Pemodelan Regresi Logistik untuk Diagnosis Dini Infeksi Covid-19 di Indonesia

Abstract

Controlling the spread of the Covid-19 virus in Indonesia, the government continues to strive for a comprehensive 3T (Testing, Training, and Tracing) implementation. Massive testing is often constrained by several things, including cost and affordability of access. This study aims to create a model for early diagnosis of Covid-19 infection cases through several characteristic symptoms and experiences of close contact with positive patients. By using a binary logistic regression model, it was found that symptoms of anosmia, feverish symptoms, and close contact experience were significant in influencing Covid-19 infection. From the odds ratio value, it is known that anosmia is the most influential variable. Someone who has anosmia tends to be infected by 31 times compared to those who do not. Validation of the strength of the model in classifying is done by making predictions on the resulting model is good, because the measurement of each criterion of the strength of the model consists of accuracy, sensitivity, and specificity of the model both on the data testing each produces a value of 0.8 (close to 1). The area under the Receiver Operating Characteristic (ROC) curve for testing data is 0.8462, which means that the model already has good criteria for classifying.

Keywords

• covid-19
• early diagnose
• logistic regression
• receiver operating characteristic