Abstract
BACKGROUND: Antimicrobial resistance has emerged as one of the foremost public health troubles of the 21st century. This has ended in a public health disaster of the global situation, which threatens the exercise of present-day remedy. There is an urgent requirement for a cost-effective strategy to reduce antimicrobial resistance. Infectious disease control researchers most often analyze and predict antimicrobial resistance rate data that includes zeros or ones. Commonly used time-series analysis such as autoregressive moving average model is inappropriate for such data and may arrive at biased results. OBJECTIVE: This study aims to propose a time-series model for continuous rates or proportions when the interval of series includes zeros or ones and compares the model with existing models. DATA: The Escherichia coli, isolated from blood cultures showing variable susceptibility results to different antimicrobial agents, has been obtained from a clinical microbiology laboratory of a tertiary care hospital, Udupi district, Karnataka, during the years between 2011 and 2019. METHODOLOGY: We proposed a Degenerate Beta Autoregressive model which is a mixture of continuous and discrete distributions with probability mass at zero or one. The proposed model includes autoregressive terms along with explanatory variables. The estimation is done using maximum likelihood with a non-linear optimization algorithm. An R shiny app has been provided for the same. RESULTS: The proposed Degenerate Beta Autoregressive model performed well compared to the existing autoregressive moving average models. The forecasted antimicrobial resistance rate has been obtained for the next 6 months. CONCLUSION: The findings of this article could be beneficial to the infectious disease researchers to use an appropriate time-series model to forecast the resistance rate for the future and to have better or advance public health policies to control the rise in resistance rate.