Results of the study selection process
A total of 1289 publications were retrieved for this review, including 1189 from the databases and 100 from Google search. After deleting duplicates, 1013 remained for screening the titles, of which 301 were selected for screening of abstracts. After reading abstracts, 82 full texts were considered potentially relevant, of which 32 met our criteria. In addition, 4 papers were identified from reference lists. As a result, 36 studies were included for the data extraction (see PRISMA flow diagram in Fig. 1). A complete list of all studies can be found in Additional file 1.
Descriptive results of geographic distribution and study designs
The studies included came from twenty different countries (see Fig. 2 Geographic distribution of mobile phone-based studies). Eighteen studies were conducted in the American region (Colombia, United States, Brazil, Guatemala, Peru and Mexico), of which one study was conducted in four countries (El Salvador, Honduras, Dominican Republic and Guatemala). Twelve were conducted in the Asia region (Nepal, Singapore, Sri. Lanka, India, China, Malaysia and Pakistan), four in the Africa region (Kenya and Madagascar) and only two were identified in other regions (Fiji and Spain). Most studies were focussing on urban areas where our target diseases are prevalent, only three were specifically conducted in a rural area. Brazil and United States were the countries with the highest number of publications (each one with six), however the studies identified in the United States were not performed under real-world conditions, but rather under controlled conditions, laboratory facilities in particular. Most studies were published in the last three years (n=22), reflecting a recent increase in the use of mobile phones for the prevention and control of arbovirus diseases.
From 36 identified studies, most of them had a descriptive approach (n = 12), of which some provided preliminary results with small groups of people who "tested" the mobile technology in controlled environments and a few studies described their lessons learned after being conducted at a large scale. Some studies included pilot/feasibility studies (n = 6), diagnostic test studies (n = 6), retrospective studies (n = 4), cross sectional studies (n = 4), randomized controlled trials (n = 3), quasi-experimental studies (n = 2), non-randomized control trial (n = 1), and a qualitative study (n = 1). Regarding our target diseases, the majority of the 36 studies focused on dengue (n = 15), six studies on Zika and no study on chikungunya specifically. However, seven publications covered arboviral diseases in general or Aedes vectors. Seven studies on the mobile health technology targeted more than one infectious disease including arboviruses.
Mobile phone services
With respect to the mobile phone technology, we classified each service of a cell phone to identify which type of mobile phone category were most frequently used in terms of our outcomes. Five mobile phone categories were identified: mobile applications (mobile apps, smartphone apps, mobile software), SMS (Short Message Services), mobile phone tracking data (call detail records, mobile phone signals), camera phone (camera module/image sensor) and simple communication service (calls). An overview is presented in Table 3. The most widely used mobile phone category was mobile applications (n = 18). Simple mobile communication (e.g. voice communication) were used less often.
Purpose of mobile phone use in health programmes
To analyse the support that mobile phones are promoting, we noticed that the included studies in this review were focused on four major purposes: surveillance, prevention, diagnosis, and management which are summarized in table 4. Three studies were identified for both purposes: surveillance and prevention [55, 59, 66], thus those were assigned for both purposes in table 4, resulting in 39 studies. This review also identified specific aims in each purpose which are presented in table 4. Some mobile applications were able to perform more than one aim in surveillance such as data collection, taking mosquitoes photos, geolocation, among others (e.g. The App, Mosquito Alert) .
In total, the mobile phone-based studies included 25 for surveillance, 7 for disease prevention, 6 for diagnosis and 1 for management (communication). The mobile phone technology, mainly taking advantage of mobile applications, has been most frequently used for multiple aims in surveillance, followed by prevention and diagnosis. The use of mobile phone numbers focussing on communication between health staff and patients was less explored. Short message services were used for surveillance (data collection and reporting) as well as disease prevention (health education and promoting behavioural change). Camera phones coupled to diagnostic platforms and/or assays were aimed at diagnosis of arboviruses and identification of mosquito species.
Among the included studies, we assessed the different target groups or users of the mobile phone technology. Health workers were the main target group for receiving mobile phone services (n = 12). This group consisted of vector control staff, healthcare workers, physicians, practitioners, health managers and other health specialists. The second most frequent group were researchers (n=11) who conducted studies that used mobile phone tracking data or designed platforms with smartphone cameras under controlled settings. The third most frequent group was the general public (n = 9), which includes communities and specific population groups (students, athletes, police officers). Three mobile phone interventions targeted both groups, general public and health workers. Only one mobile phone service was designed for patients.
This review assessed the following outcome dimensions: performance, acceptance, feasibility, usability, costs and effectiveness. A description is given in table 5 summarizing the scope of expected outcomes in the 36 studies. Although, the description was developed following prior definitions [48,49]; some adjustments were developed deductively from the included articles.
The analysis of outcome dimensions (table 6) showed that a large number of studies assessed the performance of their mobile phone services (52%), particularly mobile applications, followed by studies that assessed feasibility (30%). It can be seen that few studies have provided information on acceptance, usability, and effectiveness. Costs analysis or at least estimated prices by mobile phone services were the least explored in this review. Mobile applications were the only service that assessed all outcome dimensions. Usability was only described by mobile apps-based studies. Table 6 summarizes the number of mobile phone services dealing with one or more outcome measurements.
