Based on the above analysis, we discuss below our observations and thoughts for developing a successful WGPHSS in terms of health data collection and update, data analysis, user interface design, use of maps and charts, system maintenance and compatibility, privacy issues of health data, development costs, and training. We then discuss the limitations of this review.
4.1. Health Data Collection
Data collection methods vary widely among the WGPHSSs. Although most systems still use official health data (e.g., EpiScanGIS
uses the meningococcal case registry data from the National Reference Centre for Meningococci (NRZM) [49
]), emerging technologies, especially Web 2.0, have profoundly revolutionized the way through which health data is collected.
First, data from news and social media, and Web-user search, have been extensively used for PHS. In addition to the six WGPHSSs analyzed in this paper, previous studies also demonstrated the widespread use of this informal online information. For example, Twitter has been used to monitor dental pain [50
], H1N1 [51
], cholera [52
], and Enterohemorrhagic Escherichia coli (EHEC)/ hemolytic uremic syndrome (HUS) [53
]; Web-user search data has been proven to have the possibility to effectively monitor H1N1 [54
] and toxicological outbreaks [55
]; and Google Flu Trends
has been employed for flu surveillance in various countries worldwide [56
]. Unlike official data, these informal data are much less costly to obtain, more timely (often 1–2 weeks earlier [52
]) thus allowing for early outbreak detection and responses, more frequently posted, and are generated worldwide making large spatial scale PHS a reality [50
]. Second, the crowdsourcing technology has enabled public participation in PHS, hence making it possible to collect information not included in traditional datasets [63
]. The public can engage via smartphones (usually equipped with a built-in Global Positioning System (GPS) and apps for PHS [64
]), online social network applications [65
], or other mobile devices (e.g., tablets and netbooks). For example, end users of HealthMap
can submit information on local outbreaks via the website or the app—HealthMap: Outbreaks Near Me
—on mobile devices with Apple iOS or Android platform. More in-depth and detailed information about crowdsourcing technologies for PHS can be found in the review by Kamel Boulos et al.
]. Data from this voluntary participation can be viewed as proactive, while conversely, data from news and social media, and Web-user search can be considered passive.
Nevertheless, problems exist in both approaches. First, certain biases are probably inevitable in informal data sources [52
], including geographical biases (media are much more prevalent in developed regions), demographic biases (a certain gender or age group may contribute more data via the social media platform), and cultural biases (different meanings in different regions or languages [67
] can result in misguided reporting and therefore lead to inaccurate or even false information). Thus, besides existing platforms such as SwiftRiver
], more filtering and validating models are needed for informal data before it can be used effectively and accurately [66
]. There are encouraging signs, however, that informal data can be used reliably, as preliminary experiments from DIZIE
’s development team have verified that respiratory data in the U.S. from Twitter correlates well with data from the CDC [70
]. Second, challenges arise in extracting and integrating information from data in “different file formats, schemas, naming systems” [71
] and languages, and are subjects being discussed in several studies [72
]. Third, existing ethical issues are rarely explored in passive data collection (e.g., from Twitter). To use the application or service, consumers are likely to accept the agreement for sharing their personal information without fully understanding associated risks [63
]. Despite these limitations and potential issues, informal data is a promising source for PHS and at minimum can be a highly useful complement to official data. As in the case of an outbreak, where even if social media cannot confirm an outbreak, it can “contribute to an investigation” [69
]. Furthermore, for resource-constrained regions where official data is scarce, informal data may become a vital source of health information [74
]. As such, future WGPHSS development will certainly continue to increasingly leverage these resources.
