Introduction

The rising disease burden from non-communicable diseases (NCDs), particularly in low- and middle-income countries (LMICs), constitutes a major threat to the health and well-being of communities worldwide. The “big four NCDs” (cardiovascular disease (CVD), diabetes, cancer, respiratory disease) account for 87 % of all NCD deaths and 54 % of NCD disability adjusted life years [1]. The United Nations (UN) high-level meeting on NCDs in 2011 demonstrated a historic commitment to NCD control, and in 2012, the World Health Assembly set in its 2013–2020 Global Action Plan a voluntary target of 25 % relative reduction in mortality due to NCDs by 2025 (“25 × 25 target”) [2]. Recent modeling suggests that a 25 % reduction in the prevalence of six NCD risk factors alone (tobacco, alcohol, salt, blood pressure, obesity, and glucose) could almost achieve the 25 × 25 target [3].

Although public health interventions will play a critical role in meeting this target, improved access to appropriate health care will also make a major contribution [4]. The World Health Organization Package of Essential NCD interventions (WHO-PEN) is a systems-oriented framework to tackling NCDs in low-resource primary health care settings [5]. It outlines (1) a conceptual framework for improving PHC equity and efficiency; (2) identification of core technologies, essential medicines, and risk prediction tools; (3) evidence-based protocols for implementation of a set of essential cost-effective NCD interventions; and (4) a technical and operational outline for integration into PHC and for monitoring and evaluation purposes [5]. Despite the potential for comprehensive primary health care (PHC) to deliver reductions in NCD burden, spiraling costs both to system planners and to consumers themselves is making health systems unsustainable to meet the growing NCD burden. Innovative strategies to implementing WHO-PEN are urgently needed in order to meet the 25 × 25 target.

Such strategies cannot merely duplicate the paths taken to improve health care quality in high-income countries, which are beset by rising costs, variation in care, and low uptake of evidence-based practices. LMICs are looking to “leap frog” some of the dilemmas experienced by high-income countries through novel health care delivery models that leverage low-cost, innovative technologies [6]. Just as mobile phones overcame barriers to communication caused by limited fixed line access, these technologies are now being applied to assist in health care delivery where access to traditional health care services is limited. mHealth is a multidimensional field encompassing a wide variety of tools, technologies, and models of health care delivery. Despite the bold promise of mHealth to be transformational, delivering high-quality care at a fraction of the cost incurred in high-income countries, there is a paucity of evidence to substantiate such a claim. mHealth is accused of being afflicted with “pilotitis” [7], in which piecemeal seed projects have been conducted with a lack of attention to scalability, poor integration into health care systems, sparse robust studies demonstrating effectiveness, and few evaluations of costs and benefits [8, 9].

Most reviews into the effectiveness of mHealth interventions have been dominated by studies conducted in high-income country settings. Although the number of studies included in these reviews have been few in number, improvements have been observed in health care service delivery processes [10], behavior change (particularly smoking cessation) [10, 11], and use of geographic information systems to support improved health care [12]. In LMIC settings, the mHealth literature is dominated by interventions in maternal and child health and sexual health, with a particular focus on use of mobile phones for data collection [13–15]. Braun et al. identified 25 studies exploring community health workers’ use of mobile technology [14]. Most studies were small scale and of the few that reported outcome evaluations, some demonstrated improvements in quality of care. Goel et al. conducted a narrative synthesis of 28 studies to examine the role of mHealth in bridging human resource gaps [15]. The authors found mHealth to be widely used in PHC settings for varying purposes including data collection, health surveillance, health education, supervision, and monitoring. Despite the breadth of use, there were little data on the impact of these interventions. Only one review has specifically looked at mHealth and NCDs in LMICs, and this focused only on the use of text and automated voice interventions [16]. Of the nine controlled studies analyzed, there were significant improvements in clinical outcomes (e.g., glycemic control for diabetics, lung function for asthmatics, and heart failure symptoms) and processes of care (e.g., attendance rates for follow-up appointments), and a limited number of studies showed improvements in costs and quality of life measures.

Although these reviews have looked at specific components of mobile health interventions, they have not examined the full breadth of mobile interventions as health care system strengthening tools. In this review, we take a systems-oriented approach to critically appraising the role of mHealth in improving health care quality for NCDs in LMICs. Specific aims are to the following: (1) characterize the spectrum of mobile health interventions that have been used for NCD management and prevention in LMICs, (2) evaluate the impact of mobile health interventions on health care quality, and (3) identify gaps in knowledge around mHealth research that need to be addressed.