A variety of operational characteristics were assessed in performance studies. Mobile applications and simple voice communications (calls) reported improvements in terms of completeness, for example, reporting more houses where vector control activities were conducted [53, 56, 88]. Familiarity of health workers with the application and using well known apps (Google maps) and geographic information systems (GIS) helped in locating more houses in real-time. It was also demonstrated that mobile applications were more useful in ensuring data quality and timeliness rather than traditional capturing methods. For instance, Chaak app reported a 19% reduction in the time spent per survey, along with fewer errors in data transfer in comparison with the pen-and-paper data capturing methods . The use of different modes of data transmission from mobile phones to the central server (transference with or without internet), good storing capacity of mobile phones, design of the app (white background and black lettering for better visibility), easy navigation (use of predefined terms, radio buttons and buttons in data entry fields instead of free text input) and trained health workers favoured the good performance of this mobile phone service . Mobile apps also showed good agreement (concordance) between syndromic data reported by participants and by nurses during home visits . Question algorithms with simple terminology and visual aids were key elements to facilitate the self-reporting. The use of smartphones has also led to the development of innovations to identify mosquito species using the acoustic sensor of mobile phones. For example, the Abuzz application was capable of sensitively identifying mosquito species at 10 to 50 mm distance, including Aedes aegypti .
SMS also demonstrated good performance in terms of completeness. Two studies conducted in Madagascar achieved to transmit more than 70% of patient’s data within 24 hours [72,75]. However, timeliness and data quality were yet an issue depending on the surveillance procedure and capacities of health workers to use SMS. Lack of guidelines and trainings, high workload, and technical problems (e.g. poor telecommunication network) were the main challenges reported by health staff .
Mobile phones have been used for tracking users through mobile phone data based on the Signalling System 7 (SS7) and Call Records Details in combination with different datasets (e.g. epidemiological data, environmental data). This mobile phone service showed a strong performance in terms of predictive values, identifying areas with high transmission risk of dengue . Its use has also allowed the integration of mobility models to predict the spread of disease epidemics .
Recently, smartphone camera-based diagnostic platforms have been explored to acquire images or read assays such as ELISA tests , RT-LAMP reactions [78, 80, 81], RT-PCR and RT-RPA tests . They have demonstrated high accuracy in terms of sensitivity and specificity (range between 95% to 100%) as well as a rapid detection of arbovirus (range between 10 to 20 minutes) [78-81,83]. Using a mobile application is an enabler for processing data and interpreting various tests in these diagnostic platforms. For example, Thiha and Ibrahim (2015), developed an ELISA reader for point-of-care dengue detection using the smartphone camera and mobile app. As a result, high performance was demonstrated, with 95% sensitivity and 100% specificity for dengue detection in comparison with standard ELISA microplate readers . However, these prototypes of smartphone-based diagnostic platforms could require qualified personnel to take biological samples and further studies to validate its performance and impact in a real working environment (patient's home or clinic).
Mobile apps interventions have been shown to achieve their aims under real conditions. They were particularly used for collecting and transferring entomological information to assess the transmission risk of arboviral diseases. For example, the entomological data (collected by Vectos app, OruxMaps, AutoNavi Navigation and Baidu Map) were analysed in a web platform or central server that successfully identified the level of vector infestation (larval indices) as well as the most abundant breeding sites [53,65]. Moreover, mobile phones together with traditional methods (ovitraps) and GIS technologies were able to track and monitor mosquitoes, identifying the index of female Aedes aegypti .
Mobile applications have also proved to be feasible for early detection of arboviral disease based on participatory surveillance which engages users directly in reporting and monitoring of symptoms [55, 57, 63]. This approach required medical staff and scientific experts to validate data reported by users and checked their health status during the intervention. Mobile applications were accompanied by a web-based application or desktop software to facilitate the data management in real time.
Mobile phone tracking data through CDRs and mobile phone signals have also demonstrated to be a feasible service for dengue surveillance in Asia region [84, 86, 87]. This service -when integrated with multiple datasets- has the potential to estimate human mobility in order to predict the spread of arbovirus diseases and outbreaks. Many mobile phones numbers are required to have a better representation of population.
Mobile apps were generally well received in studies conducted in India, Fiji and Guatemala [57, 66, 67]. General public were the main target group who assessed the acceptance of this mobile phone service. User’s satisfaction with mobile interventions offered was commonly based on how they felt using the app, whether they found it helpful or useful, and whether they would recommend it to others. Although high user satisfaction was reported in most mobile applications, its results depended more on connectivity to the internet and availability of mobile phone in households. For example, a study in Fiji showed positive feedback on user satisfaction in areas with good internet connectivity . Moreover, low socio-economic level of population might be related to people who did not accept the mobile phone intervention .
