The World Health Organization (WHO) estimates that 67% of the 56 million deaths that took place globally in 2012 were due to noncommunicable diseases (NCDs), in particular, cardiovascular diseases (42.6%), cancers (21.7%), chronic respiratory diseases (10.7%), and diabetes (4.0%) [1]. About 75% of all NCD deaths took place in low- and middle-income countries (LMICs) where nearly half of all deaths occurred in persons aged under 70 years [1]. The WHO projects a rise in the number of NCD deaths from 36 million in 2008 to 55 million by 2030 if effective steps are not taken to curb the epidemic [2]. Furthermore, deaths from NCDs for people younger than 70 years old is projected to increase from 10.8 million in 2010 to 15.4 million in 2050 [3]. Four key risk factors responsible for a majority of NCDs are tobacco use, unhealthy diet, sedentary lifestyle, and excessive use of alcohol—all behavioral and largely modifiable [4]. These factors, as well as loss of employment due to NCD-related disability, combined with the long duration and complexity of NCD treatment pose additional challenges to poverty reduction and sustainable development [5].

Key to global efforts to prevent and control NCDs is national surveillance. At the national level, surveillance has three primary components: exposures (behavioral, physiologic, metabolic, social determinants), outcomes (NCD-specific morbidity and mortality), and health system capacity and response (infrastructure, policies, plans, access to health care, and partnerships) [4]. The WHO STEP-wise approach to NCD risk factor surveillance (STEPS) provides a standardized framework and toolkit for monitoring national levels of major NCD risk factors (exposures) [6]. STEPS consists of three levels of data collection: questionnaires, physical measurements, and biochemical measurements [6]. Similar to other household surveys, the STEPS questionnaires are conducted through face-to-face interviews, thus requiring significant time and resource commitments.

We propose a research and development agenda for NCD risk factors and MPS. The eventual purpose of the proposed agenda is to help standardize operating procedures for MPS, which will allow for comparisons of NCD risk factors within and across sites, and over time. Within this overall purpose, we propose specific goals. First, to answer key questions on how to design and deliver robust MPS to collect NCD risk factors; second, to assess the comparative effectiveness and costs of MPS modalities (eg, IVR, CATI, SMS); and third, to determine key ELSI in the development, in conduct, analysis, and reporting of MPS in LMICs. Each is further described below with proposed research objectives. These goals and specific objectives are also summarized in Table 2.

Under the first goal (to answer key questions on how to design and deliver robust MPS on NCDs), one research objective is to assess the usability of different ways of delivering MPS (eg, voice calls, recorded voice surveys—commonly referred to as interactive voice response (IVR)—or by text message) and community perceptions and willingness to complete MPS. This will probably require qualitative methods such as key informant interviews, focus group discussions, and user-group testing. Participants’ perceptions and willingness to complete and ability to navigate MPS delivery modalities such as IVR, SMS, and CATI can be assessed in order to provide a user-designed MPS [24]. These formative efforts can help inform translation and wording of questions, and suggest ways to improve MPS response rates. Results can be used to improve MPS platform design and questionnaires.

A second objective is to assess the role of different incentive amounts in MPS. Quasi-experimental or randomized controlled trial designs can be used to assess the effect of incentives on standardized contact, response, completion, and refusal rates in a MPS [25]. Studies will need to assess how different monetary incentive amounts; timing of incentive delivery (eg, beginning of MPS vs end); and nature of incentive structure (eg, fixed amount vs lottery) affect MPS metrics. These studies will need specific sample size calculations and calculation of key survey metrics most likely stratified by key demographic characteristics. Important issues around sampling, weighting, and stratification also remain understudied for MPS under conditions of varying mobile phone ownership and/or access [26].

A third objective will be to assess the impact of MPS introductory messages. This will, for example, necessitate studies that assess the effect of MPS introductory message on contact, response, completion, refusal, and attrition rates. Research could further examine the introduction content (motivational vs informational), voice type (male vs female), and the introduction’s character (casual voice vs formal) on key survey metrics by demographic characteristics. Exploring different sampling frames of mobile phone numbers is a fourth objective to assess the effect of the sampling frame on representativeness, response rate, completion rate, and costs of MPS. For example, the benefit of using random digit dialing (RDD) compared with a list of preexisting phone numbers, for example, from mobile network operators or previous surveys has not been characterized. Finally, the impact of NCD-specific questions (in modules) on survey attrition needs to be studied. NCD modules in existing surveys are groups of common-themed risk factor questions, such as diet and alcohol, and can be delivered in different order (or randomized) to assess survey impact by question number and content (also tests cultural sensitivities) by key demographics. The research agenda under this first goal will help optimize the delivery of MPS.

Under the second goal (to assess the comparative effectiveness of MPS modalities), it is important to understand the role and impact of different channels of communication with phone owners. One objective is to assess the effect of MPS modalities (eg, CATI vs IVR) on key metrics and performance indicators, for example, by administering the same survey content to a list of participants using two or more modalities. The costs of modality setup, deployment, and maintenance also need to be assessed to perform intermodal cost-comparisons. A second objective is to compare national or subnational estimates of NCD risk factor indicators between MPS modalities—this could be done by comparing NCD risk factor distributions captured using different MPS modalities. Variability in response to different questions, modules and risk factor prevalence captured by multiple modalities can be ascertained.