Methods

Database Search

A systematic search of the literature was performed current to May 2014 using the following electronic databases: PubMed, PsychInfo, EMBASE, CINAHL, Cochrane, and the Latin American and Caribbean Health Science Literature Database (LILACS). A gray literature search was also conducted examining articles and websites from relevant organizations including WHO, International Telecommunications Union, the m-Health Alliance (mhealthalliance.org), HealthUnbound (healthunbound.org), mHealthKnowledge (mhealthknowledge.org), Global mHealth Initiative (jhumhealth.org), and Google and Google Scholar searches. We also searched for registered trial protocols in the WHO International Clinical Trials Registry Platform which includes 15 approved trial registries and supplementary searches in Clinicaltrials.gov. Keywords used in these searches included the following: cellular phone, mobile phone, telecommunication, mHealth, telehealth, telemedicine, patient education, point of care system, medical registries, electronic health records, clinical decision support system, data collection, provider-provider communication, provider scheduling, provider training, human resource management, supply chain management, financial transactions, primary prevention, secondary prevention, developing countries, underserved areas, and all of the LMIC names. Details of the search can be found in supplementary Tables S1, S2, S3, S4, S5, S6, and S7.

Inclusion/Exclusion Criteria

We included articles on any mobile technology health care interventions used in LMICs that were relevant to NCD management and prevention. LMICs were defined based on World Bank criteria [17]. NCDs included CVD, respiratory disease, cancer, diabetes (“the big four”), and mental health. Articles were included if they were (1) randomized controlled trials (RCTs), (2) quasi-experimental empirical studies with or without a comparator group, (3) descriptive studies without any outcome measures reported, (4) reviews of mHealth interventions (systematic or non-systematic), or (5) registered RCT protocols. There was no language exclusion to the articles retrieved. Telehealth, telemonitoring, and telephone coaching studies were only included if they explicitly drew on mobile technologies as part of the overall intervention strategy. If these interventions were delivered via a standard fixed phone line or via the internet using a desktop computer, they were excluded.

Classification Framework

The mHealth interventions used in each study were characterized using a framework proposed by Labrique et al. in an analysis of maternal and child health mHealth interventions (Table 1) [18]. A key strength of this framework is its focus on health systems rather than specific technologies. The framework was developed in consultation with mHealth stakeholders including academics and program and policy implementers. It was then applied to illustrate where mHealth opportunities and health system constraints lie across a continuum of care for maternal and child health. It is useful in determining who might be the beneficiary targets of particular mHealth strategies and making explicit the particular health system barriers that are being targeted.

Table 1 Common mHealth applications
Full size table

Outcome measures were assessed according to the WHO dimensions of quality of care (Table 2) [19]. These domains serve as building blocks for identifying tools and strategies for quality improvement at the level of policy makers, service providers, and consumers across whole health systems. Assessing outcomes in these particular domains allowed us to identify where the gaps were in the evidence base for mHealth interventions across the whole health system.

Table 2 WHO quality outcome dimensions
Full size table

Data Extraction

Two reviewers independently evaluated and excluded articles at the title/abstract review stage. Full-text articles whose abstracts met the inclusion criteria were then reviewed. An Excel template was developed which outlined study characteristics, the mHealth domains, and WHO quality outcome domains. The reviewers performed test data extractions using this template to check for any inconsistency in interpretation of definitions, and the extraction template was refined following this. If the article met the final inclusion criteria, reviewers populated the data extraction template. For the RCTs, methodological quality was assessed using the Cochrane Risk of Bias Assessment Tool. Discrepancies in article inclusion, data extraction, and bias assessment were solved by team consensus.

Results

We retrieved 1,569 articles using the search terms, and 177 articles were selected for full-text review (Fig. 1). Of these, 129 articles were excluded for the following reasons: not specifically using mobile technology, in particular internet and fixed line telemedicine interventions (n = 86); not pertaining to LMICs (n = 29); not relevant to NCDs (n = 12) and study protocols (n = 2) (see supplementary Table S8 for more details on excluded studies).

