Implementation of an Adaptable COVID-19 Utilization and Resource Visualization Engine (CURVE) to Depict In-Hospital Resource Forecasts Over Time

Overview

1. Instructions for running the CURVE app in RStudio:

To run CURVE, we suggest using RStudio (https://www.rstudio.com/). After installing RStudio, double click the “COVID19-FORCASTING Sandbox.Rproj” file to open the app within the RStudio IDE. Once RStudio is open, click "app.R" file under the Files tab to open the file, and hit Run App on the upper-right portion of the source window. The screenshots below should help guide this process.

The R Shiny app will be displayed as the following figure.

Within CURVE, users can view the COVID-19 hospital, ICU, and ventilator censuses that are forecasted by the ARIMA model. The metaparameters for the ARIMA model can be manually adjusted by entering p, d, and q, or switching the toggle for Optimized ARIMA Parameters followed by clicking the Submit Parameters button. Note the later option uses the auto.arima() function provided by the forecast package (Hyndman, Athanasopoulos, Bergmeir, Caceres, Chhay, O'Hara-Wild, Petropoulos, Razbash, Wang, Yasmeen 2021; Hyndman and Khandakar 2008). Users can also change the date or modify the number of days ahead in the forecast. All the charts will be adjusted accordingly.

2. Customizing CURVE:

• Changing the logo:

  Users can modify the logo by adding their own logo into the WWW folder. Currently, we have named our logo figure "AH_logo.png". Users can then change the name of the figure in the "ui.R" file (see the figure below).

• Changing COVID-19 bed counts:

  All the bed counts are stored in the "bed_stat.csv" file, including the COVID-19 hospital census (named all_bed), ICU bed census (named icu), mechanical ventilator usage (named vent), and virtual hospital acute beds (named virtual_acute). If users don’t have virtual hospital bed counts, then they can put 0s in the virtual_acute column.

• Changing bed capacity data:

  The bed capacity data are stored in the "parameters.csv" file. The 1st column (named item) stores the categories of the beds, including currently occupied non-COVID-19 beds (named all), additional unoccupied surge beds (named surge), unoccupied virtual hospital acute beds (named virtual_acute), unoccupied acute care medical facility beds (named morecrot_bed), and projected non-COVID-19 beds based on known hospital metrics (named pred_non_covid).

  The 2nd column contains total numbers of beds. For example, on 3/11/2020, we had 100 surge beds, followed by surge bed counts of 583, 1,744, and 2,000 on 4/21, 4/23, and 5/1/2020, respectively. Note the 3rd column, entitled start_date, contains the aforementioned dates.

  Users can also change the beds labels and colors (4th and 6th columns). Note that a bed category should have a consistent label and color. For example, all surge beds have the same label (Surge Beds) and color (rgba(183, 236, 234, 0.5)).

  In the 5th column, a stackgroup value of one defines a group of bed types that is used by the area chart in the Bed Capacity section of CURVE. Projected non-COVID-19 beds were pooled with the COVID-19 hospital census and the forecast and assigned a stackgroup value of two. This group was displayed using a bar chart that was overlayed on the area chart. Thus, the user can quickly ascertain whether the total number of occupied beds will exceed the health system’s bed capacity. Users can add or remove a different category of bed to a group of bed types.

To reference CURVE, please cite the paper:

@article{chou2020covid,
  title={COVID-19 Utilization and Resource Visualization Engine (CURVE) to Forecast In-Hospital Resources},
  author={Chou, Shih-Hsiung and Kearns, James T and Turk, Philip and Kowalkowski, Marc and Roberge, Jason and Priem, Jennifer S and Taylor, Yhenneko J and Burns, Ryan and Palmer, Pooja and McWilliams, Andrew D},
  journal={medRxiv},
  year={2020},
  publisher={Cold Spring Harbor Laboratory Press}
}

References

• Hyndman R, Athanasopoulos G, Bergmeir C, Caceres G, Chhay L, O'Hara-Wild M, Petropoulos F, Razbash S, Wang E, Yasmeen F (2021). forecast: Forecasting functions for time series and linear models. R package version 8.14, <URL: https://pkg.robjhyndman.com/forecast/>.
• Hyndman RJ, Khandakar Y (2008). “Automatic time series forecasting: the forecast package for R.” Journal of Statistical Software, 26(3), 1-22. <URL: https://www.jstatsoft.org/article/view/v027i03>.