-
Notifications
You must be signed in to change notification settings - Fork 14
/
prep hospitals accessibility.py
441 lines (326 loc) · 16.7 KB
/
prep hospitals accessibility.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
import numpy as np
import pandas as pd
import geopandas as gpd
import os
from scipy.spatial import cKDTree
############
#
# Globals & Functions
#
############
data_dir = ".\\data"
GBcrs = {"init":"epsg:27700"}
# OS Open Roads accessed from: https://www.ordnancesurvey.co.uk/opendatadownload/products.html
open_roads_dir = os.path.join(data_dir, "OS Open Roads")
# Hospital Names and Locations: https://data.gov.uk/dataset/f4420d1c-043a-42bc-afbc-4c0f7d3f1620/hospitals
hosp_data_dir = os.path.join(data_dir, "hospital-data")
eng_hosp_file = "Hospital.csv"
wales_hosp_file = "Wales hospital list.xls"
# LSOA 2011 Pop Weighted Centroids: https://geoportal.statistics.gov.uk/datasets/lower-layer-super-output-areas-december-2011-population-weighted-centroids
lsoa_pop_weighted_centroids_file = "Lower_Layer_Super_Output_Areas_December_2011_Population_Weighted_Centroids.csv"
# Wards 2019 boundaries
ward_boundaries = "Wards_December_2019_Boundaries_UK_BGC.shp"
# LSOA 2011 Boundaries
lsoa_boundaries = "Lower_Layer_Super_Output_Areas_December_2011_Boundaries_EW_BGC.shp"
# Island wards to exclude - Isle of Scily and Wight wards
island_ward_boundaries = "Wards_December_2019_Boundaries_Scily_Wight_BGC.shp"
##########################
#
# Load separate data sources for Engalnd and Wales Hospitals
#
# These might be more outdated than the data above but the England file is geocode and only contains hospitals and the Welsh one
# contains info on type of hosptal (eg major/community/etc)
#
# Can use these to filter out just the hospitals I am interested in from the data above
#
# England hospitals: https://data.gov.uk/dataset/f4420d1c-043a-42bc-afbc-4c0f7d3f1620/hospitals
# Wales hospitals: https://www.whatdotheyknow.com/request/list_of_hospitals_3?unfold=1
#
##########################
dfEngHosp = pd.read_csv(os.path.join(hosp_data_dir, eng_hosp_file), delimiter="¬")
dfWlsHosp = pd.read_excel(os.path.join(hosp_data_dir, wales_hosp_file))
########################
#
# Filter hospitals by type
#
# Want to remove entries that are unlikely to provide support to
# patients suffering from COVID-19.
#
# This approximate because 1) the data doesn't contain hospital classification for all entries and
# 2) I do not have a good understanding of what type of hospital is able to provide care to COVID-19 patients.
#
# This page gives some information on Welsh hospital classifications: http://www.datadictionary.wales.nhs.uk/index.html#!WordDocuments/hospitalclassificationcategories.htm
#
#######################
# For the English hospitals, only three categoires given: 'Hospital', 'UNKNOWN', 'Mental Health Hospital'
# Filter out 'Mental Health Hospital'
dfEngHosp = dfEngHosp.loc[ dfEngHosp['SubType'] != 'Mental Health Hospital']
# Remove closed Wales hospitals
dfWlsHosp = dfWlsHosp.loc[ dfWlsHosp['Type'] != 'Closed']
# For the Welsh Hospitals there are many hospital types given. Select the following, based on http://www.datadictionary.wales.nhs.uk/index.html#!WordDocuments/hospitalclassificationcategories.htm
dfWlsHosp = dfWlsHosp.loc[ dfWlsHosp['Type'].isin([ 'Major A&E Units',
'Community',
'Acute',
'Major acute',
'Minor A&E Units'])]
##########################
#
# ONSPD Feb 2020
#
# Used to geocode postcodes in the Wales hospital data: https://geoportal.statistics.gov.uk/search?collection=Dataset&sort=name&tags=all(PRD_ONSPD%2CFEB_2020)
#
##########################
onspd_dir = os.path.join(data_dir, "postcodes\\Data\\")
onspd_chunks = pd.read_csv(os.path.join(onspd_dir, "ONSPD_FEB_2020_UK.csv"), chunksize=10000)
