The term “flattening the curve” has burst in popular terminology with the onset of COVID-19. The search expression emerged in early March, especially as the pandemic began affecting North America. Flattening the curve refers to a situation where the daily rate of increases in infection or deaths in a given country or sub-national locality is no longer increasing. If the COVID-19 country curves are indeed expected to be bell shaped, the flattening of the curve should be the precursor to quashing the curves depending in part on the quality of public action.
COVID-19 curves are widely available on-line via a region of global tracking, as well as national databases. This data blog tries to dive a little deeper into the current COVID-19 trajectories for ten selected Asian countries, each with somewhat different starting points in terms of exposure, as well a timing and mix of corresponding public action. Our sample is not intended to be representative, but serves to explore some comparative approaches to visualization as a way of framing more granular country level analytics. Our sample includes the more advanced economies of China, Malaysia, S. Korea, Singapore, as well as the East Asia emerging economies Indonesia, Myanmar, the Philippines, Thailand, and Vietnam, and the large and diverse setting of India as a comparator.
COVID-19 curves have to-date focused on confirmed infections and death, as well as recoveries. Depending on the state of testing in a given country setting, these series may significantly under-report exposure. Timelines for COVID-19 curves are either presented from the day of the first exposure, or beyond some threshold like 100 deaths.
The first panel compares absolute confirmed infection and deaths for our Asia-10 countries of focus. The graph shows the prominence of China, as well as the fact that it was the first country to be exposed.
The next set of graphs present the graphs on a per capita basis, mapped in days since the first exposure. Based on the available data, China reported its first case on January 22, 2020, now over 80 days ago. On a per capita basis, Singapore has been the most exposed on the basis of confirmed infections. This data must of course be interpreted with significant caution, as it may also reflect the country’s higher capability and public actions to test suspected cases. The mortality class shows the relatively high toll on this basis experienced by South Korea. Although the steep curves for countries such as Vietnam show the toll being imposed on those countries. These mortality rates are of course a function of the degree of local contagion, the quality of healthcare, and the age structure of the populations.
For a different view, we can also present the graph based on some threshold, for example 100 deaths. This number is in some way arbitrary, but also for some view of what infections or deaths can become a more major issue for public policy or taxing the health case system.
The Asia-10 dashboard visualizations presented are limited to national summaries. As the case of China suggests, however, both infections and deaths have been highly concentrated in spatial terms (or in terms of first or second . In China, the epicenter has been Wuhan. The spread of public actions has been truly national. For a drill-down visualization, see the more in-depth review of available evidence for the India sub-observatory link.
The visuals of course raise more questions than answers, especially as the situation remains so fluid. Both the policy makers and citizens of these countries are currently being overwhelmed by a flood of data, much of it requiring validation. The challenge is therefore how to overcome this infodemic risk.
The Google Trends capability shows the intriguing power of big data text analytics to get an leading sense of what issues the wider populations are picking up. The source however is limited to reference to specific text searches in english. Similar approaches have been adopted to text mining to Twitter, including machine translation. But text mining for our Asia-10 could also begin to look behind the curves more systematically to break down public action based on continuous monitoring, as traditional lists risk going stale in a situation as fluid and rapidly developing as set out above. For example, text mining allows searching for instances where contagion populations have been raised in terms of slums or prisons.
This data blog has looked only at the metrics of infections and deaths. As countries in the region do manage to flatten their curves, attention will shift to socio-economic disruption and recovery metrics. At the same time, all countries will be subject to close monitoring in terms of whether some countries do not appear at risk of “second waves” of infections.