Measuring the Potential Spread of the Coronavirus (COVID-19) Pandemic

Introduction

COVID-19, the topic on the minds of billions globally. What could even be thought of as a short-term shock a little over a month ago is now being reframed as an economic readjustment of a scale no one can confidently predict, though many still try. Just this week the World Bank published an estimate for economic growth in 2020 for East Asia and the Pacific in the range of -0.5 percent and +2.1 percent, as compared to an estimated 5.8 percent in 2019.So, the region that is home to the world’s second largest economy (China) could go into recession, or it could yet put up reasonable growth figures despite this complete global economic disruption that is COVID-19.

Where COVID-19 is concerned, there is still high uncertainty. Maybe it’s peaked in China for good; maybe it hasn’t yet. It certainly hasn’t peaked for the hundreds of millions who live in Western Europe, the United States, and the Middle East. And, once it has, what then? Will it resurge come September, October, November, as working hypotheses suggest is possible? The scientific community simply lacks sufficient data to make reliable predictions, notably including information about the virus’ reproduction and incubation period, the proportion of asymptomatic cases, and not only the timing and strictness of government responses but the willingness of people whose lives we’re trying to save to abide by these largely unprecedented restrictions.

Using available data on health care systems and risk factors around the world, we decided to examine the crisis from a different angle. We wanted to look at the underlying framework of nations—the age of the populations, strength of health care systems, and more—to examine how ‘at risk’ individual countries are to the spread of coronavirus.

In this article, we would like to share our key findings and the resulting COVID-19 country level risk assessments as well as methodology and data that stands behind our calcualtions.

Index Methodology

Our goal: Measure the susceptibility of a country to the spread of the COVID-19 pandemic based on statistical data that reflects the quality of healthcare systems, availability of healthcare resources, demographic trends, economic interconnectedness, and readiness to shift to digital interactions.

‘Susceptibility’ in this case is a composite qualitative state, and that susceptibility reveals itself through (i) the number of confirmed COVID-19 cases, (ii) the number of COVID-19 related deaths, and (iii) COVID-19 death rates. No one of these measures can be taken separately to measure a country’s susceptibility.

The number of cases or deaths alone can’t say how severe the disease is if we don’t know the death rates. For example, the number of cases and deaths due to COVID-19 is incomparably low relative to the ordinary seasonal flu. By the same token, if we know only the death rate, we have no sense of the disease spread.

We also know that data lies. What’s the saying? “Lies, damned lies, and statistics.” Whoever said it, you can be sure that with COVID-19, too, the statistics are biased. Not only does the number of confirmed cases depend heavily on the number of tests actually given, but the capacity to produce and deliver tests is unequal across countries. The age and socioeconomic bias of those tested also varies within communities and across countries.

You may be thinking, “at least we know the number of deaths.” Do we? Information on the cause of death is not always clear or clearly specified for patients who had COVID-19 and suffered from chronic diseases.

Based on the above understanding of susceptibility, we constructed our index as the weighted average of individual indicators that have positive correlations with the number of confirmed COVID-19 cases and/or number of COVID-19 related deaths and/or death rates.

We collected our set of indicators for 155 countries based on the methodological notes summarized by health experts and economists from health and epidemiology research institutions.2,3,4 The original set of indicators that we tested for correlation with the COVID-19 pandemic and their final weights are shown in Table 1.

Table 1. Coronavirus Susceptiblity Index Composition

VerticalIndicatorSourceWeight
 Healthcare Quality and Resources Per capita total health expenditures at PPPWHO3,6
 Index of healthcare system access and qualityIHME4,0
 Physicians (per 1,000 people)World Bank4,0
 Nurses and midwives (per 1,000 people)World Bank3,9
 Hospital beds (per 1,000 people)World Bank25,0
Economic Interconnectedness (external and internal) Total trade to GDP ratioWorld Bank
 Air transport passengers carried per capitaWorld Bank
 Global Connectedness IndexDHL
 Manufacturing and trade value added to GDP ratioUNSD9,7
Digital Infrastructure ICT development indexITU5,1
Demographic Susceptibility Urban population, % of totalWorld Bank4,7
 Population densityWorld Bank7,7
 Population 80+World Bank22,0
 Population 80+, % of totalWorld Bank
Trust in Government Government effectivenessWorld Bank5,9
 Rule of LawWorld Bank4,4
                                                                                                                                         Total100
– Indicator was eliminated from the composite index

