How mathematical models can predict Covid's next step
Mathematical modeling of the Covid-19 spread can inform our level of urgency and equip us to anticipate non-obvious, second-order effects, some of which can be mitigated with proper preparation and timely interventions. For this to happen, policymakers and scientists must work closer together.
Here are some of my current thoughts and suggestions.
“If one had told me that in two weeks, we would have created 500 new ICU beds or completely reorganized our hospital system, I would have said, okay, you are crazy, and it's happening now” is what Dr Giacomo Grasselli said a few days ago (see video) about the Italian’s response to the Covid-19 outbreak in Lombardy.
One can only but imagine Giacomo’s unfortunate realization that a wealthy region like Lombardy would be struck by this virus so hard and that drastic measures and actions were needed.
Are we prepared?
Public health policy decisions during a potential crisis are difficult as they must balance a zoo of scientific, budgetary, social, and political considerations. In an ideal world, each of these elements should have been considered well in advance and in a transparent fashion before reaching a decision or implementing a specific policy. In practice however, the reaction to the Covid-19 outbreak has to date been one of cautionary prevention. Almost all affected countries have announced measures that restrict travel and impose social distancing; a strategy that will almost surely have a strong social and economic impact as they disrupt people’s lives and business, but that will also minimise the number of fatalities and flatten the curve thus not overwhelming their already stretched healthcare resources while slowly but steadily building up herd immunity.
What's our strategy?
Already, officials all over the world like Dr Grasselli are faced with evermore difficult situations and questions. For example, can contact tracing and testing translate into population-level virus containment effects, and if so, would its scale-up have sufficient impact to justify the added cost? Is screening individuals at airports an effective method of infection control? (btw it wasn’t for the H1N1 pandemic). Can we evaluate the expected epidemiological and economic impact of different strategies for scaling up Covid-19 screening and testing (e.g., centralized, at airports, or at individual clinics), and can we estimate the required ICU bed capacity in each country/region given the observed reproduction number (R0), incubation and recovery time of the virus?
Such questions can be, and have been, addressed effectively using mathematical modeling methods that connect dynamic SEIR (susceptible - exposed - infected - recovered) models to specific intervention strategies and their socio-economic costs and benefits. See for example some early results by Hellewell, Kucharski and Cowling et al (2020).
A case for mathematical models.
Unfortunately, such mathematical modelling methods are rarely constructed in some parts of the world because they are perceived to be “too complex”, “too dependent on assumptions”, and “not accurate enough”. These misconceptions stem mostly from a tradition of insufficient communication between public health officials and expert scientists in academic environments.
It’s a trust and science communication problem.
As a scientist myself, I am personally compelled to passionately promote quantitative, data-driven and systematic methods that can help us take the right sequence of public health policy decisions. How else can one evaluate a priori the comparative effectiveness and cost-effectiveness of possible interventions at a population level? How else will we make the right choices and avoid taking unnecessary and ineffective actions?
Through my experience in modelling complex systems (see my publications page), there rarely exists a one-size-fits-all solution. Each country should thus develop or adapt mathematical models to fit their own requirements, circumstances, and peculiarities. Understanding the effectiveness of control measures in different settings is paramount for understanding the dynamics of the outbreak, and the likelihood that transmission can eventually be contained or effectively mitigated. Moreover, once a model is constructed, it can iteratively incorporate updated data, leading to better estimates and highlighting existing weaknesses in the model assumptions and in the available data so that it may also be used during the next epidemic/pandemic outbreak.
OK, but how?
For mathematical models to be useful for decision makers, they must be both relevant and methodologically sound. According to Knight et al. (2016) developing a useful model requires one to identify:
a set of useful questions and their epidemiological contexts,
a framework through which these questions can be addressed,
the parameters required to address the specified questions,
the empirical evidence available to inform those parameters.
Each of these 4 steps necessitates early engagement and transparent communication from local, national, and/or international policymakers themselves. They are the ones who should inform scientists and academics of the key public health questions to be considered. They are also the ones who can provide access to accurate data and lay out all possible public instruments, measures, and resources available as ingredients to the intervention.
Similarly, the modelers should guide those with less methodological expertise and explain their results and mathematical insights in a transparent and accessible manner so not to be perceived as a ‘black box’ that is susceptible to manipulation and is too difficult to understand but as a useful and meaningful tool for making informative decisions for the benefit of society.
Let’s be smart about Covid-19
The last pandemic (H1N1 in 2009) lasted about 20 months and killed about 0.5M people. The current pandemic (Covid-19 in 2019) will last X months and kill Y people.
There are several non-biological factors that make the Covid-19 pandemic different from all previous ones and we should think carefully what role these can play in its effectiveness and how they should be integrated into mathematical models and utilized in our interventions to tackle the virus.
- Mobile telephony global positioning system (GPS) data and location services data from social media could be used to monitor and inform outbreak control in real-time.
- Remote work, telemedicine, e-learning, online (food) shopping and social media can facilitate for social distancing while not drastically impacting work, schools, hospitals, food, and friendships.
- Our arsenal of Artificial Intelligence models will now be challenged to predict further outbreaks, identify best practices and potential weaknesses in our critical infrastructure, and even accelerate our search for a vaccine.
In summary
While governments have ramped up efforts to improve readiness and invested heavily in R&D, Covid-19 has taken us by surprise and demonstrated that we can never be too prepared. In addition to cautionary prevention measures, quantitative scientific methods do exist and should be engaged by policymakers to create tailored responses that can contain and mitigate the outbreak. Each country/region has its own characteristics and should not blindly copy/paste generic blanket solutions.
References and useful articles
Kucharski, Adam J., et al. "Early dynamics of transmission and control of COVID-19: a mathematical modelling study." medRxiv (2020).
Cowling, Benjamin J., and Gabriel M. Leung. "Epidemiological research priorities for public health control of the ongoing global novel coronavirus (2019-nCoV) outbreak." Eurosurveillance 25.6 (2020).
Knight, Gwenan M., et al. "Bridging the gap between evidence and policy for infectious diseases: How models can aid public health decision-making." International journal of infectious diseases 42: 17-23, (2016).
Hellewell, Joel, et al. "Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts." The Lancet Global Health, (2020).
Bill Gates "The next outbreak? We’re not ready" – TED talk 2015 www.youtube.com/watch?v=6Af6b_wyiwI
Microsoft Interactive map of confirmed cases: www.bing.com/covid
Read the blog in LinkedIn here: https://www.linkedin.com/pulse/how-mathematical-models-can-predict-covids-next-step-orestis-georgiou/