Rethinking of value of early-stage infectious disease modelling to public health: a real-world data validation of SIR models

Research Square (Research Square)(2022)

引用 0|浏览0
暂无评分
摘要
Abstract Objectives Performance of SIR model in the early stage of a novel epidemic may be hindered by data availability. Additionally, the traditional SIR model may oversimplify the disease progress, and knowledge about the virus and transmission is limited early in the epidemic, resulting in a greater uncertainty of such modelling. We aimed to investigate the impact of model inputs on the early-stage SIR projection using COVID-19 as an illustration to evaluate the application of early infection models. Design, Setting and Methods We constructed a modified SIR model using discrete-time Markov chain to simulate daily epidemic dynamics and estimate the number of beds needed in Wuhan in the early stage of COVID-19 epidemic. We compared eight scenarios of SIR projection to the Real-world data (RWD) and used root mean square error (RMSE) to assess model performance. Participants There are no participants involved in this study. Results According to the National Health Commission, the number of beds occupied in isolation wards and ICUs due to COVID-19 in Wuhan peaked at 37,746. In our model, as the epidemic developed, we observed an increasing daily new case rate, and decreasing daily removal rate and ICU rate. This change in rates contributed to the growth in the needs of bed in both isolation wards and ICUs. Assuming a 50% diagnosis rate and 70% public health efficacy, the model based on parameter estimated using data from the day reaching 3,200 to the day reaching 6,400 cases returned a lowest RMSE. This model predicted 22,613 beds needed in isolation ward and ICU as on the day of RWD peak. Very early SIR model predictions based on early cumulative case data initially underestimated the number of beds needed, but the RMSEs tended to decline as more updated data were used. Conclusions Early-stage SIR model, although simple but convenient and relatively accurate, is a useful tool to provide decisive information for public health system and predict the trend of an epidemic of novel infectious disease in the very early stage. However, model inputs should be frequently revisited considering the fluctuation of early-stage data and the impacts of policy-related factors should be reviewed cautiously.
更多
查看译文
关键词
infectious disease modelling,public health,models,early-stage,real-world
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要