Comparison of Data Mining Methods for Survival Analysis of High-dimensional Data for Children with COVID-۱۹

Background and aims: Data collected from pandemic diseases, such as COVID-۱۹, are oftenhigh-dimensional, censored, and heterogeneous, presenting a lack of robust statistical analysis.Ridge and Lasso regression are proposed methods that can overcome the difficulties of modelingthis complex data. This study aims to apply these two methods to identify influential variables inthe survival of children with COVID-۱۹.Method: In this follow-up study, from February ۲۰۲۰ to July ۲۰۲۱, ۱۸۵ children aged betweenone month and eighteen years old with confirmed COVID-۱۹ in Shahid Sadoughi Hospital inYazd followed until the end of day ۳۰ after discharge, except for the expired cases. All patientshad complete records of demographic, including comorbidities, clinical symptoms before andat admission time, laboratory factors, and treatment variables. Time to death after one month isdefined as survival time, and more than ۵۰ variables are considered in Ridge and Lasso regressionindependently. Analysis was done using R ۴.۱.۱ software.Results: Data analysis using Ridge regression in comparison with Lasso regression indicated thatLasso regression estimated the coefficients of influential variables more precisely consideringlambda ۰.۰۹.Conclusion: In the analysis of high-dimensional data, traditional statistical methods cannot beused. Lasso regression can solve this problem and estimate the coefficient of each variable robustly.

کلیدواژه ها:

COVID-۱۹ ، High-dimensional ، Ridge ، Lasso ، Survival