A third objective is to document MPS intermodal reliability and response consistency for NCD risk factors. This could be done by allocating (or randomizing) participants into two or more arms (eg, IVR and CATI) to establish congruency in responses for the same questions. Participants in one arm could receive one modality followed by another modality a few days later. Crossover designs might also be possible for an assessment of response consistency, adjusted for the risk of priming after exposure to the prior modality. Differences in the reported prevalence of NCD risk factors can also be evaluated to quantify possible incongruences in responses between MPS modalities. These objectives under the second research goal will answer questions on comparative effectiveness assessment in MPS in LMIC settings, as well as questions as to which modality provides better information on key survey metrics.

Under the third goal (determine key ELSI) relevant issues need to be explored in the development, conduct, analysis, and reporting of MPS in LMICs. Objectives under this goal are to (1) determine key ethical issues encountered in the conduct of MPS; (2) identify common and preferred practices for obtaining individual consent or permission for MPS to ensure voluntary, informed data collection; (3) document commonly encountered regulatory complexities and challenges that arise when conducting MPS and how they have been addressed; and (4) identify key societal goals and values that are supportive of MPS and how these are balanced against other important interests (eg, expectations of privacy and confidentiality).

A systematic review of the literature is needed to collect the most common ELSI and collate them according to broad themes. Data collection with researchers, public health practitioners, programmers, policymakers, and other stakeholders of MPS may help define and assess the key ELSI, how they have been addressed in practice, and how well they are reflected in existing reports and theoretical frameworks identified by global stakeholders. Additional conceptual work may be required in this space to understand which ethics principles and values are being satisfied and which may be tested using MPS. Equally important is the critical examination of factors that influence whether, and to what extent, various ELSI apply across different types of mHealth and digital health activities. This research goal will provide crucial perspectives and data to inform the responsible implementation of MPS in, by and for LMICs. The three research goals and possible objectives described above comprise an initial proposed research agenda for MPS and NCD risk factor data collection with a special focus on LMICs (Table 2).

Implementing the proposed research agenda will require both resources and creativity, either as standalone activities or integrated into other agendas. As such, we introduce in this section, an opportunity for integrating this agenda into a stream of work around data for health and NCD. The Bloomberg Data for Health Initiative (BD4HI) is supported by Bloomberg Philanthropies and the Australian Government’s InnovationXchange to improve the health of populations through strengthening public health data [27]. It involves multiple partners working in several sites including at national and subnational levels (Table 3). The project has an arm focused on strengthening vital statistics including birth and death registration in several countries. It also has an NCD arm implementing both nationally representative WHO STEPS survey and MPS using different mHealth platforms to generate comparable data across countries. The MPS are being designed to use or closely approximate the standardized STEPS survey with questions on demographics, tobacco use, alcohol consumption, diet, physical activity, and history of blood pressure and diabetes diagnoses and medications [6]. Finally, the project also has an arm focused on strengthening data impact and use of data for policies in countries.

It is in the context of this project that we proposed to integrate an initial research agenda for MPS (Table 2). Key to success of our research agenda is an understanding that this research would seek to inform MPS activities at the national level by feeding results to implementing partners and governments through our in-country collaborating institutions. In order to make this a realistic program of work, we decided to (1) initially focus on a few (2-4) selected countries in either the African, Middle Eastern, or Asian regions; (2) select two MPS modalities only for testing—IVR and CATI for now; and (3) work only at the subnational level. These initial research efforts will help inform future national level rollout in other countries considering adopting use of MPS. Table 2 includes a summary of our project-specific research agenda based on these decisions; this agenda incorporates elements of each of the three goals described above.

Currently, there are over 7 billion mobile phone subscriptions around the globe [10]. The majority of subscriptions are in LMICs, which are undergoing demographic and epidemiologic transitions where NCDs are catching up to infectious diseases as leading causes of death. To address this challenge, countries need cost-effective, timely, and systematic means of conducting surveillance and tracking progress toward globally agreed targets for NCD prevention and control [6,28]. Mobile phones present an opportunity to conduct rapid risk factor surveillance that can be used to inform resource allocation, as well as evaluate NCD prevention and control efforts and policies [29]. Addressing unresolved questions about maximizing key metrics for MPS, comparative effectiveness of MPS, and ELSI in the development, conduct, analysis, and reporting of MPS, would aid collection of timely and quality information about NCD risk factors [17,18]. The eventual goal of the proposed research agenda is to help standardize operating procedures for MPS, which will allow for comparisons of NCD risk factors within and across sites, and over time.

Despite the many advantages they present, it is important to note that NCD risk factor MPS probably would not obviate the need for in-person data collection. Physical measurements such as blood pressure, weight or height, and waist circumference, or obtaining blood samples for analyses (eg, for lipids, glucose), need to be conducted in-person. The MPS can complement an NCD surveillance system that includes a STEPS or similar survey done less frequently, which together can inform prevention and policy decisions [30]. The surveys, of course, have to be coupled with other aspects of the WHO surveillance framework including outcomes (ie, disease rates, deaths), health system capacity, and response (ie, infrastructure, policies or plans, access to care or medicines, partnerships) [4].

The BD4HI offers a unique opportunity to test MPS technology and delivery options in real world settings, while enhancing our understanding of the norms and values that are likely to influence global implementation. Proof of concept on the use of this technology for NCD risk factor data collection could help countries to complement existing mechanisms for monitoring progress in their interventions toward achieving commitments for NCD global targets. We welcome the global NCD and mHealth communities to provide further input on the proposed research agenda and hope it will inform research projects in the coming years.