Background: Covid-19 has raked a devastating trail across continents with the highest mortality to infected ratio ever for a pandemic. The key containment strategy identified has been social seclusion. Analyzing extensive live statistics with a fully predictive global Covid-19 model, we quantify implications of lockdown and strategic quarantine on Covid-containment. Methods: We analyze Covid-19 infection kinetics using a 6 dimensional (healthy, pre-existing, infected, recovered, immuned, dead) continuum model using a probabilistic Machine Learned (ML) kernel for parameter prediction using data from the Johns Hopkins repository. The ML platform itself uses a double filtration process, first using statistics for infected only, followed by statistics for the infected and dead combined. We use this multi-scale, multivariate ML architecture to categorize 19 countries into 4 different infection classes according to mortality-to-infected ratio. Finding: The model almost unerringly predicts the number of infected and dead (within 1 s.d.) for all infection classes, consistently up to 30 days beyond the last ML training date (10 May 2020). The Reproductive Number R0 estimated from the model makes catastrophic predictions (1.5 < R0 < 2.7) for countries resorting to categorical de-prioritization of lockdown (such as UK, US, Sweden and India). Countries in other infection classes show smaller R0 (1.06 < R0 < 1.5) with faster recovery time, reflecting the impact of variable lockdown measures, with numbers matching almost perfectly with real data. The model also predicts the date of secondary relapse. Interpretation: The model provides robust future estimation for the number of dead for all 19 infected countries that we studied, unfailingly up to 4 following weeks. Early versus later lockdown clearly has major impact on mortality statistics as also on the timelining of secondary relapse, both accurately predicted by the model. The critical importance of correct lockdown span and timing is thus accentuated and quantified. Funding Statement: All authors have been resourced through their affiliating organizations. Declaration of Interests: There is no known conflict of interest. Ethics Approval Statement: This is an epidemiological study with data collected from open access sources. Ethical approval is not required.
Read full abstract