuring the current incidents of infection, the portions of severe cases should be calculated in relation to the contributing factors such as social, health and industrial and living conditions and assessing the impacts of social distance measures on incidents. This will provide better data for future prediction, thus helping policy adjustments and allocation of resources, which have been in shortage everywhere. As capacities expands further all cases, including those who are asymptomatic and those who were in close contact with the infected, should be tested.
The spreading of coronavirus will likely be impacted by the growing density of people with post-infection immunity in a given population. Compared with transient existence of the virus in the infected, antibodies, especially IgG, stay in the serum for much longer time. Therefore, coronavirus-neutralizing antibodies should be tested in the populations and the data should be included in modeling for future estimation. Antibody testing can easily be administered in population on regular basis using a testing devise as easy to use at home as a pregnancy test and a phone app to report test outcomes. In addition to helping predict future incidents of infection, antibody testing can also help provide comfort to those with positive outcomes, release growing man power to conduct essential services and fuel economic recovery as more people recover from the infection, and expand donor pools for therapeutic plasma.
If pandemic is a nationwide wildfire, healthcare staff are the fire fighters. Given the shortage of PPE supplies and inadequate protection measures, they are at high risk for infection. In the meantime, they have been working at such intensity for long hours endlessly, battling fatigue, which not only reduces their effectiveness but also increase their own probability of infection. We need to monitor virus infection among healthcare workers, the incidents of their transmitting the virus to their family members and the outcomes of adjusting their working hours per shift and duration of their assignment in coronavirus patient care. In addition, we also need to measure the incidents of their transmitting virus to their family members as a lot of them go home after their shift and determine quickly whether we should keep healthcare workers away from their families while they are on pandemic assignments.
Coronavirus is far different from the viruses that human beings had encountered before. The lack of effective virus-suppressing drug and the often abrupt deterioration of patients’ conditions force health care workers to try whatever the method they can come up with to help their patients. Consequently, a lot of measures taken are based upon anecdotal observations without validation. To improve the effectiveness of the treatments for severe cases and of reducing the number of mild cases turning into severe ones, we need to use data and AI to identify effective measures in relation to patient’s conditions and indications of such measures so that therapeutic outcomes can be improved.
Winston's Blog 