Open AccessAssessing the likelihood of contracting COVID-19 disease based on a predictive tree model: A retrospective cohort study
Author(s) -
Francesc X Marín-Gomez,
Mireia Fàbregas-Escurriola,
Francesc López Seguí,
Eduardo Hermosilla Pérez,
Mència Benítez Camps,
Jacobo Mendioroz Peña,
Anna RuizComellas,
Josep Vidal-Alaball
Publication year - 2021
Publication title -
plos one
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.99
H-Index - 332
ISSN - 1932-6203
DOI - 10.1371/journal.pone.0247995
Subject(s) - medicine , decision tree , logistic regression , medical diagnosis , retrospective cohort study , contact tracing , disease , medical record , covid-19 , data mining , computer science , pathology , infectious disease (medical specialty)
Background Primary care is the major point of access in most health systems in developed countries and therefore for the detection of coronavirus disease 2019 (COVID-19) cases. The quality of its IT systems, together with access to the results of mass screening with Polymerase chain reaction (PCR) tests, makes it possible to analyse the impact of various concurrent factors on the likelihood of contracting the disease. Methods and findings Through data mining techniques with the sociodemographic and clinical variables recorded in patient’s medical histories, a decision tree-based logistic regression model has been proposed which analyses the significance of demographic and clinical variables in the probability of having a positive PCR in a sample of 7,314 individuals treated in the Primary Care service of the public health system of Catalonia. The statistical approach to decision tree modelling allows 66.2% of diagnoses of infection by COVID-19 to be classified with a sensitivity of 64.3% and a specificity of 62.5%, with prior contact with a positive case being the primary predictor variable. Conclusions The use of a classification tree model may be useful in screening for COVID-19 infection. Contact detection is the most reliable variable for detecting Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) cases. The model would support that, beyond a symptomatic diagnosis, the best way to detect cases would be to engage in contact tracing.