4.3. Data Analysis
Data analysis capabilities in existing WGPHSSs should be strengthened and extended as these analytical tools are vital in transforming raw data into knowledge. The ultimate goal of PHS is to inform the health planners and decision makers who are responsible for potential disease control, intervention, and prevention. This goal requires substantially more information than what raw health data provides. Without sound analyses, and in particular those supported by proper statistical models, to produce new and value-added information, WGPHSSs will remain “data rich” but “information poor” [77
]. In fact, only by combining the use of mapping and data analysis techniques can spatial disease patterns be identified [78
]; however, our findings suggest that most of the identified WGPHSSs focus solely on data sharing and visualization. To uncover more useful, value-added information, WGPHSSs should employ more spatial and non-spatial analysis methods, especially when they are used by decision makers. As an example, Bayesian spatial analysis can be incorporated to provide more reliable estimates of disease risks in small areas [79
]. Some WGPHSSs use standalone data analysis software as a solution to address the lack of embedded analysis tools. For example, EpiScanGIS
uses SaTScan to perform cluster analysis [49
]. Although more evidence is needed, we believe embedding analysis functionalities in WGPHSSs can streamline the process of PHS. Given that the implementation is much more complicated in a Web context than in the desktop context, a possible solution would be for WGPHSSs to incorporate an existing suite of online analytical tools. For example, Google Flu Trends
enables comparison of flu trends across regions with the online data analysis tool—Google Public Data Explorer. Breast Health Portal
and Project Safety Net
will similarly enable the use of Google Public Data Explorer in upcoming versions [80
]. Likewise, Bernier et al.
] have successfully combined GIS and Spatial On-Line Analytical Processing (SOLAP)—a technology devoted to spatio-temporal surveillance data analysis, employed primarily in Business Intelligence (BI)—for improved monitoring of climate-related health issues. An alternative variation is FluBreaks
, a WGPHSS that uses Google Flu Trend
data while providing its own sophisticated suite of statistical analysis tools. Future WGPHSS development should explore how to further leverage these combination modes.
4.4. User Interface Design
An effective WGPHSS relies on a good user interface. As stated by Frank [82
], “the user interface is the system”. WGPHSS users come from vastly different backgrounds, and most are not technical, cartographic, or statistical experts. A good WGPHSS should allow end users to expediently access and query data even if they are unfamiliar with GIS [83
]. As such, potential end users are key stakeholders and their needs and opinions regarding user interface design are of paramount importance and must be solicited if a WGPHSS is to be successful. For instance, non-technical people were consulted to help design the user interface of the Multi-Agency Internet Geographic Information Service
(MAIGIS) project [83
]. Furthermore, we found that interactive mapping has become an indispensable functionality of a WGPHSS. However, improper map customization options risk misinforming end users, making appropriate customization-settings design a necessity. WGPHSS designers should follow sound cartographic and statistical principles when designing the default customization settings in an interactive Web context [84
]. A good user-customized health map should enable not only easy interpretation, but also effective communication of health information, especially in instances where the raw data used to produce the maps is not tabulated. In a pilot study exploring public health practitioners’ preferences and visual perceptions about geographic information, Koenig et al.
] found that novice users of disease maps had difficulties comprehending data classifications and information expressed by charts (e.g., histograms and plot boxes) and understanding linkages between tabular data and disease maps. Moreover, novices frequently disagreed on what constituted appropriate color schemes. Therefore, WGPHSS designers should carefully configure these settings and determine whether end users will be allowed to change data classification methods (e.g., equal intervals, natural breaks, quantiles, optimal classifications, etc.
) or color schemes (sequential or diverging color scheme), and the extent to which these elements can be changed.
4.5. Use of Maps and Charts
Current WGPHSSs can benefit from incorporating different types of maps and charts to represent health phenomena for analytical purposes. We found that animated maps are being underused in existing WGPHSSs compared with static and interactive maps. However, we believe animated maps should be much more widely adopted in PHS due to their ability to capture the inherent temporal feature of health phenomena. It has also been demonstrated that users can more accurately identify space-time clusters from animated maps than from multiple static maps [85
]. Several factors facilitate the increased use of animated maps in WGPHSSs, namely, decreased costs, increased bandwidth, better streaming technology, smaller-sized format (vector rather than raster-based animation), automated creation procedures [86
], and available animation technologies (e.g., Adobe Flash, Microsoft Silverlight). Some researchers asserted that animated maps could be further improved and even more useful if they provided more options, such as a pause functionality and pace manipulation [87
]. However, rapidly improving technologies have since rectified these missing functionalities and rendered those concerns non-issues. For example, U.S. Cancer Statistics: An Interactive Atlas
, Diabetes Interactive Atlases
, and Interactive Atlas of Heart Disease and Stroke
were developed using Adobe Flash to provide animated mapping capabilities and all of them support play/pause functionalities. Particularly, users of Interactive Atlas of Heart Disease and Stroke
also can manipulate the pace of playing animated maps.