Before the correlation tests, we normalized all indicators from Table 1 to the 0-100 range. Each indicator for a given country at a given time was calculated as the ratio of the difference between the raw indicator value and the minimum value divided by the difference between the maximum and the minimum. Indicators were inverted if growth in the indicators’ original values represent reduced risk.

We also imposed limitations on maximum values to avoid strong bias in the index. For example, US healthcare expenditures are nearly $8,000 per capita; however, values above $4,000 per capita do not appear to result in improvements in the quality of the US health care system. Therefore, we used $4,000 as the upper bound value in our formula.
Since data is not available for all time points for all countries from the original data sources, we also rolled forward values from the last reported date to cover gaps. Indicators’ values for the 2019-2020 period are estimates extrapolated from the 2010-2018 compound annual growth rate.

The results of our correlation tests are shown in Table 2. For our composite index, we included only those indicators with stable positive correlation with the COVID-19 measures for all country groupings.

Table 2. Correlation Test Results*

IndicatorCOVID-19 Confirmed Cases COVID-19 Deaths       COVID-19 Death RateRatio of confirmed cases to deaths
All countries (155)>100 cases  (72)>500 cases  (37)>1000 cases  (25)All countries (155)>1       (75)>10     (28)All countries (155)>0%    (75)>2%    (31)>4%    (24)
123456789101112
Air Transport (ration of passengers carried to population)0,09-0,01-0,21-0,320,01-0,05-0,23-0,108-0,294-0,295-0,296
Government Effectiveness-0,21-0,100,110,27-0,10-0,040,100,1350,4420,4590,482
Health Expenditure per Capita in PPP-0,26-0,150,030,23-0,19-0,16-0,080,0490,3320,2920,268
Hospital Beds (per 1,000 people)-0,14-0,08-0,040,07-0,050,000,060,0280,1900,2220,199
IHME Health Care Access and Quality Index-0,26-0,20-0,090,05-0,19-0,18-0,180,0170,4040,3320,274
Manufacturing and Trade to GDP Ratio0,100,120,190,220,090,090,14-0,038-0,156-0,199-0,146
Nurses and Midwives (per 1,000 people)-0,12-0,010,180,38-0,040,030,160,1030,3130,2910,286
Physicians (per 1,000 people)-0,20-0,10-0,020,13-0,16-0,13-0,140,0570,3760,3160,291
Population Density0,110,110,120,070,100,110,10-0,034-0,149-0,352-0,369
Rule of Law-0,16-0,030,200,42-0,050,010,150,1130,3730,3270,323
Trade in Goods and Services to GDP Ratio-0,16-0,25-0,31-0,40-0,13-0,17-0,25-0,229-0,316-0,369-0,482
Urban Population0,170,100,05-0,300,130,120,08-0,027-0,378-0,429-0,409
Population 80+0,740,720,780,770,490,470,450,074-0,032-0,141-0,164
Population 80+ %0,280,180,10-0,010,240,220,16-0,005-0,280-0,291-0,265
ICT Development Index Value-0,23-0,16-0,020,27-0,16-0,13-0,030,0700,4520,3850,392
Global0,190,07-0,09-0,290,130,07-0,01-0,119-0,491-0,517  -0,533
*Pearson correlation coefficients based on data from Johns Hopkins University as of March 24, 2020.

We assigned equal weights (⅓) to each of the three COVID-19 pandemic measures. Weights of individual indicators were calculated as a ratio of averages of correlation coefficients to the sum of averages of correlation coefficients within each COVID-19 measure multiplied by ⅓.

