Use of IT tools to search for a correlation between weather factors and onset of pulmonary thromboembolism

TL;DR

This study investigates the potential correlation between weather factors, especially atmospheric pressure changes, and pulmonary embolism hospitalizations using IT tools and high-resolution weather data, revealing a strong negative correlation.

Contribution

The paper introduces a novel framework using Python and econometric tools to analyze high-resolution weather data in relation to PE hospitalizations, highlighting a significant correlation.

Findings

Strong negative correlation between atmospheric pressure and hospitalizations

Increase in hospitalizations following periods of high pressure changes

Framework and tools developed for high-resolution weather and health data analysis

Abstract

Pulmonary embolism (PE) and deep vein thrombosis (DVT) are gathered in venous thromboembolism (VTE) and represent the third cause of cardiovascular diseases. Recent studies suggest that meteorological parameters as atmospheric pressure, temperature, and humidity could affect PE incidence but, nowadays, the relationship between these two phenomena is debated and the evidence is not completely explained. The clinical experience of the Department of Emergency Medicine at AOUC Hospital suggests the possibility that a relationship effectively exists. We have collected data concerning the Emergency Medicine Unit admissions of PE patients to confirm our hypothesis. At the same time, atmospheric parameters are collected from the Lamma Consortium of Tuscany region. We have implemented new IT models and statistic tools by using semi-hourly records of weather time high resolution data to process the dataset. We have carried out tools from econometrics, like mobile means, and we have studied anomalies through the search for peaks and possible patterns. We have created a framework in Python to represent and study time series and to analyze data and plot graphs. The project has been uploaded on GitHub. Our analyses highlighted a strong correlation between the moving averages of atmospheric pressure and those of the hospitalizations number (R= -0.9468, p<0,001) although causality is still unknown. The existence of an increase in the number of hospitalizations in the days following short-to-medium periods of time characterized by a high number of half-hourly pressure changes is also detected. The spectrograms studies obtained by the Fourier transform requires to increase the dataset. The analyzed data (especially hospitalization data) were too few to carry out this kind of analyses.