Abstract:
Novel Corona viruses are a viral family that causes Severe Acute Metabolic Syndrome. The Virus
was strictly increasing throughout in the World. From the total cases(91.5%) were cured in the
world.In Africa(60%) were cured and In Ethiopia (98.61%) were cured from march 2020 to March
2021. Many scholars conducts prognostic factors of Covid-19 by using coxph,non parametric,
and logistic regration model.But loglogistic regration does not account censoring observation and
Coxph model and non parametric models were used in the independent and identically distributed
covariates.For those model hetrogeneity does not considered. The objective of this thesis was de velop various parametric frailty model to detect random effect on time–to-cure of covid-19 in the
two treatment care center. Data were collected in Jimma university and Shenen Gibe covid-19
Center. Appropriate model that describes the Covid-19 data were Gamma and Inverse Gaussian
frailty with exponential, log-logistic, log-normal, and Weibull baseline function were compared.
Based on Akaike information criterion criteria all models were compared.Data were analyzed by
using R version 4.0.5 software.298 covid-19 patients 246(82.65%)were cured with the median cur ing time of 19 days.The log-logstic model with Gamma frailty distribution has the smallest Akaike
information criterion(1609.625) value compared with the others. Clustering effect is significant
on the modeling time to cure from covid-19 with in test of unobserved hetr0geneity in all models.
From this finding, age group, severity ,co morbidity , diabetics, lung-cancer, and oxygen are prog nostic (significant) factors for time to cure of covid-19. From this thesis almost of patient were
cured in two treatment care center around Jimma zone.log-logistic Gamma frailty model was best
fit of Covid-19 dataset.There is heterogeneity between the two treatment center for time to cure
of covid-19. Researcher recommend that there is some limitation of parametric frailty model.For
further study who have interest to compare parametric model the simulation is appropriate.
