One of the most critical steps in controlling a pandemic disease, which has a significant mortality rate, is testing for the disease. Though this may appear relevant and obvious, its ability to forecast mortality in the current COVID-19 pandemic has not been tested.
A new study published on the preprint server medRxiv* in May 2020 shows that testing can identify the probable mortality in the current outbreak, via a rapid, policy-oriented technique of analysis.
Study: Early and massive testing saves lives: COVID-19 related infections and deaths in the United States during March of 2020. Image Credit: stock / Shutterstock
*Important notice: medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.
Epidemic-Relevant Metrics
There are three types of measurable parameters, namely, case counts, disease counts, the prevalence of disease over location or time, and the population density of infected locations. The real prevalence of a disease that causes a large number of asymptomatic infections can be assessed only if everyone is tested using a test with high sensitivity. As a result, the term apparent prevalence is used, denoting the ratio of cases that test positive to the total number of tested individuals. This ratio expressed as a number of positive cases per million residents can help to compare the disease burden between different geographical units.
The apparent prevalence method describes the ratio of test-positive cases to all tested individuals. If expressed per million residents, the apparent prevalence can compare different geographical units, e.g., each and all states of the US.
However, to conduct comprehensive studies that investigate numerous states, a protracted research program is required. To rapidly provide policy-makers with usable information, here, a quasi-real-time assessment was designed, which captures both nationwide and state-specific dimensions. Analyzing the epidemic data reported in all 50 states of the USA, during March of 2020 (the month when testing started), we investigated whether testing-related variables –including extensive and early testing− predict mortality.
What Parameters Were Used?
The variables assessed include:
- Number of diagnostic tests during the first week after testing
- The above number as a proportion of the total number of tests
- The total number of positive tests until March 31, 2020
- The apparent prevalence rate, or the number of cases per million residents
The data was then analyzed using a machine learning strategy called KRLS (kernel regularized least squares) regression. Pattern recognition was the basis of the analysis. Any distinct pattern led to the selection of a threshold that matched the upper limit of a data compartment that was linearly distributed to allow three groups of data to be identified from the meeting of two orthogonal lines.
What Did the Study Show?
The researchers found that 93.5% of the variance in number of deaths and 86.7% of the variance in deaths/million cases was due to these six predictors. The two that had statistical significance were the total number of confirmed cases and the apparent prevalence rate, both of which were comparable when it came to predicting the number of deaths. However, when the deaths/million are to be predicted, the apparent prevalence is better by a factor of 3.5.
When the first variable is analyzed, it differentiates the states into three groups based on predicting the deaths per million.
The Advantage of Multiple Metrics
The researchers also found that measures including two or more of the variables which interact with each other are more helpful as reporting measures. This is because single metrics do not convey the changing situation or the geographical factors that influence the outcome. Instead, the use of composite measurements allows for the interaction between multiple dimensions that make geography-specific interventions to be executed.
The current study used KRLS regression analysis only to arrive at an instantaneous map of the mortality. However, this is a powerful and versatile tool that could be used over shorter time intervals to provide a picture of changing epidemic parameters.
The results of the current analysis show how important it is to use interacting measurements to analyze an epidemic. One instance is the use of several metrics, such as the number of tests performed in week I/million citizens/population density, which clearly shows how large-scale testing early in an epidemic can prevent deaths.
Improving the Results of Testing
The result is affected by other factors like the availability of diagnostic equipment and personnel, hospital beds, and of intensive care units, and the interplay of demographic and geographic interactions at local and regional levels. Population-dense regions will have a more significant number of interindividual contacts and better connectivity via road, air, and waterways, which both tend to encourage the spread of an epidemic.
Pandemics can thus be analyzed as a cluster of processes at the local and regional level, which interact with each other, currently published data relating to COVID-19 is often poorly referenced concerning the location and lacks sufficient detail. The simple putting together of multiple geographical lines and points which characterize surface-based data leads to the exclusion of internal processes that hit only one city or neighborhood.
Only with such minutely detailed data can the location and time of intervention be accurately identified. The researchers conclude, “To optimize these approaches and reduce the COVID-19 related mortality, the collection, and reporting of high-resolution, geo-temporal data constructed as interactions are recommended.”
*Important notice: medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.
Research shows how public criticism during COVID-19 impacted teacher wellbeing