In uva-bi-sdad/dc.hifld.urgent: Social Data Commons: Urgent Care Facility Access Scores

knitr::opts_chunk$set(echo = TRUE)

libraries

library(sf)
library(tidygeocoder)
library(dplyr)
library(community)
library(readr)

clean df

#start clean
rm(list=ls())

working directory

setwd("~/VDH/Floating_Catchment_Areas/va/1. urgent_care_va_model/")

load data provider

#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.va_pl_hifld_2020_urgent_care_facilities "
)

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

#c) Disconnect
RPostgreSQL::dbDisconnect(con)

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

load data provider

provider <- read.csv("urgent_care_va.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)))

under construction

library(tidycensus)
library(tidyverse)

census_api_key("eba406410c653b81d6a795ac4e989221f7bdf302")

# Bring in census tract data. 
va <- get_acs(geography = "block group", 
                        year = 2019,
                        variables = c(population = "B01003_001E",
                                      medinc = "B19013_001E"
                        ),
                        state = "VA",
                        survey = "acs5",
                        output = "wide",
                        geometry = TRUE)

# data combined
#centroid and coordinates

data_combined <- data.frame(
  GEOID = va$GEOID,
  population = va$population,
  medinc = va$medinc,
st_coordinates(st_centroid(va$geometry))
)

data_combined <- data_combined %>% filter(!is.na(data_combined$X))

data_combined <- data_combined %>% filter(GEOID != 517000323001 )
data_combined <- data_combined %>% filter(GEOID != 517000323002 )
data_combined <- data_combined %>% filter(GEOID != 517000323003 )

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
  )
}

traveltimes <- read.csv("traveltimes_exercise.csv", row.names = 1)

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 This is because the Geography-GEOID from initial file may be outdated

library(tigris)
library(maps)
library(sf)
# 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_DC <- readRDS("~/VDH/Floating_Catchment_Areas/dmv/blocks_DC_2014" )
# blocks_MD <- readRDS( "~/VDH/Floating_Catchment_Areas/dmv/blocks_MD_2014" )
blocks_VA <- readRDS("~/VDH/Floating_Catchment_Areas/dmv/blocks_VA_2014" )

# blocks <- rbind(blocks_VA, blocks_MD)
# blocks <- rbind(blocks, blocks_DC)

blocks <- blocks_VA
# 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

#providers w geoid
provider$GEOID <- substr(provider$geoid_blk_new, 1, 12 ) 
#data_combined with geoid: ok
num_providers <- provider %>% group_by(GEOID) %>% summarise(prov_cnt = sum(doctors) )

#join providers to block groups
data_combined$GEOID <-  as.character( data_combined$GEOID)
data_combined <- data_combined %>% left_join(num_providers, by= "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, prov_cnt, near_10_mean, near_10_median)
# 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)

uva-bi-sdad/dc.hifld.urgent documentation built on Jan. 30, 2023, 5:42 p.m.