##########################
#
# Gather postcodes to to geocode
#
# All the postcodes in the NHS Trust Sites data and the postcodes in the Wales hospitals data
#
# There should be overlapping postcodes in these.
#
##########################
def clean_pcd(pcd_string):
# Remove all spaces
while " " in pcd_string:
pcd_string = pcd_string.replace(" ","")
pcd_string = pcd_string.upper()
return pcd_string
# Check there are no missing postcodes
assert dfWlsHosp['Postcode'].isnull().any() == False
dfWlsHosp['postcode_clean'] = dfWlsHosp['Postcode'].map(lambda s: clean_pcd(s))
pcds_to_geocode = dfWlsHosp['postcode_clean'].unique()
dfONSPDFiltered = pd.DataFrame()
for dfONSPD in onspd_chunks:
# Clean postcode
dfONSPD['pcds_clean'] = dfONSPD['pcds'].map(lambda s: clean_pcd(s))
# Filter just the ones we want
dfONSPD = dfONSPD.loc[ dfONSPD['pcds_clean'].isin(pcds_to_geocode), ['pcds', 'pcds_clean','lat','long']]
dfONSPDFiltered = pd.concat([dfONSPDFiltered, dfONSPD])
# Check that all postcodes have been found
assert pcds_to_geocode.shape[0] == dfONSPDFiltered.shape[0]
# Geocode wales hospitals
dfWlsHosp = pd.merge(dfWlsHosp, dfONSPDFiltered, left_on = 'postcode_clean', right_on = 'pcds_clean', how = 'left')
assert dfWlsHosp['lat'].isnull().any() == False
assert dfWlsHosp['long'].isnull().any() == False
# Create geo data frames
gdfWlsHosp = gpd.GeoDataFrame(dfWlsHosp, geometry = gpd.points_from_xy(dfWlsHosp['long'], dfWlsHosp['lat']), crs = "EPSG:4326")
gdfWlsHosp = gdfWlsHosp.to_crs(GBcrs)
gdfEngHosp = gpd.GeoDataFrame(dfEngHosp, geometry=gpd.points_from_xy(dfEngHosp.Longitude, dfEngHosp.Latitude), crs = "EPSG:4326")
gdfEngHosp = gdfEngHosp.to_crs(GBcrs)
gdfEngHosp.dropna(subset = ['Latitude','Longitude'], inplace = True)
##################################
#
# Combine Welsh and English hospitals into a single hospitals dataframe
#
##################################
gdfWlsHosp = gdfWlsHosp.rename(columns = {"Hospital":"OrganisationName", 'lat':'Latitude', 'long':'Longitude', 'Type':'SubType'})
gdfWlsHosp['OrganisationCode'] = None # Don't have this info for the Welsh hospitals
gdfEngHosp['postcode_clean'] = gdfEngHosp['Postcode'].map(lambda s: clean_pcd(s))
gdfHospitals = pd.concat([gdfEngHosp, gdfWlsHosp], join = 'inner')
# Create unique index that works for both english and welsh hospitals (since welsh hospitals data doesn't include organisation codes)
gdfHospitals["NamePCD"] = gdfHospitals['OrganisationName'].str.replace(" ","_") + "_" + gdfHospitals['postcode_clean']
# There are two two pairs of hospitals with the same name and same postcode, treat these as duplicates
gdfHospitals.drop_duplicates("NamePCD", inplace = True)
# Save this
gdfHospitals.to_file(os.path.join(hosp_data_dir, "England_and_Wales_Hospitals.shp"))
##############
#
# Read and clean data
#
##############
##
## Wards
##
# Load ward boundaries and get ward centroids
gdfWards = gpd.read_file(os.path.join(data_dir, "boundaries", ward_boundaries))
gdfWards = gdfWards.to_crs(GBcrs)
# Load island wards - selected manually using QGIS
gdfWardsIsland = gpd.read_file(os.path.join(data_dir, "boundaries", island_ward_boundaries))
gdfWardsIsland = gdfWardsIsland.to_crs(GBcrs)
