2. Collect Empirical Data for Forecasting

Using NHSN Hospitalization Data

Set parameters

state_name <- "Maryland"
geo_ids <- "md"
forecast_disease <- "influenza"
forecast_date = "2024-12-01" #"2025-10-12" # Sunday

Load packages

knitr::opts_chunk$set(echo = TRUE)

library(AMPHForecastSuite)
library(tidyverse)
library(jsonlite)
library(epidatr)
library(epiprocess)

Save and check API key

1. Go to https://api.delphi.cmu.edu/epidata/admin/registration_form and register for a psuedo-anonymous account.

2. Run the command below to save your API key in your .Renviron file. As it says in the prompt, add a line DELPHI_EPIDATA_KEY=yourkeyhere to your .Renviron file. Save and close the file, then restart R or RStudio. After restarting, run the command epidatr::get_api_key() to check that your API key is saved.

If the API key does not come, you can use this one until it comes: 04e7369e1541a

# Follow instructions about opening `.Renviron` file
epidatr::save_api_key()

# Check that the API key is saved
epidatr::get_api_key()

Pull NHSN hospitalization data

target_data <- get_nhsn_data(
  disease = forecast_disease, #"influenza"  or "rsv" or "covid"
  geo_values = geo_ids,
  forecast_date = forecast_date,
  save_data = TRUE
)

## Important: forecast_date is more than 1 week ago. Pulling data issued prior to forecast_date.

## Pulling data issued on or before 2024-12-01

## Warning: No API key found. You will be limited to non-complex queries and encounter rate
## limits if you proceed.
## ℹ See `?save_api_key()` for details on obtaining and setting API keys.
## This warning is displayed once every 8 hours.

Or pull directly with epidatr

Note: To pull data directly that pertains to a previous forecast date, you need to filter by issue date. The code below is not set up to do that, so it will pull the most recent data. To pull data as of a previous forecast date, use the get_nhsn_data() function above.

# Map disease names to NHSN signal names
# Based on epidatr NHSN signals for respiratory diseases
signal_map <- list(
  "influenza" = "confirmed_admissions_flu_ew",
  "covid" = "confirmed_admissions_covid_ew",
  "rsv" = "confirmed_admissions_rsv_ew"
)
signal <- signal_map[[forecast_disease]]

# Call epidatr to get the data
target_data <- epidatr::pub_covidcast(
  source = "nhsn",
  signals = signal,
  geo_type = "state",
  time_type = "week",
  geo_values = tolower(geo_ids)
)

target_data <- target_data %>%
  dplyr::filter(time_value >= lubridate::as_date("2020-09-01"),
                time_value < lubridate::as_date(forecast_date)) %>%
  dplyr::mutate(disease = forecast_disease, signal = signal) %>%
  dplyr::select(geo_value, source, disease, signal, issue_date = issue, time_value, value)

# convert to match hubverse

data("loc_data")

target_data <- target_data %>%
  dplyr::mutate(abbreviation = toupper(geo_value)) %>%
  dplyr::mutate(target_end_date = time_value + 6) %>%
  dplyr::mutate(target = paste0("wk inc ", disease, " hosp")) %>%
  dplyr::left_join(loc_data %>% dplyr::select(abbreviation, location, location_name),
                   by = c("abbreviation" = "abbreviation")) %>%
  dplyr::arrange(target_end_date, location_name) %>%
  dplyr::select(location, abbreviation, location_name, target, source, disease, signal,
                target_end_date, issue_date, observation = value)

Check the data

ggplot(target_data, aes(x = target_end_date, y = observation)) +
  geom_line() +
  facet_wrap(~toupper(location_name), scales = "free_y") +
  labs(title = paste("NHSN Hospital Admissions"),
       x = NULL,
       y = "Weekly hospital admissions (n)",
       alt = "Line plot of weekly hospital admissions over time") +
  theme_bw()

Save it

write_csv(target_data, file = file.path("target-data", paste0("target-hospital-admissions-", forecast_date, ".csv")))