Exploitation of Deaths Registry in Mexico to Estimate the Total Deaths by Influenza Virus: A Preparation to Estimate the Advancement of COVID-19

Abstract

Following the AH1N1 influenza virus of 2009, it was suspected that many deaths were being incorrectly registered as caused by unclassified pneumonia in Mexico. In light of the current SARS-CoV-2 (or COVID-19) pandemic, it was assumed that a similar phenomenon was occurring. To verify this hypothesis, a machine learning algorithm that can estimate the extent of false negative AH1N1 influenza virus registration in Mexico was developed. The INEGI database of deaths in Mexico in 2005 through 2008, and World Health Organization International Classification of Diseases, the deaths by influenza and deaths by unclassified pneumonia were utilized to train the algorithm in order to differentiate the expected and observed influenza deaths in 2009. By predicting the pattern of unclassified pneumonia deaths for the year 2009, it was found that the difference between the expected and observed deaths had a strong correlation with the amount of deaths of influenza virus. This reveals that the deaths recorded as influenza virus in 2009 are a statistical representation of many deaths registered as unclassified pneumonia, but attributable to the same virus. With this, it was possible to estimate the precise ratio of this correlation in 2009. Without the COVID-19 and unclassified pneumonia data for the years 2019 and 2020 available, it is not yet possible to apply the findings of this work to the current global pandemic. However, a generalization method to do so is proposed. This work was made on Python, and the code is available on GitHub (https://github.com/EByrdS/influenza_deaths_mx).