The majority of WGPHSSs use only choropleth maps; however, other types of thematic maps, such as graduated symbol maps, isopleth maps, and cartograms can provide an abundance of additional information. Unlike choropleth maps, graduated symbol maps use symbols in proportional sizes or graduated colors (e.g., in HealthMap
) to better express a health phenomenon’s severity. Isopleth maps have a much greater ability to reveal spatial patterns, as the health phenomena being mapped are not restricted to arbitrarily defined region boundaries. Moreover, isopleth maps represent health phenomena with a continuous surface, and therefore can provide local details that cannot be expressed with choropleth maps [88
]. A final alternative, cartograms, can convey information on a specific health indicator by distorting geometry or space proportionally. This helps highlight trends to the user, especially extreme values. Examples of using graduated maps, isopleth maps, and cartograms to represent health phenomena can be found on the websites [89
]. Similarly, beyond the traditional line, bar, or pie charts, unconventional chart types, including cartographic charts (e.g., pyramid charts, which have been used in the Pennsylvania Cancer Atlas
]) and charts commonly used in other domains (e.g., candlestick and bubble charts), need to be implemented more frequently in WGPHSSs for better health information communication. Despite the usefulness of these additional maps and charts, an interpretation problem (especially for those not so commonly used, e.g., cartograms) may arise for end users without a cartography background. Thus, further research is needed to examine balancing cartographic map/chart use and interpretation improvement.
4.6. System Maintenance and Compatibility
System maintenance and compatibility of WGPHSSs are important elements that still pose significant challenges. Not all WGPHSSs documented in the articles could be explored online because some systems’ Uniform Resource Locators (URLs) were invalid or could not be found. Interestingly, Disease Surveillance On-Line—the WGSPHSS developed by the Public Health Agency of Canada, for modifiable diseases, cancer, and injury surveillance—could be located in July 2013, but not on 6 August 2013 (the day we revisited all WGPHSSs’ websites for the first time). This may indicate that system maintenance is an issue. Lack of further financial and human investment might be a barrier to the ongoing maintenance, as some WGPHSSs may no longer be employed after a health emergency is over. Theoretically, any WGPHSS can be modified or extended to monitor other health phenomena, so maintaining these systems at a minimum cost for future use would be invaluable. From time to time, systems may have their URLs changed. Hence, it should be ensured that redirecting users from the old websites to the new ones is supported.
Meanwhile, some systems must be operated with a specific Web browser (e.g., AEGIS Flu
is currently supported only in Mozilla Firefox; Breast Health Portal, Project Safety Net
, and EpiSPIDER
do not work optimally in Internet Explorer), leading to browser incompatibility issues. Different end users prefer using different Web browsers (e.g., Internet Explorer, Google Chrome, Opera, Apple Safari, Mozilla Firefox, etc.
), or their options may be dictated by the different operating systems they are using (e.g., Windows, Mac, Linux, etc.
