GIS is a powerful tool for improving the understanding of data through visualisation and analysis, and is being increasingly used by public health professionals for planning, monitoring and surveillance [1]. Presenting data in maps can provide more insight than a table of the same data, enabling quick assessments of trends and interrelationships [2]. This capability can assist in targeting public health initiatives [3] as well as evaluating health programs and informing long term planning. Providing equitable minimum health services is a particular challenge in developing countries where health resources and transport infrastructure are often poor [4]. Access to health services is the primary determinant of utilisation of these services [5–7] and GIS tools are being increasingly used to evaluate the distribution of health resources (e.g. [7–11]).

Despite this potential, the use of GIS in developing countries is not widespread. There are perceived problems with GIS in low resource settings[12–14]; (i) GIS software can be expensive and complex, (ii) input data are often of low quality but presentation in GIS suggests veracity; and (iii) there may be a low capacity for data analysis with the danger of misinterpretation.

Various authors have described the potentially undemocratic nature of GIS and how it can exacerbate power imbalances through disenfranchising those without the skills and infrastructure required to work with the technology [15, 16]. Many GIS applications in developing countries rely on either proprietary software and/or a web interface requiring internet access [17–22]. Furthermore most of these applications require specialist data base skills to set up the back end data system for display and querying which limits health GIS to those with sophisticated data management skills and access to the internet.

A common response designed to democratise GIS has been the development of web based tools to allow more decentralized access to health GIS applications [18–20, 23, 24]. Paradoxically, due to low levels and unequal distribution of internet access both between and within countries, the web itself is seen by many to be an undemocratic tool in the developing world preferentially empowering an urban elite and further disenfranchising the rural poor [25–27]. For example the latest figures (2010) for internet users as a percent of population show 11% for Africa and 12.3% for Indonesia, compared to 58% and 77% for Europe and North America respectively [28]. However new opportunities for the more widespread use of GIS in low resource settings are emerging out of recent developments in GIS software and associated hardware.

This study addresses issues related to the prohibitive expense and complexity in GIS by exploring the possibilities presented by free or open source GIS software not reliant on internet access, decreasing hardware costs (e.g. GIS-ready laptops less than USD400), and the increasing availability of spatial data [17, 23] and spatial data collection tools (e.g. GPS enabled mobile phones). Through the introduction of a simple and inexpensive GIS in the developing context of eastern Indonesia, this study assessed whether these tools were appropriate for increasing the data analysis capabilities at local levels. Three GIS applications were implemented over a nine month period with mentoring support from Australian researchers. Training materials were tailored to the existing technical skills common in the young health department staff, and basic equipment was provided. The evaluation included an assessment of the uptake of the technology in this context and the sustainability of this uptake.

Study Area

Eastern Indonesia faces the challenges of providing adequate and equitable health services to a largely remote, rural population. Health in the eastern Indonesian province of Nusa Tenggara Timur (NTT) is generally poor, with high incidence of malaria [29], high infant mortality rate (54/1000, compared with 44/1000 nationally), and child malnutrition averaging 39% and reaching 50% in some areas [30].

In Indonesia, the responsibility for health service delivery is mainly at the district level. Due to decentralization since 2000, there has been devolution of budgeting and planning responsibility to the district level [31] without a corresponding provision of the base data and analysis skills required for evidence-based decision making. With the devolution of much authority and many functions to the district level, there have been disruptions to coordination and planning mechanisms, and difficulty in providing levels of public service delivery as high as those before decentralisation [32]. Currently, district and subdistrict health officers collect and collate health data for analysis at the provincial or national level (Figure 1a). However, with limited involvement in the subsequent processes of analysis, there is little sense of ownership of the data and data quality is often poor. Improving the quality of health data is a goal of a range of projects in Indonesia, yet there is little incentive at the local level to assure data quality when there is a limited understanding of how these data are to be used.

Increasing the capacity of district health officers to conduct data analysis could improve their understanding of the data they collect (Figure 1b) and so encourage the collection of more reliable data. District staff could base their health programs (Figure 1c) on the collected data and be empowered to advocate for targeted programs to address gaps in existing health services (Figure 1d).

Site Locations

The trial sites were located in three districts (kabupaten) in the eastern Indonesian province of Nusa Tenggara Timur (NTT), one of the five poorest provinces in Indonesia [33]. These districts were South Central Timor (Timor Tengah Selatan, TTS), in West Timor, and Ngada and Nagekeo, in central Flores (Figure 2). Prior to 2005 Nagekeo was part of Ngada district.

