In uva-bi-sdad/dc.hifld.ems: Social Data Commons: Hospitals Access Scores

knitr::opts_chunk$set(echo = TRUE)

libraries

library(sf)
library(tidygeocoder)
library(dplyr)
library(community)
library(catchment)

clean df

#start clean
rm(list=ls())

working directory

setwd("~/VDH/Floating_Catchment_Areas/dmv/6. ems_stations_dmv_model/")

load data provider. Do not run

#provider <- read.csv("pediat.dmv.geo.csv", row.names = 1)
#from pgadmin
#a) conn
con <- RPostgreSQL::dbConnect(drv = RPostgreSQL::PostgreSQL(),
                               dbname = "sdad",
                               host = "postgis1",
                               port = 5432,
                               user = Sys.getenv(x = "DB_USR"),
                               password = Sys.getenv(x = "DB_PWD"))

#b) query
provider <- sf::st_read(
  con, 
  query= "
SELECT  *
FROM dc_health_behavior_diet.ncr_pl_hifld_2021_ems_stations  "
)

provider <- provider %>% select(unique_ID, facility_name, address, city, state, county, fips, lat, lon)

#c) Disconnect
RPostgreSQL::dbDisconnect(con)

#write
write.csv(provider, "ems_stations.csv" )

load data provider

provider <- read.csv("ems_stations.csv", row.names = 1)

fix data

## collapse by location
provider$doctors <- 1
provider$location <- paste0(provider$lat, ",", provider$lat)
counts <- tapply(provider$doctors, provider$location, sum)
locations <- which(!duplicated(provider$location))
provider <- provider[locations,]
provider$doctors <- counts[provider$location]

## assign IDs just to be explicit
provider$ID <- paste0("l", seq_len(nrow(provider)))

data combined

## data combined =
county_names <- list(
  dc = "District of Columbia",
  md = c("Charles", "Frederick", "Montgomery", "Prince George's"),
  va = c("Alexandria", "Arlington", "Fairfax", "Falls Church", "Loudoun", "Manassas", "Manassas Park", "Prince William")
)
data <- list()
shapes <- list()


for(state in c("dc", "md", "va")){
  # get county shapes to identify block groups
  counties <- download_census_shapes(".", state, "county", paste0(state, "_counties"))
  blockgroups <- download_census_shapes(".", state, "bg", paste0(state, "_blockgroups"))

  # save only selected counties
  counties <- counties[counties$NAME %in% county_names[[state]],]
  counties[counties$NAME == "Fairfax", "NAME"] <- c("Fairfax City", "Fairfax")
  shapes[[state]] <- list(
    counties = counties,
    blockgroups = blockgroups[substr(blockgroups$GEOID, 1, 5) %in% counties$GEOID,]
  )

  # population data
  data[[state]] <- download_census_population(".", state, 2019, counties = counties$GEOID)
}

data_combined <- do.call(rbind, lapply(names(data), function(state){
  d <- data[[state]]$estimates
  s <- shapes[[state]]$blockgroups
  rownames(s) <- s$GEOID
  total <- d$TOTAL.POPULATION_Total
  total[total == 0] <- 1
  data.frame(
    GEOID = d$GEOID,
    population = d$TOTAL.POPULATION_Total,
    percent_female = d$SEX.BY.AGE_Female_Female / total * 100,
    percent_white = d$RACE_Total_White.alone / total * 100,
    percent_over_49 = rowSums(d[, grep("[5-8][05]", colnames(d))]) / total * 100,
    female_over_14 = rowSums(d[, grep("Female_(?:[2-9]\\d|1[58])", colnames(d))]),
    pediatric_pop = rowSums(d[, c(4:7,28:31) ]),

    st_coordinates(st_centroid(st_geometry(s[as.character(d$GEOID),])))
  )
}))
data_combined <- data_combined[!is.na(data_combined$Y),]

