Predictive and Etiologic Analysis of Typhoid Fever Using Multivariate Logit Regression and GPT Data Analyst

Typhoid fever is one of the dangerous diseases and many sufferers. This disease attacks humans regardless of age and in principle can be cured. The chance of a patient's recovery can be known, among other things, when someone has historical information or the patient's history during the time of treatment. Historical data from various typhoid patients can be used to predict the recovery of other typhoid patients, if they are in similar conditions. This study aims to apply multivariate logistic regression from the concept of prediction and etiology to examine the recovery factors of typhoid patients. This study's etiological notion is restricted to using a single exposure variable. The data analysis uses quantitative research. The results of the predictive concept show that the regression equation model is y=0.465 -1.174Age(2)-0.646Age(3)-0.888Age(4)+0.211Hemoglobin(1)+0.317Platelet(1)+0.308Calcium(1)-0.330Current_Medication(1) +0.500Current_Medication(2). The chance of a typhoid patient recovering is a maximum of 75%, which can occur when the patient's platelet and calcium conditions are normal. Meanwhile, the lowest chance of patient recovery is 23%, which can occur when the patient is 31-40 years old and the treatment applied is ceftriaxone. From an etiologic concept, two best models were found, namely a model where age is the exposure variable, current medication is the confounder, and treatment_outcome is the dependent variable, and the second model where age is the exposure variable, current medication is the confounder, and treatment_outcome is the dependent variable. From the etiological concept, it can be seen that the variables that have the most influence on patient recovery are age and the treatment used. In addition, the use of GPT Data Analyst was concluded to be unable to directly analyze logistic regression data for typhus cases, but it can be used to help simplify data analysis for researchers by using logistic regression coding.