). Therefore, browser incompatibility is highly problematic for end users, as they may be unable to properly view the systems. Realistically however, enabling WGPHHSs to be compatible with all Web browsers requires additional development efforts, which may be unfeasible. One compromise is to develop WGPHSSs to fit a Web browser such as Google Chrome that can be used on multiple operating systems. Using Rich Internet Application (RIA, usually refers to “Web applications that provide a rich and engaging user experience comparable to desktop applications” [93
]) technologies such as Adobe Flash/Flex, Microsoft Silverlight, and JavaFX to develop WGPHSSs is another solution for Web-browser and operating-system incompatibility issues. Uniquely, RIA requires installation of browser plug-ins before launching the application, which remains a debatable issue among Web application developers. Nevertheless, since almost all Web browsers support installing plug-ins, RIA-based WGPHSSs can be operated across different browsers. For example, WGPHSSs developed with Adobe Flash (e.g., EpiScanGIS
) can be operated in any Web browser on any operating system that supports Adobe Flash Player. However, loading RIA-based WGPHSSs demands higher network bandwidth than non-RIA ones, an issue requiring special attention in WGPHSS development.
4.7. Privacy Issues of Health Data
Health data privacy concerns remain prominent barriers to efficient data analyses and system accessibility. Previous research has demonstrated the possibility of re-identifying patients’ location information from low-resolution maps [94
]. Therefore, geo-referenced health data is always aggregately released at a small map scale (e.g., provincial, regional, county, or health unit level) over the Web, severely limiting their ability to outline local trends, and identify precise locations of areas of interest (e.g., disease clusters) and their corresponding characteristics (associated with the clusters). These privacy issues also lead to restricted WGPHSS access, thereby limiting the use of these systems in health planning. For example, AEGIS
is accessible only by public health professionals. These exclusive WGPHSSs are more comprehensive in terms of health data and analysis functionalities; and as such, the information they provide would be invaluable for evidence-based health planning. With the implications of the built environment on public health being heavily considered in planning processes, urban planners must be able to access the detailed health information contained in these WGPHSSs so they can make evidence-informed decisions. Thus, measures and policies allowing this access while safeguarding confidentiality and security need to be developed as soon as possible. The literature suggests that harmonizing privacy legislation with public health research demands, developing and providing toolsets, algorithms and guidelines for utilizing disaggregate health data, educating the community, and simplifying bureaucracies [95
] are potential approaches to balancing the protection of health data privacy and the provision of detailed health information.
4.8. Development Costs
The resource requirements for development and training are perhaps the greatest challenges in the global acceptance and use of WGPHSSs. Developing a WGPHSS can in fact be unaffordable, as it is a time, money, and human resource consuming process, involving very technical complications such as “data conversion, cartographic design, and system design” [96
]. Hence, WGPHSS development occurs in developed countries much more often than in developing ones. Findings from our review show that development costs can be reduced in the following three ways.
Collaborating within public health sectors, whether locally, regionally, nationally, or internationally, can reduce development costs. Besides enabling monitoring health phenomena at different spatial scales, this approach can also benefit the development and management of integrated systems by reducing duplicated efforts. The importance of collaboration was emphasized in previous research [4
] and recent studies have illustrated the adoption of this collaboration paradigm in WGPHSS development. For instance, the CDC developed Rapid Data Collector
, a Web-based system for sharing health information with stakeholders while ensuring that other governmental agencies could utilize the system [97
]. Likewise, Canada initiated the Canadian Integrated Public Health Surveillance (CIPHS)
], attempting to remove barriers to integrating health surveillance data from different levels and systems. In addition to reducing costs, collaboratively developing a WGPHSS by health sectors at various organizational levels has the potential to avoid the inaccuracy and inconsistency issues in data collected at different spatial levels, in different data formats, and from different sources [99
]. Another added benefit attributable to health sector collaboration is that PHS is no longer limited to arbitrarily-defined administrative regions. Given that diseases have no political boundaries [17
], this makes more sense, especially in the context of conducting surveillance and data analysis. Public health sectors in different countries have collaborated to develop WGPHSSs for cross-border use. For example, Gao et al.
] and Moreno-Sanchez et al.
] developed WGPHSSs to monitor diseases across the Canada-USA and USA-Mexico borders, respectively.
