JMIR Publications

ABSTRACT

Background:

Here we analyze a vast cohort comprising over 25 million COVID-19 patients collected by the Mexican Government between 2020 and 2023. The dataset contains valuable information on attributes and comorbidities, enabling us to investigate clinically relevant associations.

Objective:

Our objective is to unravel the intricate network of relationships between variables within the entire cohort and specific patient subsets, with a particular focus on associations involving fatalities.

Methods:

We employ the Odds Ratio (OR), estimated from Fisher’s test on 2×2 contingency tables, as a measure of association between variable pairs. We compute a total of 3,899 such measures by examining all possible variable pairs within 25 patient groups. The results are ordered and presented as networks, where variables are depicted as nodes (vertices) and associations at a specific OR threshold are represented as links (edges). The recoded data, along with the results and data mining functions, are publicly accessible.

Results:

Our findings demonstrate that hospitalization, gender, and age significantly influence disease outcomes, as do comorbidities such as pneumonia, chronic renal problems, diabetes, and hypertension. Interestingly, we observe that the associations of comorbidities are diminished in pregnant women, suggesting a pro- tective effect of pregnancy against the detrimental impact of these comorbidities on COVID-19 patients.

Conclusions:

Our analysis of variables in Mexican COVID-19 patients reveals a complex network of associations. By visualizing these associations as networks, we provide a clear and accessible representation that enhances our understanding of the factors contributing to fatalities in this population.