SMS interventions were highly acceptable for the prevention and surveillance of arbovirus. Their acceptability was assessed on how much participants enjoyed the service and whether they perceived it an informative and trustworthy strategy . Another study also checked if their health workers could easily use the service . As result, SMS showed to be a user-friendly service. The participation of stakeholders was useful to promote SMS as media for the prevention of dengue and facilitate its acceptance amongst the community.
Most mobile applications showed a good proportion of active users out of all participants who downloaded the app [55, 63, 67], but some researchers recommended more incentives, educational campaigns/trainings and constant communication with study/health personnel to keep to users motivated [57, 59, 63]. Some concerns related to additional expenses of mobile technology (e.g. mobile data plan), mobile phone features (less storage space, slow internet connection), lack of interest and knowledge regarding purpose of mobile phone intervention were associated with a proportion of users who did not use it . Fear and mistrust of adopting a new technology were other reasons for low usability in health workers . In addition, external factors such as period of high staff turnover, cellular tower collapse and socio-politic events caused the decreased use of mobile apps .
Cost calculations were done in different ways. One study described the market cost of a mobile device , another presented estimations of the mobile phone network including calculations of staff salary , another estimated the costs of coverage of mobile service during its implementation  and another analysed cost-effectiveness for the whole intervention, identifying cost savings .
Most studies on costs compared their mobile phone intervention with standards methods for vector surveillance. For example, Mosquito Alert app based on citizen-science initiatives demonstrated a reduction in the cost of coverage in comparation with ovitraps (Mosquito alert costed 1.23 Euros per km2 per month while ovitraps costed about 9.36 Euros per km2 per month). Vector surveillance with ovitraps required much effort to be installed and checked by qualified staff, while mobile application was mostly associated with community buildings and non-recurring investments in technology . Similar economic benefits were briefly mentioned by Bhadra et al. (2018) . However, Chaak app, reported costs equal or slightly higher than traditional capturing methods . Their costs were an issue associated with the type mobile phone network (cost per household were U.S.$0.10 for the pen-and-paper method compared with a cheap mobile phone plan U.S.$0.10 or an expensive mobile phone plan U.S.$2.13 for Chaak app). Additionally, a software developer or a person with technological skills could be required to manage the central server, adding costs to the mobile phone intervention. On the other hand, one study analysed cost-effectiveness of the MI-Dengue system using multivariate models to estimate the median cost savings per case prevented which was median $58 . This system based on the concept that vector control strategies should be applied in targeted areas with higher densities of gravid female mosquitoes, showed a better allocation of resources, saving hundreds of thousands of dollars in direct costs (health care and vector control) as well as lost wages . The cost analysis of this system not only included estimations on mobile phone technology but also costs associated with vector control inspections and other technologies (e.g. computers).
For the diagnosis of arboviral diseases, Chan et al., (2016) mentioned that smartphones are a more affordable alternative to collect fluorescent signals for point-of-care detection of arboviruses in comparison with other portable devices (ESEQuant Tubescanner) . However, information regarding the cost of these diagnostic platforms for point-of-care detection was limited.
Few studies showed effective m-health interventions in terms of reducing the vector densities through improved dengue prevention and behaviour change and/or performing as an early warning indicator for outbreaks. The analysis of effectiveness was based on well-defined methodologies (randomized controlled trials or quasi-experimental designs); however some studies were conducted in specific setting with a short interventional period.
SMS-based studies were the only ones that reported effectiveness in term of improving knowledge and practices of arboviral disease. Preventive messages via mobile phone were able to produce positive changes in human behaviour improving dengue practices and consequently affecting vector densities in households. Dammert et al. (2014), showed that households exposed to repeated preventative messages in Peru reported an increase in the use of vector control methods (window screening and/or mosquito bed nets), and a reduction in the infestation level (e.g. vector water containers testing positive for dengue larvae was 1.44% in the exposed group with SMS vs 2.47% in non-exposed group) . Additionally, SMS with conventional education methods were able to bring a major effect in the prevention of arboviral diseases. In Nepal, SMS together with a prevention leaflet were sent to the community, which increased knowledge and practices of people towards dengue prevention . Availability of mobile phones in households and shared responsibility of the community and other companies were identified as enablers of SMS interventions. In contrast, limited network access in remote areas, reaching private network users and lack of knowledge concerning the purpose of using mobile phones were the main obstacles perceived in the implementation of this mobile phone service.
For surveillance, the use of SMS has also demonstrated to be effective for reporting immediately notifiable diseases . Likewise, mobile applications plus traps were effective for monitoring of Aedes Aegypti in real time [58, 60, 61]. Their integration with geographic information systems (GIS) enabled the development of early warning mechanisms. For example, GIS datasets obtained from mobile application provided early warning signals in low endemicity areas where traditional surveillance was limited . Positive results were also observed in MI-Dengue system using a website platform, a mobile device (plus mosquito traps) and vector control inspections. Researchers showed that, in Brazil, the system was able to identify high risk areas which were then targeted for vector control and consequently prevented 27,191 cases of dengue fever . Using both approaches together (standard surveillance methods and mobile apps) are effective as entomological surveillance instruments for decision-making in the control of Aedes mosquitoes and subsequent action.