Abstract-Yilun Sun - Department of Biostatistics, St. Jude Children’s Research Hospital
Title: %HOSP_Sim: A SAS macro for projecting hospital bed census from daily new cases of COVID-19
Title: %HOSP_Sim: A SAS macro for projecting hospital bed census from daily new cases of COVID-19
Abstract:
Abstract:
It is critical to plan for hospital resources during the COVID-19 pandemic so that local health partners can take coordinated action to ensure the availability of sufficient bed capacity to accommodate critically ill patients in their communities. To help address this need, we developed a SAS macro %HOSP_Sim to project the short-term COVID-19 hospital bed census (occupancies) based on daily new cases using a microsimulation approach. The simulation utilizes the probability of hospitalization and the distribution of hospital length of stay. The %HOSP_Sim allows users to customize the number of simulation experiments, the length of the monitoring period, and time from diagnosis to hospitalization. Users can choose parameters that best describe regional observations, and those parameters can be altered to fulfill various assumptions. In addition, %HOSP_Sim could be used to conduct sensitivity analyses to check the robustness of findings to varying parameters and scenarios. It can be extended to simulation experiments using subgroup information such as population characteristics. We illustrate the use of %HOSP_Sim by projecting the short-term hospital bed census as a function of daily new cases of COVID-19 in Shelby County, Tennessee. The extension to a ShinyR app is ongoing.
It is critical to plan for hospital resources during the COVID-19 pandemic so that local health partners can take coordinated action to ensure the availability of sufficient bed capacity to accommodate critically ill patients in their communities. To help address this need, we developed a SAS macro %HOSP_Sim to project the short-term COVID-19 hospital bed census (occupancies) based on daily new cases using a microsimulation approach. The simulation utilizes the probability of hospitalization and the distribution of hospital length of stay. The %HOSP_Sim allows users to customize the number of simulation experiments, the length of the monitoring period, and time from diagnosis to hospitalization. Users can choose parameters that best describe regional observations, and those parameters can be altered to fulfill various assumptions. In addition, %HOSP_Sim could be used to conduct sensitivity analyses to check the robustness of findings to varying parameters and scenarios. It can be extended to simulation experiments using subgroup information such as population characteristics. We illustrate the use of %HOSP_Sim by projecting the short-term hospital bed census as a function of daily new cases of COVID-19 in Shelby County, Tennessee. The extension to a ShinyR app is ongoing.