Malaria diagnosis and mapping with m-Health and geographic information systems (GIS): evidence from Uganda

Nowadays m-Health technology has gained attention because of the latest developments in medical technology [7]. The application of m-Health on chronic disease outcomes in low- and middle-income countries shows positive impacts [8]. Projects based on m-Health are increasing in Africa and their effectiveness has been proven at several levels, including specialist advice in remote areas [9], drugs supply and stock management, and contribution to national health management [10]. One of the main challenges for m-Health applied to malaria is the reliability of the parasites screening technologies.

Several kinds of m-Health technology and, specifically, of mobile microscopy applications are already described in the scientific literature regarding the detection of soil-transmitted parasite worms [11] or hematologic and infectious diseases [12]. The Cellscope mobile microscope (a mobile phone connected through a frame to a lens) meets the Digital Imaging and Communications in Medicine (DICOM) standards [13]. A reversed-lens mobile phone microscope can generate imaging with little distortion and sufficiently high resolution to be employed for the detection of parasites in blood and stool samples [14]. This solution has excellent potential for further developments especially considering the minimal cost of the apparatus (approximately 30 USD, excluding the mobile phone). Potential application to diagnose malaria was demonstrated through the detection of haemozoin crystals in the blood smear [15]. Therefore, m-Health, and in particular mobile microscopy, can provide a low-cost method of detecting malaria in remote areas, but also could become a precious source of real-time and geo-located data on malaria. Consequently, the integration of m-Health and GIS with existing health systems will boost the chances of identifying the spatial and temporal pattern of malaria and responding accordingly [16]. This will also highlight the environmental [17], climatic [18], and socio-economic [19] risk factors for malaria. As a matter of fact, GIS has become a central component of the vector-borne disease risk-assessment process in public health and epidemiology, and it is recognized as a functional element in the roadmap for malaria elimination [20]. However, the adoption of m-Health technologies in rural areas is subject to the ICT infrastructure bottleneck. In the case of an inadequate ICT backbone, m-Health could widen the existing gap between urban and rural populations.

This paper aims to evaluate theoretically if the ICT infrastructure in a developing setting could be able to support m-Health technology in reaching the portion of the population with limited access to proper health diagnosis. Furthermore, the paper examines the feasibility of m-Health technology for malaria detection in rural settings.

According to World Bank statistics [21], Uganda has a population of 37.78 million people (2014), out of which 19.5% are below the poverty line (in 2012, down from 33.8% in 1999). Life expectancy at birth is 59 years. Average monthly temperature (1901–2009) ranges from 23.9 °C in February to 21.6 °C in July, whereas average monthly rainfall peaks in April (149.5 mm) and floors in December (34.4 mm). According to WHO, 48% of the population in 2013 is aged under 15 and only 4% is over 60 [22]. About 85% of the population lives in rural areas.

According to IndexMundi [23], which refers to CIA World Factbook (August 2014), the major infective diseases in Uganda are:

However, Malaria is recognized as the leading cause of morbidity in Uganda with 90–95% of the population at risk and it contributes to approximately 13% of the mortality of under five-year-old children [24]. According to the President’s Malaria Initiative (PMI) [25] malaria prevalence in children (up to 59 months) is 42%, but in rural areas in the northern region, this can climb up to 63%. Malaria is highly endemic in most of the country, and Uganda has some of the highest transmission rates in the world. Falciparum is the major source of infection, responsible for 99% of malaria cases. Pyrethroid and carbamate insecticide resistance has been documented in the country. Therefore, malaria places a huge burden on the Ugandan health system accounting for 30–50% of outpatient visits and 15–20% of hospital admissions and 9–14% of patient deaths [25]. The overall malaria-specific mortality is estimated to be between 70,000 and 100,000 child deaths annually in Uganda [26].

The Ugandan national health system is organized at the national level and in district level health centres [25]. At the top of the national health service are the two national referral hospitals, which offer general hospital and comprehensive specialist services, in addition to training and research facilities for medical students. Regional Referral hospitals (14 in total) offer general hospital plus some specialist services. At district level health facilities are organized in general hospitals and four categories of health centres: HC-I, II, III, and IV, which respectively serve at the community, parish, sub-county and county level. Only general hospitals, HC-III and HC-IV have laboratory facilities. HC-II only provide outpatient care and community outreach, while HC-I does not even have physical infrastructure but consist of groups of volunteers (Village Health Teams, VHT) that engage mainly in primary healthcare, prevention campaigns and promoting health services at community level.

In Uganda there are 25.3 million mobile phones, thus a SIM penetration of 64% (2014) and mobile network coverage of 75% of the population and 65% of the land (2013) [27]. Connections, SIM card penetration, and mobile network coverage are growing fast.

To assess m-Health feasibility in Uganda, the authors mapped the distribution of cellular base stations (2G and 3G), the clinical incidence of falciparum, the modelled population density and the travel time to main cities.

