Almost every day during the summer I was on a Zoom call with the COVID 19 epidemiologists’ taskforce, reporting on the performance and the details of the epidemiological surveillance machine learning tool that my team and I had built. Our work, on top of targeting high risk travel profiles, mainly served as a surveillance and early warning system for a country that decided to open its borders on July 1st, in order to save as much of the tourist season as possible, given the centrality of the tourism sector to the Greek economy, while keeping the pandemic under control.
This tool traces its research origins to MIT where, together with my advisors Asu Ozdaglar and John Tsitsiklis, I spent most of my PhD years tackling the problem of allocating curing resources on a network epidemic, under the regime of full visibility of the epidemic progression. Adapting the allocation decisions online, as the epidemic progresses offers significant performance improvements hence underlining the need for constant learning and dynamic programming when dealing with such a dynamic phenomenon. This idea turned out to be the foundation of our team’s collaboration with the Greek government: collecting information dynamically and in a targeted way so that you make the most “bang for your buck” when testing at the ports of entry is naturally limited, hence leading to dynamic and pre-emptive political decisions for the protocols to be followed regarding visitors to Greece from any particular country.
The tool or system we introduced in Greece this summer combined real time testing results from prior tests performed at Greece’s point of entries to learn the risk profiles of travelers, from particular countries, and optimize the allocation of tests over time. This operation required the cooperation of different kinds of dedicated civil servants: doctors and nurses at the points of entry, firefighters and policemen who helped with operations, three different teams of IT developers to manage the different technological aspects, professors who met with me daily to provide insight and modeling advice, three different Ministries managing the different components, from procurement of tests to managing the labs to coordinating operational and staffing decisions.
I cannot recall a time after the Athens Olympics of 2004 that so many different groups of people and institutions had seamlessly collaborated on a single project in Greece. What made this happen? We were given a place to stand and a lever, and, at least from my perspective, we did move the freaking world. Let me tell you the story of what brought us here.
From me, it all started on March 21st, when Prof. S. Tsiodras, head of Greece’s COVID 19 epidemiological taskforce and one of the most impressive individuals I have ever met, said on TV that the lives of our elders were at risk, “It is our mothers and our fathers”. This guy is fighting for our mothers and our fathers --- and you can tell from his eyes that he is actually fighting. On April 30th, 2020, I Googled the name of Greece’s Prime Minister, “Kyriakos Mitsotakis”, and sent an email to the first address that popped up, entitled “In case you need help with Epidemics Analytics”. Three hours later I received an email to chat and chat we did. We talked about the need for analytics in managing this crisis and agreed on a potential project (disclaimer: this project is pro-bono for my team and myself). The chat concluded with a “Let’s do it”. Having grown up in Greece, I would have bet my fortune that we would never “do it”.
Looking back, I admit I was a fool to be a pessimist. I spent the next month meeting with every single member of Greece’s COVID-19 team: the epidemiologists, the public health officials, the Civil Protection taskforce. We spent sleepless nights designing our strategy to open the borders and giving shape to our system, having deep technical discussions with the scientific team, creating contingency plans for all operational scenaria and ironing out technical glitches and potential fails. The scientific discussions with the international caliber committee of epidemiologists (S. Tsiodras, P. Lagiou, G. Magiorkinis, D. Paraskevis, C. Chatzichristodoulou) were critical to the development of the project and to building the confidence that the added value from this system would not be just “fancy math” but could produce actionable insights and the much needed control over the country’s borders.
It was not all bells and whistles. Micro-politics were always there, the sarcastic attitude “This is Greece not the US” would halt development for a few days. But then General Secretary of Civil Protection Nikos Hardalias would impress upon the group that there will be no compromises in terms of adhering to mine and my team’s adoption of exacting technical standards. Similarly, incredible individuals such as my colleague Prof. Pagona Lagiou made the earth tremble when she talked to the President of the Hellenic Republic to ensure that the project would have the full support of everyone involved during those initial stages. And then there was the anonymous, with the nickname “ccm.tgsd” (standing for “cc me to get s*** done”), who ensured that micro-politics will not stand in the way of the common good.
The day after the project went live, I decided to fly to Greece to overlook all the components of the system since every step was critical for the accuracy and the efficiency of the whole system. I started from the bottom and spent two days at the Athens Eleftherios Venizelos airport next to the doctors and nurses who took samples and the firefighters who were checking the paperwork and “decoding” the test/no-test decision. I was there from 6am to 7pm and I saw the same three firefighters working with the arriving passengers for the whole 11 hours, shattering my stereotype of Greek civil servants spending the last half hour looking at their watch counting down to the 8-hour mark. When I asked one of them, Vasilis Katsiadas, where he is getting all this energy, the answer struck me “Our COVID-19 success story has been the only big win of our generation. I will do everything to keep it this way.”.
I ended up staying in Greece for the whole summer, until end of September, exactly because such projects need constant overlook and tuning while the analytics need to be translated to the decision makers so that tough decisions are made on time and with confidence. Physical presence did make a difference and I would like to advise all of my colleagues that would like to help out in different aspects of Greek governance to actually be present physically. It does make a difference, and besides it is a win-win; it is the summer after all. Of course, for such a thing to happen one would need the backing of their home-institution, the Marshall School of Business at the University of Southern California in my case, that gave me complete support both financially and practically from the very beginning of this adventure and I am extremely grateful for that.
In terms of the policy outcome, in my judgment Eva proved to be a valuable tool. It enabled Greece to adopt PCR (i.e., COVID 19) test requirements in time, hence prolonging the tourist season way beyond other European, tourist-dependent countries. On the technical side, the benchmark would be improvement compared to random testing, that we indeed significantly improved over. I sincerely think that the true value of such a system is that it offered a -unique in the world- snapshot of the true, unbiased incidence of COVID 19 (percentage of positive cases per population, i.e. country-specific cohorts of visitors to Greece) hence flagging to the Greek authorities, about two weeks prior, an upcoming epidemic outbreak originating from visitors from a particular country. This early warning, in turn, allowed Greek authorities to bar entry to visitors from high risk countries while letting in visitors from low risk ones during the summer season.
What am I trying to say with this story? We were given a place to stand by the leadership: K. Mitsotakis, N. Hardalias, the Public Health Department of the Ministry of Health but we also had an enormous lever: the Sotiris, Pagones, Vasilis of our country. When the political will is combined with the amazing talent that our country possesses in all levels, from the incredible scientists of the epidemiological committee, to the over-qualified IT developers and engineers and all the workers on the ground, miracles can happen.
Kimon Drakopoulos is an Assistant Professor in the Data Sciences and Operations department at USC Marshall School of Business.
The team which he led comprised by himself, Hamsa Bastani, Vishal Gupta and Jon Vlahoyannis developed Eva, a tool based on machine learning to distribute Greece’s scarce testing resources, for Covid 19, at the country’s borders based on visitors risk profile, so that early warning be provided to Greek state authorities as well as surveillance of the counties from which visitors originated.