Ghana

2. Ghana#

Author: Andres Chamorro
Date: March 15, 2023

The objective of this analysis is to construct health facility access indicators over the general population, and across the socioeconomic distribution for a set of GFF countries with available geolocalized facility level information.

Indicators of interest

Percentage of population within 2h of driving to the nearest primary care facility (population level, and by SES quintile).
Percentage of population within 2h of driving to the nearest district hospital (population, and by SES quintile).
Percentage of health facilities with direct access to an all season road.
Percentage of health facilities within 2km of an all season road.

Data Sources

Health facilities: National Registry Master List, we geo-coded entries with missing coordinates
Administrative Boundaries: World Bank GAUL Level 1/2
Global Friction Surface (2019)
Population 2020: WorldPop 1km Grid Unconstrained

2.1. Data preparation#

This first section conducts some common data preparation tasks:

Import libraries
Load input data (this notebook was ran from the DEC JNB server)
Merge population and friction surface raster data
Prepare origins (population grid) and destinations (health facilities)

Define inputs

##  health and admin data
iso3 = "GHA"
geocoding_output = f"{iso3}_geocoding_2.7.23.json"
input_dir = "/home/public/Data/PROJECTS/Health"
out_folder = os.path.join(input_dir, "output", iso3)
if not os.path.exists(out_folder):
    os.mkdir(out_folder)

Summary of health facilities by type

Hospital tag was assigned if "hospital" or "Hospital" is in facility name

None        9121
hospital     556
Name: facilitytype, dtype: int64

2.2. Calculate travel time to nearest health facility#

We calculate driving time from every populated place (using a 1 km population grid) to the nearest health facility, using the global friction surface.

2.2.1. Travel Time Maps#

../../_images/356a00809497d747d5e7d893e1b8b6dd69a109d33300fc73feb1351e87fc35a1.png

../../_images/b3528aaa960444ca575c08c5faf4248877f8bbf9e508157cd2cca77cdf2b0c00.png

2.3. Summarize population within 2 hours#

Summary stats

Summary of % of pop. within 2 hr. of health facility 
 count    110.000000
mean       0.991290
std        0.025454
min        0.779150
25%        0.992899
50%        0.998995
75%        1.000000
max        1.000000
Name: health_pct, dtype: float64
Summary of % of pop. within 2 hr. of hospital 
 count    110.000000
mean       0.975085
std        0.055193
min        0.670728
25%        0.980969
50%        0.996126
75%        1.000000
max        1.000000
Name: hospital_pct, dtype: float64

2.4. Access to roads#

Percentage of health facilities having direct access to an all-season road, by district (admin-2 level).

All-season road defined as primary, secondary or tertiary using the OpenStreetMap classification.

Direct access defined as being within 100 meters of a road.

2.4.1. Load OSM roads and define classification#

{
    'motorway': 'OSMLR level 1',
    'motorway_link': 'OSMLR level 1',
    'trunk': 'OSMLR level 1',
    'trunk_link': 'OSMLR level 1',
    'primary': 'OSMLR level 1',
    'primary_link': 'OSMLR level 1',
    'secondary': 'OSMLR level 2',
    'secondary_link': 'OSMLR level 2',
    'tertiary': 'OSMLR level 2',
    'tertiary_link': 'OSMLR level 2',
    'unclassified': 'OSMLR level 3',
    'unclassified_link': 'OSMLR level 3',
    'residential': 'OSMLR level 3',
    'residential_link': 'OSMLR level 3',
    'track': 'OSMLR level 4',
    'service': 'OSMLR level 4'
}

Show code cell source Hide code cell source

roads = gpd.read_file(join(input_dir, 'osm', iso3, 'gis_osm_roads_free_1.shp'))

adm2_json = json.loads(adm2.to_json())
epsg = epsg_get(adm2_json)

roads = roads.to_crs(epsg)
roads['OSMLR'] = roads['fclass'].map(osm.OSMLR_Classes)

def get_num(x):
    try:
        return(int(x))
    except:
        return(5)
roads['OSMLR_num'] = roads['OSMLR'].apply(lambda x: get_num(str(x)[-1]))

## Calculate buffers
roads_100m = roads.copy()
roads_100m['geometry'] = roads_100m['geometry'].apply(lambda x: x.buffer(100))

roads_2km = roads.copy()
roads_2km['geometry'] = roads_2km['geometry'].apply(lambda x: x.buffer(2000))

health = health.to_crs(epsg)
# hospitals = hospitals.to_crs(epsg)

## Intersect roads and buffer
# roads_1_100m = roads_100m.loc[roads_100m.OSMLR_num<=1].unary_union
roads_2_100m = roads_100m.loc[roads_100m.OSMLR_num<=2].unary_union
# roads_3 = roads.loc[roads.OSMLR_num<=3].unary_union
# roads_4 = roads.loc[roads.OSMLR_num<=4].unary_union

# health.loc[:, "bool_1_100m"] = health.intersects(roads_1_100m)
health.loc[:, "bool_2_100m"] = health.intersects(roads_2_100m)
# health.loc[:, "bool_3"] = health.intersects(roads_3)
# health.loc[:, "bool_4"] = health.intersects(roads_4)

# roads_1_2km = roads_2km.loc[roads_2km.OSMLR_num<=1].unary_union
roads_2_2km = roads_2km.loc[roads_2km.OSMLR_num<=2].unary_union
# roads_3 = roads.loc[roads.OSMLR_num<=3].unary_union
# roads_4 = roads.loc[roads.OSMLR_num<=4].unary_union

# health.loc[:, "bool_1_2km"] = health.intersects(roads_1_2km)
health.loc[:, "bool_2_2km"] = health.intersects(roads_2_2km)
# health.loc[:, "bool_3"] = health.intersects(roads_3)
# health.loc[:, "bool_4"] = health.intersects(roads_4)

health = health.to_crs(adm2.crs)
facilities = gpd.sjoin(health, adm2[['OBJECTID', 'WB_ADM2_CO', 'geometry']], how='left')

## Get percentages
# res_osmlr = facilities[['bool_1_100m','bool_2_100m','bool_1_2km','bool_2_2km','WB_ADM2_CO']].groupby('WB_ADM2_CO').sum()
res_osmlr = facilities[['bool_2_100m','bool_2_2km','WB_ADM2_CO']].groupby('WB_ADM2_CO').sum()
res_count = facilities[['WB_ADM2_CO','bool_2_100m']].groupby('WB_ADM2_CO').count().rename(columns={'bool_2_100m':'count'})
res_osmlr_pct = res_osmlr.apply(lambda x: x/res_count['count'])

res = res.merge(res_osmlr_pct, left_on="WB_ADM2_CO", right_index=True)

2.4.2. Maps of Access to Roads#

../../_images/68af3931c648153bc32c192719f12039d1ec3fdc57a6451977961a1711ca9d27.png

../../_images/76f6e98ac53d202724d86575be5fa8bbdb8f065c68075ede27e590b89e83e755.png

2.5. Access to health disaggregated by wealth quintile#

Categorize population grid by wealth quintiles, and then summarize the population with access to health (within 2 hours of health facility or hospital)