travel time

library(osrm)
options(osrm.server = Sys.getenv("OSRM_SERVER"), osrm.profile = "car")
if(!file.exists("traveltimes_exercise.csv")){
  traveltimes <- osrmTable(
    src = data_combined[, c("GEOID", "X", "Y")],  #population-demand
    dst = provider[, c("ID", "lon", "lat")]     #providers supply
  )$duration
  write.csv(
    cbind(GEOID = rownames(traveltimes), as.data.frame(traveltimes)),
    "traveltimes_exercise.csv", row.names = FALSE
  )
}

add1. Define geography id. This is because the Geography-GEOID from initial file may be outdated

library(tigris)
library(maps)
# add block geoids
# get US blocks shapefile
blocks_VA <- st_as_sf(blocks(state="VA"))
blocks_MD <- st_as_sf(blocks(state="MD"))
blocks_DC <- st_as_sf(blocks(state="DC"))
blocks <- rbind(blocks_VA, blocks_MD)
blocks <- rbind(blocks, blocks_DC)

# lon and lat to geo-points
geopts <- provider %>%
  st_as_sf(coords = c("lon", "lat"), crs = 4269)
# indeces of bgs which contain a geopoint
inds <- st_within(geopts$geometry, blocks$geometry, sparse=T)
blk_list <- c()
for (i in inds){
  if (identical(blocks$NAME[i],character(0))){
    blk_list<- append(blk_list, NA)}
  else{
    blk_list <- append(blk_list, blocks$GEOID[i])}
}
provider['geoid_blk_new'] <- blk_list


# This assumes that you have longitude and latitude of each location and if you want bg or tract or county just replace 'blocks_VA <- st_as_sf(blocks(state="VA"))' with a corresponding value, e.g. 'tracts_VA <- st_as_sf(tracts(state="VA"))'. I believe it uses either tigris or maps package

add2 count providers per geography using matching codes

#obtain geoid
provider$GEOID <- substr(provider$geoid_blk_new, 1,12)
data_combined$GEOID <- as.character(data_combined$GEOID)
#summary by geoid: counts
provider_counts <- provider %>% group_by(GEOID) %>% summarise( hosp_cnt= sum(doctors))

#join hosp_cnt to geographies
data_combined <- data_combined %>% left_join(provider_counts, by= c("GEOID") )

add3 mean and median of 10 nearest drive times

#mean of 10 nearest
top_mean <- function(x) {  
   mean(head(sort(x ), 10) ) }
#median of 10 nearest
top_median <- function(x) {  
   median(head(sort(x ), 10) ) }
#apply rowwise
traveltimes_near <- data.frame(near_10_mean=apply(traveltimes, 1, top_mean), 
                               near_10_median=apply(traveltimes, 1, top_median)) 
#rownames_to_column(traveltimes_near, var = "GEOID")
traveltimes_near$GEOID <- row.names(traveltimes_near) 
#join mean median traveltimes to geographies
data_combined <- data_combined %>% left_join(traveltimes_near, by= "GEOID")

prepara data for save

#raw traveltimes: traveltimes matrix already estimated and with colnames arranged 
traveltimes <- read.csv("traveltimes_exercise.csv", row.names = 1)
#population: always recheck relevant population: ie. for pediatrics: pop 0-17 years
population <- data_combined %>% select(GEOID, population,hosp_cnt, near_10_mean.y, near_10_median.y)
# realign travel times
traveltimes <- traveltimes[as.character(population$GEOID), provider$ID]

save new data

write.csv(provider[, c("ID", "address", "lat", "lon", "doctors")], "provider.csv", row.names = FALSE)
write.csv(cbind(GEOID = rownames(traveltimes), traveltimes), "traveltimes_trimmed.csv", row.names = FALSE)
write.csv(population, "population.csv", row.names = FALSE)

next: catchment areas - ems

uva-bi-sdad/dc.hifld.ems documentation built on June 8, 2022, 9:11 a.m.