The health system in Nusa Tenggara Timur

In 2009, Nusa Tenggara Timur province (NTT) had a population of 4.2 million, and health facilities included 33 hospitals, 284 clinics (puskesmas), and more than 10,000 other local health facilities (pustu, polindes, poskesdes and posyandu) [34]. National government funding is allocated directly to the 21 city and district health departments for establishing and operating health clinics. The provincial health department has a role in coordination of health resources (staff and infrastructure) and systems for surveillance of health indicators, however subdistricts can apply directly to the national government for funding to support health posts.

Free and simple GIS health applications, without reliance on access to the internet, were deemed appropriate for low resource settings such as rural eastern Indonesia. District and clinic health staff demonstrated ready uptake of health GIS, and instigated and implemented a range of new health mapping applications, independent of external expertise. Maps were also being used by district health departments in NTT as tools to advocate for improved resource allocation, and in assessing the efficacy of public health programs, thus creating the potential for health mapping to inform policy [47]. Many district health staff are now offering training in health mapping to other staff members, independent of the original project. This will support the sustainability of this capacity at the district and clinic level.

We suggest that the effectiveness of the training in this pilot study was largely because simple software was used and the training was contextualized, i.e. based on exercises using data from the participants' districts. The use of local data in the training exercises prepared the trainees for applying the skills in their workplace. The step-by-step training materials, with Indonesian narration, supported continued learning after the training.

Uptake of health mapping by district and clinic staff was further enabled due to it being easily integrated into existing health information systems. These systems use spread sheets (district level) and hardcopy ledger (clinic level). Data from both these sources were easily imported into the open source mapping software (Open Jump) without additional data management systems.

Health data in developing countries are often of poor quality and the district staff who collect the data have little opportunity to use the data they collected to inform their own planning and resource allocation. Furthermore, data sharing between province, district and subdistrict is generally poor [48]. Rapid collection of health infrastructure data in the field over large areas empowered district health staff, for the first time, to collect reliable and complete health infrastructure data.

This study did not evaluate the quality of the patient health indicator data reported by the clinics. It is expected that timely mapping of health data by the district and clinic staff who collect and report these patient health data will highlight possible anomalies in the data set [49] and possibly lead to improvements in patient health data. Also, it is suggested that, if the staff reporting the data are also using the data to inform the public health programs they implement, there will be added incentive to ensure high quality data is collected. However even though spatial visualisation can be a tool for critical data analysis, this skill does not guarantee critical thinking. The ability to question the veracity of data, search for correlations, trends or inconsistencies are skills that need to be actively taught alongside the more technical competencies. Experience has shown that when trainees present the results of their health mapping to peers, critical thinking is encouraged both in the presenters and the audience. Although the results to date have been promising, a longer timeframe is required to evaluate the impacts of health mapping on policy and program development within district and provincial governments, and on the delivery and efficacy of health care.

Some cautionary observations were made about the specific software and hardware used. Data collection using Cybertracker is simple, however we found that creating a database within Cybertracker was limited to those with reasonable IT literacy. Some hardware problems were encountered, including difficulty maintaining a Bluetooth connection between the PDA and GPS, and a short battery life of the PDA. To ensure no field time was lost due to hardware problems, some surveys were augmented by using a standard GPS to log points, with data recording manually. Alternative PDA hardware with inbuilt GPS and a supply of batteries has largely overcome this problem.

Open Jump was effective for visualisation and preliminary analysis of recorded data. Some initial limitations with the Open Jump charting and printing functions were overcome through collaboration with the open source software engineer originally responsible for developing these components. Open source software is, by its nature, open to collaboration and sharing, and there is a large community of software developers ready to improve and refine open source tools based on intelligent feedback.

The requirement for licence software to run AccessMod^© is a significant limitation for its wider adoption although it effectively demonstrated the utility of simple modelling using free and easily collectable data. A further constraint was the limited availability of the demographic data required for comprehensive analysis. Whilst AccessMod^© provided a general indicator of coverage of health facilities, the lack of population distribution data precluded, for example, the delineation of patient catchment areas for a particular health facility. However, staff of the district health departments have local knowledge about the location of population in their district. With mapping skills developed at the district level, this local knowledge can be incorporated into the interpretation of resource allocation maps, even in the absence of precise population information.

The effectiveness of free and simple GIS was demonstrated at district and clinic levels in the low resource setting of rural, eastern Indonesia. The uptake and continued use of these technologies was a demonstration of decentralised GIS, empowering those with local knowledge and public health skills. This technology was deemed appropriate because it did not require internet access, a centralized database or a need for database development or maintenance.

This low cost, internet-free approach has broad relevance in the context of the move towards the decentralisation of governance and service provision in the developing world [50, 51]. We argue that whilst most of the world is still without internet access, particularly outside urban centres, dependence on web-based health GIS applications will tend to re-centralise the power of GIS and thus work against many political reforms occurring in the developing world. To truly democratize the power of health GIS, appropriate 'disconnected' technologies need to be available for the least empowered in the developing world.