Epidemiology Network Analysis

Abstract

<p><i>Research questions</i> The primary objective of this doctoral dissertation is an explorative investigation into the social network dynamics within eight high schools, located in Tromsø and Balsfjord (North Norway), and the extent to which these dynamics contribute to the overall health and well-being of the students, such as in the context of infectious disease spread and the transmission of negative or positive health effects, and also in comparison with non-social host factors such as sports or recreational drug frequencies. Secondarily, we aim to develop new analytical methods and provide a framework for enabling agnostic evaluation of social networks in epidemiological studies and faster iterations of developing scripts for general statistical research. <p><i>Methodology</i> Using the Fit Futures gathered data on friendship, we used simulations, homophily, &Chi;² tables, logistic regression, and random forests as the main methods to analyze social influence in our topics of interest. We applied classical database normalization and data cleaning to the original data and developed scripts for automatic analysis in R and Python exporting results directly in plain text, Latex, and HTML. <p><i>Results</i> We found that the social network influences significantly the spread of <i>Staphylococcus aureus</i>. Students close to the network tend to have similar inflammatory biomarkers, 25OHD, and BMI levels. Some high schools tend to consume similar levels of over-the-counter medicines and tend to share the same brand of prescribed medicines. There is also a bias on recreational drug usage by high schools. <p><i>Conclusions</i> Social influence is shown to be significant in every analysis. These findings emphasize the importance of considering social network dynamics in understanding and addressing health and well-being issues among students. Further research and interventions targeting social network influences can contribute to developing more effective health strategies. <p><i>Originality</i> Use of non-parametric simulation and machine learning methods to estimate social influence. We are measuring social influence on 25OHD, in an inflammatory proteomic assay. <p><i>Significance</i> Social influence, whether from virtual friends or physical ones, is a growing area of interest in many fields. In Epidemiology in particular we saw a boost in popularity after the Sars-Cov-2 pandemic.