In rplzzz/CovMitigation: COVID-19 infection model for Virginia

Infection response to reopening

We will pull the model for the localities that have colleges that have reopened in or near them. For each such locality we will look for any obvious change in model parameters around the time of the relevant reopenings.

library(CovMitigation)
library(dplyr)
library(stringr)
library(ggplot2)

read_model_data <- function(dir) {
  filepat <- 'filter-fit-(.*)\\.rds'
  savefiles <- list.files(path=dir, pattern=filepat)
  localities <- str_match(savefiles, filepat)[,2]
  savefiles <- file.path(dir,savefiles)
  mods <- lapply(savefiles, readRDS)
  names(mods) <- localities
  mods
}

models <- read_model_data(here::here('analysis','filter-updates.2021-05-23'))
models <- models[names(models) %in% va_college_reopening$locality]

makeplot <- function(loc, var='beta') {
  reopendata <- filter(va_college_reopening, locality==loc)
  modelfit <- models[[loc]][['modelfit']]
  modelfit$date <- as.Date(modelfit$time, origin='2020-01-01')
  modelfit$value <- modelfit[[var]]
  dmax <- max(modelfit$value)

  ggplot(modelfit, aes(x=date,y=value)) + 
    geom_line(size=1.2) +
    geom_vline(data=reopendata, mapping=aes(xintercept=date, color=School), size=1.2) +
    ggtitle(loc) +
    ylab(var) +
    coord_cartesian(ylim=c(0,dmax)) +
    #coord_cartesian(ylim=c(0,1.5)) +
    theme_bw()
}

beta_plots <- lapply(names(models), makeplot)
for(plot in beta_plots) {
  print(plot)
}

import_plots <- lapply(names(models), makeplot, var='import_cases')
for(plot in import_plots) {
  print(plot)
}

enrichment_plots <-  lapply(names(models), makeplot, var='b0')
for(plot in enrichment_plots) {
  print(plot)
}

For both beta (transmissibility) and import_cases (daily case importation rate), The only locality that shows an obvious response is Harrisonburg City, where beta jumps from about 0.25 to over 3 in the second week following reopening. The import rate also roughly quadruples during that time.

This makes some sense because JMU was a COVID-19 hotspot in the weeks following reopening; however, Radford and Virginia Tech also had outbreaks; yet, the Radford City, Pulaski County, and Montgomery County don't show any effect in beta, and except for an uptick from 1.5 to 3 in Montgomery County, there's no effect in import_cases either. Therefore, our next question is, what was different in the Radford/VT area as compared to JMU?

Detailed comparison: JMU vs. Radford/VT

localities <- c('Lynchburgcity','Harrisonburgcity', 'MontgomeryCounty', 'Radfordcity', 'PulaskiCounty',
                'AlbemarleCounty', 'Charlottesvillecity')
models2 <- models[localities]

dgplots <- lapply(models2, filter_model_diagnositc_plots)
caseplots <- lapply(localities, function(loc) 
  {
    plt <- dgplots[[loc]][['cases']]
    reopen <- filter(va_college_reopening, locality == loc)
    plt + geom_vline(data=reopen, mapping=aes(xintercept=date, color=School), size=1.2)
  })

for(plt in caseplots) {
  print(plt)
}