What happens when a virus-based disease spreads stealthily across the world, at an alarming spread rate, without sight of it ending anytime soon?
Relatively unheard of just a couple of months ago, the new coronavirus disease (COVID-19) that emerged from Wuhan has already triggered a global health emergency. This infection outbreak has since spread to more than 150 countries, affecting more than 170,000 cases, and resulting to more than 6,500 deaths.
Globally, the rate of infection has grown at such an alarming rate that many countries have already declared state of emergency, with those hardest hit declaring lockdowns to prevent further spread. Current statistics show a spread rate double that of seasonal flu, and fatality rate more than 10 times of that for seasonal flu.
The increasing global infection figures have shown that COVID-19 is spreading stealthily across the world. Asymptomatic carriers, scarcity of test kits and COVID-19 as a new pathogen complicates efforts in trying to contain the transmission. The increased infection cases will inevitably overwhelm the health system as doctors and nurses are overworked to keep up with new cases, as well as the number of hospital facilities especially intensive care units (ICU). Scientists have started work on finding treatments and vaccines to address the outbreak, however drug discovery is a lengthy process, which is likely to take at least a year.
The challenges posed by the coronavirus is wide-ranging: from early detection, containment and isolation, to mitigation and segmentation, and ultimately to the end of the disease.
Data science in healthcare
We are dealing here with a new coronavirus, far more complex than its predecessors, and with a spread rate (called R0 pronounced as R-naught) almost double that of normal flu. Based on current statistics, SARS-CoV-2 virus which, causes the disease COVID-19, has an R0 of 2.2 which means a single infected person will infect 2.2 others on average. With the advent of data science and advanced analytics, we are also now able to access tons of data, collected from the various facets and phases of the virus and the disease. You can see this data in various forms and shapes in charts, graphs and tables, to help visualise the facts in simpler and more practical ways.
From detecting coronavirus through computerised tomography scans, to more intelligent contact tracing, to analytics-based cluster predictions, to developing powerful drugs that can tackle this coronavirus - big data and artificial intelligence (AI) play a broad role in our fight against COVID-19. Speed is of essence in restraining the outbreak. Contact tracing is imperative to ensure that anyone who has close contact with the infected patients are traced, isolated and monitored for symptoms.
Advancements in Geographic Information Systems (GIS) tools allow the spread of COVID-19 to be modelled through various spatial and temporal scales. Structured mathematical and statistical techniques, such as Bayesian inference methods facilitate the analysis of infectious disease incidence through time and space. A spatial-temporal mapping tool can allow user to visualise and scroll by infectious disease incidences through timeline.
Sankey diagram helps to visualise the flow of patients between various phases of patient journeys. The multidimensional visualisations help to understand the flow and category relationships of patients to specific outcomes. An example is illustrated in a Sankey diagram by WHO, showing patterns of COVID-19 disease progressions in China. The thickness of arrows is indicative of the proportion of patients who died or recovered.
The historical contagion is useful in understanding and predicting the spread among a local population during this COVID-19 outbreak. Identifying ‘super-spreaders’ or super-spreading events among the infected patients can be conveniently studied and visualised from a network perspective. The degree of interconnectedness between cases and infected clusters can be visualised through a network map. A founder of a Singapore’s coding academy had built a network map of COVID-19 cases organised by known locally transmitted clusters within Singapore. Each node represents an infected person and the edge represent the transmission of the contagion through a known contact.
Enter telco analytics
What single device is considered the most personal accessory that caters to people of (almost) all ages and demographies, from the youngest to the most elderly, all genders, societal levels, and ethnicities? And with it, carries nearly all personal, behavioural, mobility and social records of information. Enter the personal electronic devices (PEDs), ranging from smartphones, tablets, smartwatch, fit-trackers and other electronic accessories that people bring around with them.
In performing contact tracing for example, health agencies normally rely on combinations of personal declarations (of affected or suspected person), travel records, CCTVs and police discovery through investigations. These sources of information are not always easy or available to gather and synchronise. This is where PEDs come in handy. Taking mobile phones alone for example, it is estimated that every single person across the world has at least one phone, on average. There are also many countries where mobile phone penetration rate is more than 150 per cent, which means a person has more than one device.
And what does the smartphone carry with you? A typical smartphone will have the owner’s personal details way beyond name, location both present and historical, mobility patterns (smartphone can intelligently know what is home and work location), browsing patterns of which sites or portals were visited, buying patterns, close circle of friends through record of most frequent contacts, social media accounts and interactions, and sentiments through posts and tweets. For a lot of countries, General Data Protection Regulation (GDPR) and other privacy laws prevent telcos from “sniffing” this goldmine of personal and social data, but the capability is always there. Imagine reducing contact tracing and cluster predictions from days or weeks, into a matter of minutes, when telcos are allowed to sniff all available information when national concern and state of health emergency requires it.
The role of big data and artificial intelligence becomes even more relevant in trying to contain the spread of the disease. More than just geolocation and mobility information, the billions of smartphones globally give both historical collection of data and real-time (or near real-time) snapshots that give better insights for analysis and more accurate predictions. More advanced algorithms and computational models can be developed to utilise all these megasets or supersets of collected information in order to get answers to questions, predict several scenarios under varied conditions, and come up with best recommendations and prescriptions.
Simple practical examples can include, but not limited to:
- Quicker assessment of probability of exposure in a given area or cluster through cross-matching of smartphone locations of affected and suspected individuals
- More accurate and detailed time-based historical location information of infected individuals and close circles (potential infections) to establish chains of transmission
- Complementing CCTV and other video records to identify people in particular infection hotspots
- Using database of medical prescriptions and purchase of medicines to trace individuals who are in risk of avoiding quarantine
- Sniffing through social media tweets and posts for potential discovery of unrecorded cases
Advanced analytics and AI can significantly accelerate the data processing required to get the insights, answers and recommendations to handle and address the COVID-19 pandemic - from days to minutes, and with better-than-expected accuracy. Telco analytics can take it a step further in providing deeper insights into people trends and patterns, making intelligent predictions about the disease possible. With responsible guidelines in place, and when it is a matter of national emergency, telco analytics, together with big data and AI, can definitely do things faster and more accurately in containing, mitigating, and eventually ending this global problem. Something to consider.
Albert Nombres, Yam Guan Goh, Teradata