To reduce the bias related to high correlation between the population 80+ and both the number of cases and deaths, in the final stage, we redistributed half the weight for the population 80+ indicator in favor of hospital beds availability.

More sophisticated econometric approaches can be applied after we accumulate more statistical data on COVID-19 pandemic. While preparing this report, significant changes in the number of reported cases and deaths worldwide were still being reported daily, reducing the quality of econometric estimates.

Not all factors that influence the spread of COVID-19 are fully known or estimated yet. For example, it is known that temperatures above 40C destroy the virus. But, it is still not clear how weather impacts virus transmission. Local media outlets may be eager to celebrate the coming high temperatures in certain latitudes around the world, but scientists cannot yet support these hopeful headlines.5,6

Key Findings and Results

Under-testing in rural areas.

COVID-19 measures, such as the number of cases and deaths, have positive correlations with the same indicators except one: urbanization rate. The urbanization rate has a positive correlation with the number of deaths but, interestingly, not the number of cases. We interpret this as a lack of COVID-19 testing in rural areas.

Strong healthcare systems are not panaceas.

The COVID-19 death rate has positive correlations with health care system quality and resources indicators, institutional factors, and the level of ICT development. We’re pointing at you, developed countries. And yet the number of cases and deaths depend on demographic factors and economic interconnectedness. This finding could imply that the high quality of a health care system, availability of healthcare resources, and high institutional standards can’t prevent the spread of the pandemic.

Decisive government action and public response matter.

Our Coronavirus Susceptibility Index shows how deep the virus could hit a country’s population holding all other factors equal. A comparison of the index risk scores with the actual observed number of COVID-19 cases and deaths could serve as an indicator of the quality of government response to the pandemic.

Implementation of strict quarantine measures at the very beginning of an outbreak (the China scenario) can thwart the spread of COVID-19 even when all factors point to a critical risk level. At the same time, absence of a robust government response in the first two months of the outbreak leads to catastrophic social and potentially economic consequences, which are even now playing out in Italy and the United States, countries with upper-medium risk scores close to 50, much lower that China’s score  of 73 (critical risk of pandemic).

We would do a disservice to you all if we did not conclude by circling back to where we began this article: No one is equipped yet to confidently predict the full scale and impact of COVID-19, and all estimates and forecasts should be taken with caution. The situation is changing very fast, devaluing the robustness of statistical approaches that underpin forecasts.

Stay well.

The Knoema Team

References

  1. Keyes, Nick, and Sanchez-Bender, Marcela. “East Asia and Pacific: Countries Must Act Now to Mitigate Economic Shock of COVID-19.” Press release. The World Bank, 30 March 2020. Link
  2. Olga B. Jonas “Pandemic Risk”. A background paper to the World Development Report 2014. The World Bank, October 2013. Link
  3. Peter Sands, Anas El Turabi, Philip A Saynisch, Dr Prof Victor J Dzau “Assessment of Economic Vulnerability to Infectious Disease Crises” The Lancet, May 2016. Link
  4. Warwick J. Mckibbin “Global Macroeconomic Consequences of Pandemic Influenza”. Lowy Institute, February 2006. Link
  5. Bukhari Qasim, Jameel Yusuf “Will Coronavirus Pandemic Diminish by Summer?” Massachusetts Institute of Technology (MIT). March 17, 2020. Link
  6. Yeomans, David, and Henrikson, Eric, “Slowing a pandemic: How summer temperatures could impact coronavirus.” Weather blog. kxan. March 12, 2020. Link

Global Recession Is Coming. But Not In 2020.

Introduction COVID-19, the topic on the minds of billions globally. What could even be thought of as a short-term shock a little over a month ago is now being reframed as an economic readjustment of a scale no one can confidently predict, though many still try. Just this week the World Bank published an estimate …

How Leading Indices Become Lagging

Introduction COVID-19, the topic on the minds of billions globally. What could even be thought of as a short-term shock a little over a month ago is now being reframed as an economic readjustment of a scale no one can confidently predict, though many still try. Just this week the World Bank published an estimate …