Consistent Comparison of Symptom-based Methods for COVID-19 Infection Detection

Abstract

Multiple COVID-19 diagnosis methods based on information collected from patients have been proposed during the global pandemic crisis, with the aim of providing medical staff with quick diagnosis tools to efficiently plan and manage the limited healthcare resources. In general, these methods have been developed to detect COVID-19 positive cases from a particular combination of reported symptoms, and have been evaluated using datasets extracted from different studies with different characteristics. On the other hand, the University of Maryland, in partnership with Facebook, launched the Global COVID-19 Trends and Impact Survey (UMD-CTIS), the largest health surveillance tool to date that has collected information from 114 countries/territories since April 2020. This survey captured various individual features including gender, age groups, self-reported symptoms, isolation measures, and mental health status, among others. In this paper, we compare the performance of different proposed COVID-19 diagnosis methods using the information collected by UMD-CTIS, for the years 2020 and 2021, in five countries: Brazil, Canada, Germany, Japan, and South Africa. The evaluation of these methods with homogeneous data across countries and years provides a solid and consistent comparison among them.

Social Media Sentiment about COVID-19 Vaccination Predicts Vaccine Acceptance among Peruvian Social Media Users the Next Day

“…The UMD Global COVID-19 Trends and Impact Survey is a daily tracking survey of Facebook users conducted in 115 countries, including Peru [95][96][97][98]. The survey is conducted through Facebook recruitment that is weighted to represent the characteristics of each national population, with non-response and missing data treated as Missing Completely at Random (MACR) and computed with non-response weights with an inverse propensity score weighting (IPSW) approach (for details on survey methodology, please refer to [91,92,99]).…”

Performance and explainability of feature selection-boosted tree-based classifiers for COVID-19 detection

“…Several methods have been developed for the detection of COVID-19-active cases based on individual features extracted from survey data. These methods can be categorized into three classes: prediction rules, logistic regression methods, and machine learning models [10] . Prediction rules identify active cases based on a specific set of symptoms.…”