# Only work with England and Wales wards for now
gdfWards = gdfWards.loc[ (gdfWards['WD19CD'].map(lambda s: s[0] =='E')) | (gdfWards['WD19CD'].map(lambda s: s[0] =='W'))]
# Exclude Isles of Scily and Isle of Whight for now - need to add in ferry crossings to connect with road network
gdfWards = gdfWards.loc[~gdfWards['WD19CD'].isin(gdfWardsIsland['WD19CD'])]
gdfWardsCent = gdfWards.copy()
gdfWardsCent['geometry'] = gpd.points_from_xy(gdfWardsCent.LONG, gdfWardsCent.LAT)
gdfWardsCent.crs = {"init":'epsg:4326'}
gdfWardsCent = gdfWardsCent.to_crs(GBcrs)
gdfWardsCent.to_file(os.path.join(data_dir, "boundaries", "ward_centroids.shp"))
##
## LSOAs
##
gdfLSOAs = gpd.read_file(os.path.join(data_dir, "boundaries", lsoa_boundaries))
gdfLSOAs = gdfLSOAs.to_crs(GBcrs)
dfLSOACent = pd.read_csv(os.path.join(data_dir, lsoa_pop_weighted_centroids_file))
gdfLSOACent = gpd.GeoDataFrame(dfLSOACent, geometry = gpd.points_from_xy(dfLSOACent.X, dfLSOACent.Y))
gdfLSOACent.crs = GBcrs
# Filter out centroids on Isle of Whilte and Isles of Scily
lsoa_cols = gdfLSOACent.columns
gdfLSOACent = gpd.overlay(gdfLSOACent, gdfWardsIsland, how = 'difference')
gdfLSOACent = gdfLSOACent.reindex(columns = lsoa_cols)
##
## Open Roads Data
##
# commented out because once data files have been combined and saved as a single file, this section
# of code doesn't need to be executed
# Combine mutiple OS Road Nodes files into a single geodataframe
road_node_files = [i for i in os.listdir(open_roads_dir) if "RoadNode.shp" in i]
gdfRoadNodes = gpd.GeoDataFrame()
for f in road_node_files:
gdf = gpd.read_file(os.path.join(open_roads_dir, f))
gdf['file'] = f
gdfRoadNodes = pd.concat([gdfRoadNodes, gdf])
gdfRoadNodes = gdfRoadNodes.to_crs(GBcrs)
gdfRoadNodes.to_file(os.path.join(data_dir, "OS Open Road Nodes.shp"))
road_link_files = [i for i in os.listdir(open_roads_dir) if "RoadLink.shp" in i]
gdfRoadLinks = gpd.GeoDataFrame()
for f in road_link_files:
gdf = gpd.read_file(os.path.join(open_roads_dir, f))
gdfRoadLinks = pd.concat([gdfRoadLinks, gdf])
gdfRoadLinks = gdfRoadLinks.to_crs(GBcrs)
gdfRoadLinks.to_file(os.path.join(data_dir,"OS Open Road Links.shp"))
########################
#
# Use road network to find distances to nearest hospitals
#
########################
import networkx as nx
import pandana as pdna
# Extract the road network topology from the gis data
print("Loading network topology")
# commented out because once network topology has been extracted from open road gis data this section doesn't need tp be executed
gdfRoadNodes = gpd.read_file(os.path.join(data_dir, "OS Open Road Nodes.shp"))
gdfRoadLinks = gpd.read_file(os.path.join(data_dir,"OS Open Road Links.shp"))
dfNodes = pd.DataFrame({"identifier":gdfRoadNodes.identifier, "x": gdfRoadNodes.geometry.x, "y":gdfRoadNodes.geometry.y})
dfNodes.set_index("identifier", inplace=True)
dfLinks = gdfRoadLinks.loc[:, ['startNode', 'endNode', 'length']]
dfNodes.to_csv(os.path.join(data_dir, "OS Open Road Nodes.csv"))
dfLinks.to_csv(os.path.join(data_dir, "OS Open Road Links.csv"))
gdfRoadNodes = None
gdfRoadLinks = None
dfNodes = pd.read_csv(os.path.join(data_dir, "OS Open Road Nodes.csv"))
dfLinks = pd.read_csv(os.path.join(data_dir, "OS Open Road Links.csv"))
