Geographic Information System Software
Most GIS products use only a small fraction of the functionality of expensive and complex commercial GIS packages. A broad range of visualisation techniques and queries can be performed using free simple open source (OS) software. It was assumed that staff who are able to update and graph data in spreadsheets (such as Excel) would have the competencies to view the same data spatially using OS GIS software.
Three open source or free GIS tools were used in the three study districts: Cybertracker, free software for field data collection on GPS-enabled PDAs (personal digital assistant), was used to collect health infrastructure data; Open Jump, Java-based, open source GIS, was used to visualise health data and for simple analysis; and AccessMod© , a free extension from World Health Organisation (WHO), was used for service availability mapping (Figure 3).
Lack of reliable baseline data is commonly cited as a difficulty in implementing health GIS in the developing world [13, 14, 21, 35]. Addressing this difficulty, training included the use of Cybertracker for the collection of up-to-date and accurate spatial data at the local level. Data collection using Cybertracker is rapid [36], includes location coordinates, and can be exported directly to a GIS format for mapping and analysis. Health mapping with Open Jump and service availability mapping with AccessMod© used both new data collected using Cybertracker and existing data available from district health departments and clinics (Figure 3).
Open Jump was chosen for this study because of its intuitive interface, broad functionality and the availability of a well designed charting plug-in. OS software is of particular interest to developing countries where resources are limited and licensing costs can be prohibitive. A report commissioned by the UK government on intellectual property rights and international development, recommended that developing countries should consider OS software alternatives in their procurement policies [37]. The free and unlimited distribution of OS software and its ability to import a broad range of pre-existing data formats makes it an attractive alternative to expensive proprietary packages. Furthermore the source code is open, allowing modifications to be made locally to suit particular applications [38].
AccessMod© is an extension to ArcView 3.x GIS software (ESRI) developed for the World Health Organisation to model accessibility and geographic coverage of health care infrastructure. The primary application of AccessMod© is to produce a spatial model of accessibility as a function of travel time to health facilities. Travel time is determined by transport infrastructure, land cover (i.e. rivers, forest, grass land) and terrain. Multiple travel time estimates based on different transport modes (e.g. walking, public transport, private transport) can be produced. These data can be combined with population distribution data and health centre capacity information to produce a theoretical catchment for mapped infrastructure. Estimating access to health services as a function of travel time, rather than as linear distance, is a significant improvement particularly in rugged terrain where modes of transport vary [7, 9, 39–43].
Although ArcView is not free software, this component of the study was included to demonstrate to training participants who showed a particular aptitude for GIS how a more sophisticated level of modelling could be conducted using the collected and free spatial data. The service availability modelling was also designed to develop a broader understanding in district health departments, once basic skills were developed and data collected, of the potential of GIS with further investment. The choice, by WHO, to develop AccessMod© for ArcView 3.x over other GIS platforms (e.g. ArcGIS) was principally motivated by its continuing widespread use and availability in developing countries [42]. We similarly choose AccessMod© due to the common availability of ArcView, its free distribution and simple interface.
Training and equipment
Each district was provided with one laptop (~USD600), one external hard drive for data backup (320 GB) and one HP iPAQ 112 PDA (~USD350) with external Bluetooth GPS (~USD50). Initially, a 3 day training workshop was conducted in the use of Cybertracker for field data collection and Open Jump for data visualisation and querying. The trainees were mostly district health department staff and some provincial and clinic staff. Their selection was based on interest in learning health mapping and likelihood that they would have opportunities to use these skills in their work. The training examples and exercises used local data from each participating district. Six months after this training, further instruction was provided to selected participants from each district in service availability mapping. Participants were given all the required software, local spatial data, and video tutorials. The tutorials provided step-by-step, screen capture instructions with Indonesian language narration, (also available at the project website [44]), enabling participants to continue self-training after the training workshop.
Pilot Health Mapping Applications
GIS technologies were trialled in each district to guide practical implementation and test their effectiveness.
1. Rapid field data collection using PDA and Cybertracker software
Before this study, there were no comprehensive audits of health infrastructure at the district level. Management of hospitals and health clinics are the responsibility of the district government, whereas other health facilities (e.g. health posts) are funded by the central government directly to the sub-district governments. This leads to the potential for the poor coordination of health resource allocation between levels of government in the absence of reliable data about existing facilities.
An audit of health infrastructure was undertaken using a simple data entry sequence in Cybertracker designed to collect information about the type of facility, working infrastructure (electricity, water, beds) and staff (numbers of doctors, nurses, midwives). In each district, two staff were allocated a motor bike to record these data at all health centres.
2. Mapping health indicators using Open Jump software
Patient health data recorded at health clinics are reported each month, in an aggregated form, to the provincial and national health departments. Before this study, district health departments had little capacity to analyse the patient health data they collect, and as a consequence these data were rarely used to inform the allocation of health resources in their district.
Subdistrict and village administration boundaries were obtained as spatial data sets from the district planning agency (BAPPEDA). All the target districts had recently updated these data to include newly formed administrative divisions. Health data, collected at the village level and collated into annual reports were entered into the data base files (.dbf) associated with the spatial data for administrative boundaries using Microsoft Office 2003 or Open Office Calc (open source spreadsheet software). These data were then mapped in Open Jump using colour themeing and charting tools.
3. Service availability mapping using AccessMod©
Most of the rural population of NTT lives in villages, often with limited access to health facilities due to rugged topography, poor roads, limited transport and seasonal flooding. Travel time from homes to health facilities was estimated using AccessMod© , taking into account terrain and seasonal variations in access (e.g. flooding). The following variables influencing travel time were provided as spatial grids for input into AccessMod© :
-
(i)
Slope as derived from a high resolution digital elevation model (DEM). These data were obtained as a free download from the ASTER Global Digital Elevation Model (GDEM) program, a joint initiative of the Japan's Ministry of Economy, Trade and industry (METI) and NASA [45]. Slope was categorised into five classes from flat to vertical.
-
(ii)
Land cover was produced as a grid with four categories savanna, scrub, forest and rivers. For TTS these data were produced by classifying Landsat satellite imagery obtained as a free download from the U.S. Geological Survey [46]. In Ngada and Nagekeo landcover data were available from previous projects.
-
(iii)
Transport infrastructure comprised road data which were classified into three categories of road quality; national, provincial and district. These data were obtained from the local department of planning.
Each cell of each grid was then allocated an average travel time based on the mode of transport to be modelled, e.g. average speed walking through a scrub cell could be 2 km/hr whilst the average speed on a district level road using public transport may be 10 km/hr. These grids were combined within AccessMod© to produce an overall travel time grid. Using AccessMod© this grid was then intersected with the location of health facilities collected in pilot study 1 to produce models of travel time to selected health facilities.
Evaluation of the pilot health mapping applications
The relevance and effectiveness of the training in health mapping, as perceived by the district and clinic staff who had received the training, were assessed by surveying all the trainees using written questionnaires with closed and open questions, immediately after and six months after training.
The implementation of the pilot GIS applications were evaluated by interviewing staff and officials of the participating clinics, and districts and provincial health departments. These interviews consisted of open-ended questions designed to discover if health mapping had been used to inform resource allocation planning, or in advocating for public health programs, and if so, whether these processes were an improvement on previous practices.