A geospatial approach to assess health coverage and scaling-up of healthcare facilities

Oinam Bakimchandra, Joymati Oinam & RK Kajal
Contd from previous issue
According to the ‘Rural Health Statistics 2012’ report under the Division of Statistics, Ministry of Health and Family Welfare, Government of India (GoI), there is a shortfall of 213 Primary Health Sub-Centres (PHSCs), 14 Primary Health Centres (PHCs) and 7 Community Health Centres (CHC) in rural areas of the state. It is to be noted that since Manipur is included under the Hill Area Development Programme, the total rural population was considered to be from hilly/tribal areas. In total, there are 420 PHSCs, 80 PHCs and 16 CHCs in the rural areas of Manipur state.
Geodatabase and methodological approach For this study, primary data and information on location of health facilities were collected by conducting GPS survey in different parts of Manipur. In addition, secondary data on healthcare facilities (both public and private) were collected from various health organizations in the state like the State Health Society, Manipur under NRHM and the Health Directorate, Government of Manipur. From the data about healthcare facilities, a geodatabase on their location was generated in which 421 PHSCs, 85 PHCs, 17 CHCs, One Sub-Divisional Hospital (SDH), seven District Hospitals (DH) and two State/Centrally funded public hospitals have been geo-tagged and field-verified for use in the study. For private hospitals and clinics registered under the Medical Directorate, Government of Manipur, locations of 54 out of 100 private healthcare facilities have been geo-tagged and updated in the geodatabase. Table 1 shows the distribution of healthcare facilities used in this study. Tertiary-level hospital infrastructure data of Regional Institute of Medical Sciences (RIMS), Imphal were collected for 2015 and 2017. In addition, elevation data (SRTM DEM, spatial resolution of 30 m), LULC map (Landsat 8–30 m spatial resolution), road network data and districts/sub-districts base vector layer of Manipur obtained from North Eastern Space Applications Centre, Government of India, Shillong, demographic/census data (according to Census 2011) and gridded population data for 2015 and projected for 2020 (1 km) were extracted from Gridded Population of the World (GPWv4).
The conceptual framework described and implemented in AccessMod (ver.5) software has been implemented in this study. It utilizes analytical capabilities of GIS to examine geographic aspects of the healthcare system.
Various geospatial parameters like LULC layer, road network layer, elevation layer, population coverage capacity (PCC), settlement layer, and geodatabase of the location of existing healthcare facilities are generated using ArcGIS. All these generated data layers, i.e. both raster and vector data are used as primary inputs for the model. For this study, spatial resolution of 30 m and projected coordinate system (UTM) are used for all the data  layers. Generation of each raster and vector layer for the model is described in the Supplementary Table 1. Input data on travelling scenario and population coverage capacity in tabular form were generated. To incorporate the mode of transportation and travelling speed of a patient through varying topography of the region, a travelling scenario text file was prepared. Based on the mode of accessibility in different LULC classes found in the region, this file describes the modes of transportation – as walking or motorized – along with their speed.
PCC for different types of healthcare facilities was computed. For population coverage estimation, norms issued by the high-level expert group report on universal health coverage for India, instituted by the Planning Commission, GoI during 2011 were adopted. It has been recommended that a PHSC should cover a population of 5000 (or 3000 in a remote, dangerous location); PHC about 30,000 or more (20,000 in remote areas) and CHC about 120,000 people in urban areas or 80,000 people in remote/tribal areas. It is to be noted that PCC for CHC according to the norms is 120,000, but as given in the Rural Health Statistic report, a CHC in Manipur covers a population of 1–3 lakhs. Health infrastructure data for the year 2015 and 2017 from RIMS, Imphal were obtained to compute PCC for 2015 and 2017. PCC was computed using the formula provided in AccessMod 5.
Here we provide an example to estimate PCC for 2015 (Table 3). PCC capacity of a tertiary-level hospital = (A x B x C)/(D x E x F), where A is the average no. of beds (1000 nos at RIMS), B the occupancy rate (55%), C the number of days in a year (365 days), D the total number of patients visiting RIMS per year divided by total population of Imphal valley (0.21), E the assumption that 15% of patients seek admission at tertiary-level hospitals (15%) and F is the average length of stay at the hospital (5 days). Therefore, average PCC of RIMS (corresponding to 2015) = 1000 x 55 x 365/0.21 x 15 x 5 = 1,274,603.1. PCC for a tertiary-level hospital (RIMS) for 2017 was computed as 900,886.36 using occupancy rate of 48.87% and average length of stay as six days. For the projected year 2020, PCC was computed as 1,801,772.72 by considering the projected population of 2020. The model was set-up for spatial catchment analysis of the existing healthcare facilities in the region using the generated input parameters as discussed earlier. The model integrates the travelling time, PCC and population distribution to generate the catchment area of each health facility. Maximum travel time of 60 min for PHSC, 90 min for PHC and 120 min for CHC has been assigned in the model. The results represent catchment areas of the existing healthcare network. Based on this analysis, a population grid layer has been generated which indicates the spatial location of the populations not covered by the existing healthcare facilities of the region. Subsequently, scaling-up analysis was performed to assess the requirement of establishing new healthcare centres in the region. This analysis seeks to address the requirement of new healthcare centres in the region to meet the expanded population. Catchment area and scaling-up analysis were performed for 2015, 2017 and the projected year 2020.
Spatial data analysis of the existing healthcare facilities network An exploratory spatial data analysis was carried out to determine the distribution of the existing healthcare facilities. Standard distance analysis, directional distribution analysis, kernel density mapping and nearest neighbour analysis were carried out. It was found that most of the existing healthcare facilities in the region are concentrated within a radial distance of 4.75 km from a mean centre located at the central part of Imphal valley. From directional distribution analysis, the direction of the axis of standard deviation ellipses appeared that the skewed distribution towards the northeast–southwest of the region, which is indicative of higher distribution of existing healthcare facilities in the northeastern and southwestern regions of the state. The output from the kernel density estimation of the existing healthcare facilities in the state is found to be highly concentrated in the Imphal valley. This is largely attributed to the very high population density in the valley compared to the hilly region of the state. It was observed from the nearest neighbour analysis that the locations of healthcare facilities are significantly spatially clustered amongst themselves, i.e. with an observed mean distance of 2.62 km and an expected mean distance of 3.56 km. It can be concluded from the analysis that given a Z-score of –12.16, there is less than 1% likelihood that this clustered pattern observed in the distribution of healthcare facilities is a result of any random choice. Existing healthcare facilities location coverage analysis Figure 5 shows the results of location coverage analysis for the existing healthcare facilities in Manipur. Based on this analysis, it can be clearly observed that all the districts in the valley area of Manipur are well covered and connected with existing healthcare facilities, i.e. Imphal East, Imphal West, Bishnupur and Thoubal districts. On the contrary, population in the hilly districts, i.e. especially in the northeastern, northwestern and southern parts of Manipur, and particularly the border areas was found to be insignificantly covered by the existing catchment area of the healthcare network. This may be attributed to the absence of proper serviceable road network connectivity in the interior border areas and sparse population density in the hilly region compared to the valley area. The state government has highlighted that lack of proper roadways poses hindrance to providing necessary healthcare facilities to people in the interior villages of the hilly districts in Manipur. Ukhrul, Churanchandpur, Chandel, Tamenglong and Senapati are such hilly districts that need intervention towards provision of basic medical facilities and opportunities, in addition to providing road and air ambulance services to the interior villages in these districts. The model results clearly show the spatial distribution and coverage area of the existing network of healthcare facilities in the region.
(To be contd)