dfNodes.drop_duplicates(inplace=True)
dfNodes.set_index('identifier', inplace=True)
gdfHospitals.set_index('NamePCD', inplace = True)
# Get largest connected component
print("Finding largest connected component")
edges = dfLinks.loc[:,['startNode','endNode','length']].values
G = nx.Graph()
G.add_weighted_edges_from(edges)
largest_connected_component = sorted(nx.connected_components(G), key = len, reverse=True)[0]
# Clean up to save memory
G = None
edges = None
# Create pandana network. It's much faster for nearest point-of-interest analysis
print("Creating Network")
# Filter nodes and edges to just include those in the largest connected componet
dfLinksLCC = dfLinks.loc[(dfLinks['startNode'].isin(largest_connected_component)) & (dfLinks['endNode'].isin(largest_connected_component))]
dfNodesLCC = dfNodes.loc[largest_connected_component]
net=pdna.Network(dfNodesLCC["x"], dfNodesLCC["y"], dfLinksLCC["startNode"], dfLinksLCC["endNode"], dfLinksLCC[["length"]])
#net.save_hdf5(os.path.join(data_dir, "pandana_network.h5"))
# Can then get the nearest 3 hospitals to each node in the network
print("Finding nearest hospitals")
search_distance = 200000
net.set_pois("hospitals", search_distance, 3, gdfHospitals.geometry.x, gdfHospitals.geometry.y)
dfNear = net.nearest_pois(search_distance,"hospitals", num_pois=3, include_poi_ids=True)
######################################
#
#
# Get distance to hospitals from Ward centroids
#
# Filter dfNear to include only the road nodes that are nearest to ward centroids
#
######################################
# Select only the Wards centroids and their nearest hospitals rather than every road node
ward_road_nodes = net.get_node_ids(gdfWardsCent.geometry.x, gdfWardsCent.geometry.y)
dfNearestHospitals = dfNear.loc[ward_road_nodes]
# Include Ward codes
gdfWardsCent['ward_road_node'] = ward_road_nodes
dfNearestHospitals.reset_index(inplace=True)
dfNearestHospitals = pd.merge(dfNearestHospitals, gdfWardsCent, left_on = 'identifier', right_on = 'ward_road_node', how = 'outer', indicator = True)
assert gdfWardsCent.shape[0] == dfNearestHospitals.shape[0]
# Check each ward Centroid has nearest 3 hospitals and
print("Checking hospitals found - wards")
assert dfNearestHospitals['_merge'].value_counts()['left_only'] == 0
assert dfNearestHospitals['_merge'].value_counts()['right_only'] == 0
assert dfNearestHospitals['poi1'].isnull().any() == False
assert dfNearestHospitals['poi2'].isnull().any() == False
assert dfNearestHospitals['poi3'].isnull().any() == False
########################################
#
#
# Get distance to hospitals from LSOA Centroids
#
# Repeate steps above but for LSOA centroids
#
########################################
# Get just the LSOA centroids and their nearest hospitals
lsoa_road_nodes = net.get_node_ids(gdfLSOACent.X, gdfLSOACent.Y)
dfNearLSOACentroids = dfNear.loc[lsoa_road_nodes]
# Include LSOA codes
gdfLSOACent['lsoa_road_node'] = lsoa_road_nodes
dfNearLSOACentroids.reset_index(inplace=True)
dfNearLSOACentroids = pd.merge(dfNearLSOACentroids, gdfLSOACent, left_on = 'identifier', right_on = 'lsoa_road_node', how = 'outer', indicator = True)