Interdisciplinary collaboration is also needed. The sources provided in the identified journal articles and the sheer volume of literature suggested that various disciplines, including (Web) GIS, spatial epidemiology or medical geography, information technology and Web engineering, and public health, have made efforts to develop WGPHSSs since 2000. In practice, it is almost impossible for a person to possess all the required knowledge to coordinate and develop a successful WGPHSS. The academic community can collaborate to help train capable personnel by providing cross-disciplinary courses, thus contributing to the development of WGPHSSs.
4.8.2. Reuse or Adapt Existing WGPHSSs
Theoretically, with slight changes, a WGPHSS can be reused or adapted to monitor any kind of health phenomena. Currently, there are a number of successful cases. For example, a host of WGPHSSs (Global Health Atlas
, and DengueNet
and AEGIS Flu
; Google Flu Trends
and Google Dengue Trends
) were developed with the same platform. Another instance is that the Hong Kong government developed an infectious-disease surveillance system by integrating health components into an existing GIS platform. Reusing or adapting existing WGPHSSs can significantly reduce the development costs, which is especially important for low-income countries where costs are prohibitive. This cost-effectiveness also makes monitoring overlooked health phenomena feasible. The WHO defines health as “a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity” [101
]. This definition highlights the necessity of monitoring the physical, behavioral, and mental health of the population. Our review, however, indicates that most WGPHSSs are concerned with monitoring infectious diseases. Therefore, determining how to reuse or adapt current WGPHSSs for monitoring ignored health phenomena is a pre-dominant issue that requires further research. Moreover, although some WGPHSSs have been reused or adapted for other surveillance purposes, evaluation is still necessary to determine how successfully the WGPHSS has been adapted to its new purpose.
4.8.3. Adopt Geomashup and Open-Source Models
], probably because using open-source platform or software demands more advanced programming knowledge compared with building geomashups [103
]. However, we believe this development model should be adopted more in the long run, as it maximizes the interoperability, reusability, and extensibility of the WGPHSSs in addition to greatly reducing licensing costs. Additional advantages of using open-source software or platform in health informatics were discussed in previous studies [16
]. Some projects, such as AEGIS
, have released their source code to the public, most likely due to the realization of the significant advantage of utilizing the open-source model. Likewise, BioCaster
has made its ontology and software resources (e.g., the rule engine that drives the text mining system) available for public access and feedback [22
Furthermore, both models can facilitate cooperation and collaborations between different sectors within or between nations at different spatial levels. For instance, information on health conditions and health determinants can be separately prepared by different sectors and provided as different Web services and then be mixed to form a geomashup-style WGPHSS. On the other hand, WGPHSSs developed with the same open-source platform or software can interoperate with each other if collaboratively monitoring cross-border health phenomena is needed.
Parallel with WGPHSS development, end user training is necessary, especially for enabling proper map customization and use of analysis functions in WGPHSSs, as not all WGPHSS users have a GIS, cartography, or statistics background. Lack of funding may impede the provision of classroom training; however, adding training materials into system websites can facilitate training, thereby making Web GIS more cost-effective than desktop GIS [8
]. This training method has already been implemented in several systems. For instance, the manual for EpiScanGIS
has been integrated into its website. Likewise, the CDC supplies an online training module for learning GIS. Other efforts have also been initiated to train end users over the Internet. For example, a Web portal named Geovisual Explication
(G-EX) has been developed to help public health professionals learn new geo-visualization tools and spatial analysis methods [104
In spite of its potential contributions, this review has limitations. One is that it is not exhaustively comprehensive. A few WGPHSSs were inaccessible or could not be found online and therefore could not be included as part of the analysis. For WGPHSSs monitoring infectious diseases, some may have been discarded after the disease outbreaks were brought under control. Moreover, we relied on the Google search engine only to locate WGPHSSs. As such, this may have generated incomplete or biased information. We attempted to mitigate this problem by searching for WGPHSSs from the bibliographic databases as well. Furthermore, we reviewed English-language journal articles and WGPHSSs (or WGPHSSs with English versions) only, therefore may have omitted relevant WGPHSSs in other languages.