The Uganda Communication Commission produced a map of the distribution of 2G and 3G cellular base stations in Uganda [27]. Semi-automatic capture screen software elaborated this map creating a mosaic of images which was assembled and geo-referenced using Esri ArcGIS (version 10.0). The position of both types of the cellular base stations was digitalized. The mobile network coverage was estimated considering that the signal from a base station can cover a radius of approximately 10 km. This assumption was adopted because of the undetermined topographic position and, most important, the undocumented type, brand, and version of the cellular base stations deployed in Uganda. Therefore, using GIS software, a 10 km buffer was mapped around each cellular base station to represent the network coverage.

The modelled clinical incidence of falciparum malaria geo-dataset [28] was multiplied by the population distribution geo-dataset [29] in Uganda using GIS. The output represents the number of persons that are affected by falciparum malaria according to the model per unit of surface.

Two final assumptions are made: first, the equipped healthcare structures are located only in major cities and second, that people would not like to spend more than 1 h to travel to healthcare facilities. The extra time that people spend waiting for testing in the healthcare facility is not considered. Using the global map of accessibility [30] calculations are made about how many patients affected by falciparum malaria live more than 1 h away from a major city and thus are unwilling to travel to the healthcare facilities.

In GIS environment, the zone statistic tool allowed calculating the number of modelled falciparum malaria patients per district, the number of falciparum malaria patients that are within 10 km from the 2G and 3G mobile network, and the number of falciparum malaria patients covered by mobile network residing more than 1 h of travel away from a major city (i.e. healthcare facility equipped with laboratory facilities).

The implementation of m-Health will drastically transform the public health management system. Figure 5 represents a traditional malaria management system (blue arrows represent physical movement, red arrows decision fluxes, green arrows information and data flow, and yellow arrows drugs supply).

Patients with malaria symptoms reach a healthcare facility where trained medical staff can diagnose malaria (step 1 in Fig. 5). However, there are some barriers to adequate diagnosis and treatment: (a) remoteness of villages and lack of reliable transportation infrastructure, (b) poor equipment, (c) little specialist training, and (d) economic constraints (endemic poverty of rural population). Alternatively, the remote population can use the rapid diagnostic test (RDT) provided by drug suppliers (step 7 in Fig. 5). RDTs, however, have several limitations (e.g. less accurate diagnosis). Data collected at the healthcare facility is analysed by spatial and decision support systems (SDSS) (step 2 in Fig. 5), computerized management systems aimed to smoothen complicated geographic issues [31]. These systems are usually based on a GIS platform and aim to back a better-informed decision of public health authorities (such as Ministry of Health) on malaria elimination (step 3 in Fig. 5).

An example of SDSS applied to malaria is represented by the system that has been developed in the South West Pacific archipelago to automatically locate and map the distribution of confirmed malaria cases, rapidly classify significant transmission centres, and guide targeted responses [32, 33]. As the SDSS relies on effective case detection, the performance of this system is dependent upon the quality of case reporting. However, healthcare facilities collect data not in real-time and with poor spatial accuracy [34]. Indeed, official records are likely to be linked to a health unit, a district, a municipality, or another level of spatial aggregation. Although data can be stored at the individual patient level, the spatial dimension is restricted to an aggregation that can hide crucial local diversity and thus hinder control efforts. Consequently, public health authorities cannot design efficient and flexible responses to malaria disease and cannot adequately direct the healthcare personnel (step 4 in Fig. 5). Moreover, doctors in healthcare facilities do not have a clear picture of transmission in the rural area because they are not onsite, and malaria testing offsite could be unreliable due to errors and inaccuracy. Therefore, they give directives to drug suppliers (step 5 in Fig. 5) to provide medicine to the population in an approximate way, resulting in inappropriate drugs delivery to the rural population and a waste of medical resources.

Figure 6 shows the same process with the innovative introduction of m-Health technologies integrated with GIS. The main innovation is the establishment of mobile points of care equipped with m-Health technologies (i.e. mobile microscopy). They help overcome some of the barriers obstructing the access of rural populations to proper malaria diagnosis and treatment (step 8 in Fig. 6). As a matter of fact, remote diagnosis helps to decongest health facilities where sick patients fuel malaria diffusion, and it significantly reduces costs and the challenges of timely transportation. In particular, in Uganda was proven that introduction of community healthcare services can reduce significantly the number of patient visits presenting as malaria and change the profile of cases seen at health facilities [35].

Moreover, the performance of any SDSS depends on a correct approach to key health system components, specifically at healthcare facilities and community levels [36]. The introduction of m-Health technology will automatically link case recording with the geographic position of the detection, almost in real time, limiting the discretion of the operators that record cases. This will help to overcome conventional SDSS challenges.

First, the m-Health/GIS integration will strengthen community engagement and malaria monitoring at local level encouraging initial treatment. As other scholars suggest vigilance at the community level is a best practice in the anti-malaria strategy in remote areas [37, 38].