#assert gdfLSOACent.shape[0] == dfNearLSOACentroids.shape[0] This test fails for LSOAs. There are 3 pairs of LSOA centroids that are v close and share the same nearest road node
# Need to drop duplicate entries that get created from mergeing with duplciated road nodes
dfNearLSOACentroids.drop_duplicates(inplace=True)
# Check each ward Centroid has nearest 3 hospitals and
print("Checking hospitals found - lsoas")
assert dfNearLSOACentroids['_merge'].value_counts()['left_only'] == 0
assert dfNearLSOACentroids['_merge'].value_counts()['right_only'] == 0
assert dfNearLSOACentroids['poi1'].isnull().any() == False
assert dfNearLSOACentroids['poi2'].isnull().any() == False
assert dfNearLSOACentroids['poi3'].isnull().any() == False
#########################
#
# Clearing up and making output more presentable
#
#########################
##
## Wards
##
# Compute average distance to nearest 3 hospitals
print("Formatting and saving wards data")
dfNearestHospitals['mean_distance_nearest_three_hospitals'] = dfNearestHospitals.loc[:,[1,2,3]].mean(axis = 1)
dfNearestHospitals = dfNearestHospitals.reindex(columns = [ 'WD19CD','WD19NM','ward_road_node',
1,2,3,'poi1','poi2','poi3',
'mean_distance_nearest_three_hospitals'])
dfNearestHospitals = dfNearestHospitals.rename(columns = {1:'distance1',2:'distance2',3:'distance3',
'poi1':'hospital_postcode1','poi2':'hospital_postcode2','poi3':'hospital_postcode3',
'ward_road_node':'ward_centroid_road_node'})
dfNearestHospitals.to_csv(os.path.join(data_dir, "england-wales-nearest-hospitals-ward.csv"), index=False)
##
## LSOAs
##
print("Formatting and saving LSOAs data")
dfNearLSOACentroids['mean_distance_nearest_three_hospitals'] = dfNearLSOACentroids.loc[:,[1,2,3]].mean(axis = 1)
dfNearLSOACentroids = dfNearLSOACentroids.reindex(columns = [ 'lsoa11cd','lsoa11nm','lsoa_road_node',
1,2,3,'poi1','poi2','poi3',
'mean_distance_nearest_three_hospitals'])
dfNearLSOACentroids = dfNearLSOACentroids.rename(columns = {1:'distance1',2:'distance2',3:'distance3',
'poi1':'hospital_postcode1','poi2':'hospital_postcode2','poi3':'hospital_postcode3',
'lsoa_road_node':'lsoa_centroid_road_node'})
dfNearLSOACentroids.to_csv(os.path.join(data_dir, "england-wales-nearest-hospitals-lsoa.csv"), index=False)
####################
#
# Make a map
#
###################
print("Creating ward map")
from matplotlib import pyplot as plt
gdfWards = gdfWards.merge(dfNearestHospitals, left_on = 'WD19CD',right_on = 'WD19CD', how = 'inner')
f, ax = plt.subplots(1,1, figsize = (30,30))
legend_info ={'label': "Mean distance to nearest 3 hospitals (m)"}
gdfWards.plot(column = 'mean_distance_nearest_three_hospitals', ax = ax, legend = True, legend_kwds = legend_info, linewidth=0)
plt.axis('off')
f.savefig(os.path.join(data_dir, 'england-wales-nearest-hospitals-ward.png'))
print("Creating lsoa map")
gdfLSOAs = gdfLSOAs.merge(dfNearLSOACentroids, left_on = 'LSOA11CD',right_on = 'lsoa11cd', how = 'inner')
f, ax = plt.subplots(1,1, figsize = (30,30))
legend_info ={'label': "Mean distance to nearest 3 hospitals (m)"}
gdfLSOAs.plot(column = 'mean_distance_nearest_three_hospitals', ax = ax, legend = True, legend_kwds = legend_info, linewidth=0)
plt.axis('off')
f.savefig(os.path.join(data_dir, 'england-wales-nearest-hospitals-lsoa.png'))