Second, the diagnosis and case reporting will be more accurate, timely, and useful along the chain from local communities to healthcare structures, district authorities, and the Ministry of Health. To this extent, the preservation and safeguarding of the information flow will require the introduction of TLC stakeholders in the malaria management scheme (step 2a in Fig. 6).

Third, timely and accurate reporting (including geographical information) granted by m-Health technologies is a key factor for the effectiveness of any surveillance-response system. Indeed, the ability of health system components to geo-reference data and properly define cases influences the capability of the SDSS to routinely map the allocation of cases and precisely categorize malaria transmission hotspots.

The combination between m-Health technologies and GIS in an integrated SDSS could enable time and space shortcuts. The intensity of edge-connecting nodes (e.g. ill patients to diagnostic centres) depends on TLC signal quality (3G versus 2G standards). The integrated SDSS provides an efficient protocol to visualize and communicate the distribution pattern of malaria transmission a high degree of accuracy. Moreover, integrated GIS queries could enable malaria programme managers to place specific positive cases at a detailed (i.e. household) level; recognize, pick and map areas of main concern (e.g. malaria hotspot) prioritising the intervention; steer the choice of suitably focused, exclusive reaction; and dig out comprehensive additional statistics and facts.

The necessity of gathering more high quality data over a broad geographical area has already been indicated by malaria specialists in Uganda [25]. A network of sentinel sites (UMSN—Uganda malaria surveillance network) has been set up to collect data, ranging from malaria morbidity to mortality of carriers after exposure to insecticides. However, the UMSN is too small to allow a meaningful comparative analysis. The new technology proposed in this paper would allow creating a rich database from which patterns and trends could be identified through Big Data analysis. In order to gain a deeper understanding of the link between malaria and environmental, social and economical conditions it is, therefore, advisable to have a multidisciplinary approach to the subject, gathering data from fields such as climatology, entomology, education, genetics, geography, economics, and more. Big Data and GIS would help discover and visualize the emerging patterns and suggest a preferred course of action.

Assembling this comprehensive dataset facilitates rapid and efficient decision-making by public health authorities in marginal and remote areas and permits rapid and well-organized budget determination, resource allotment and workforce recruitment to aid the execution of actions within acknowledged disease hotspot (step 4 in Fig. 6). Initially, an RDT approach will be integrated into the new system (step 7 in Fig. 6) with the target to be substituted by mobile microscopy. Finally, the distribution of drugs will be enhanced and will be delivered to the appropriate patients avoiding the waste of resources on persons who are not sick (step 6 in Fig. 6).

Upgrading to 3G or 4G standards can substantially improve the efficacy of m-Health technologies. Network optimization exemplified in Fig. 6, softens location problems. Physical barriers will be partly overcome by technology, and organization will change sensibly. As a consequence:

Considering the current trend to develop open source software for spatial analysis and geographic visualization, implementing such a system with those capabilities does not require significant financial investment on hardware or software, but only the human-resources investment to set up and maintain the system. A strategy of developing self-paced training packages that can be used to training individuals could possibly be used to customize some GIS software to minimize end-user skill requirements for its use [37].

The present study highlights also that the current 3G network is not sufficiently extended to support a mass application of m-Health technologies. Therefore, in remote areas where only the 2G mobile network is available, mobile microscopy could be conveniently used in offline mode if the Smartphone is equipped with automated cell-count software, which would enable field operators to diagnose malaria without a remote consultation through m-Health. In this case, the diagnosis would be instantly made on site at the remote point-of-care while the patient’s record would be transmitted to a larger health centre when faster (3G) connectivity is restored.

Future implementation of m-Health in the diagnosis and treatment of malaria should address some technological challenges:

An important concern is the quality of the data that governments have to manage and input in any disease control and surveillance system. Data transmission and management should be reliable, accurate, and safe, and at the same time the financial costs of m-Health/GIS system implementation should have the minimum impact on the poorest part of the population. If not properly addressed these issues will have three results:

Therefore, future research should focus not only technological aspects (mobile microscopy and GIS) of the innovation process but also on the social and economic impact of new technology on the stakeholders described in Fig. 6 (e.g. network analysis of stakeholders’ relationship).

Further refinement and validation of integration between SDSS and m-Health technologies should include a cost/benefit assessment of the surveillance systems. While the application of m-Health strategies involves initial costs, subsequent savings should rapidly bring to break-even and then to a positive payback.

Further studies to improve the effectiveness of the m-Health feasibility model should address healthcare facilities location—a central node for both the remote diagnosis and treatment. Identification of rural communities that have complicated access to the healthcare system is also crucial.

This study is limited to malaria in Uganda. However, exploring additional applications for various other vector-borne and infectious diseases such as dengue, filariasis, chikungunya, kala-azar, and Japanese encephalitis are very promising [39].