Fig. 1
figure 1

Included/excluded studies

Full size image

The 24 included non-protocol studies, their characteristics and the mHealth domains are summarized in Table 3. The majority of studies came from middle-income country settings with a mixture of urban- and rural-based studies. The most common disease areas were either diabetes (n = 8) [20–27] or CVD and risk factors for CVD (n = 9) [22, 28–35]. Thirteen studies tested specific mHealth interventions, but only seven used a RCT design [23, 25, 26, 33, 36–38], with the remainder using quasi-experimental designs to assess outcomes [24, 27, 28, 34, 39, 40]. Six exploratory studies were identified which described, validated, or pilot-tested various mHealth interventions but did not provide any substantive outcome data [29–32, 35–41]. Five reviews were also identified [16, 20–22, 42], of which two systematically appraised the literature [16, 21].

Table 3 Included studies
Full size table

Of the intervention and exploratory studies (n = 19), the following mHealth domains were identified: client education and behavior communication (n = 13) [23–27, 33, 34, 36–41], sensors and diagnostics (n = 5) [28–30, 32, 35], registries (n = 1) [34], data collection (n = 3) [28, 29, 34], electronic health records (n = 1) [28], decision support (n = 1) [31], provider communication (n = 3) [28, 29, 34], provider work-planning (n = 5) [28, 36–38, 40], and supply chain management (n = 1) [28]. There were no studies pertaining to provider training and education, human resource management, or financial transactions and incentives. Most studies tested only one or two mHealth domains (n = 16) with only three studies using multifaceted interventions involving three or more domains [28, 29, 34].

The outcomes of the 13 studies that did test particular interventions are shown in Table 4. Six studies reported effectiveness of the intervention for clinical outcomes (n = 6) [23–26, 33, 34], and three reported improvements in processes of care, particularly improved knowledge, attitudes, and behaviors such as medication adherence (n = 3) [23, 36, 39]. Three studies addressed costs [26, 33, 37], with one study reporting improvements in health-related quality of life [33]. Four studies reported improvements in clinical attendance rates [28, 36–38], and four studies reported various self-reported metrics related to acceptability of the intervention [24, 25, 27, 36]. No studies reported outcomes related to equity or safety, and similarly, no studies reported any qualitative or process evaluations of the interventions. For the RCTs, the majority of the risk of bias criteria was classified as either low or unclear (Fig. 2).

Fig. 2
figure 2

Risk of bias for randomized controlled trials

Full size image
Table 4 Outcome measures by WHO quality domains (intervention studies)
Full size table

The search of registered clinical trial protocols identified 24 additional mHealth RCT protocols for NCD prevention in LMICs (Table 5). Although eight of these studies were listed as being complete, we were unable to find published results for any of these. As with the published RCTs, the majority of studies focused on client education strategies using SMS systems (Table 5).

Table 5 RCT protocols registered in clinical trials databases
Full size table

Discussion

In this review, we examined the ability of mHealth interventions to improve health care quality in LMIC settings for NCD management and prevention. Specifically, we sought to identify which mHealth components have been associated with the greatest impact on health care quality dimensions. We build on previous reviews by updating searches in a rapidly evolving field, but more importantly, we examine how mHealth has been used to strengthen health care systems to address the growing NCD burden.

A number of key findings were observed from this review. The first and most important is that mHealth for NCD management remains a relatively under-explored area. The literature is characterized by a limited number of high-quality studies, mainly conducted in middle-income country settings and mainly focused on two NCDs—CVD and diabetes. Despite mHealth having a wide variety of applications, studies so far are dominated by behavior change interventions through use of text messaging systems. Few studies have applied mHealth tools as a means of strengthening health systems. Although the studies that have reported effectiveness are encouraging, few have examined outcomes across multiple dimensions of health care quality, and none have looked at equity and safety issues. Related to this, there is a dearth of process evaluations to understand the contextual factors that promote or hinder effectiveness of the interventions. Consequently, there remains a major gap in our understanding of the factors that may influence scalability, replication of outcomes in different settings, and sustainability of outcomes beyond controlled trial settings.

The paucity of literature suggests that mHealth for NCD management is still at an early stage of development. There are, however, several relevant registered studies, all of which are RCTs, which are actively recruiting. This suggests that the evidence base will grow substantially in the coming years. These registered studies generally have larger sample sizes, longer follow-up periods, and are using designs such as cluster and stepped-wedge cluster RCTs which are more conducive to understanding health system impacts. Unfortunately, most of these newer interventions remain narrowly focused on behavior change using text messaging systems. However, there are some notable exceptions which are taking a systems-oriented approach using mHealth domains such as decision support, electronic health records, and workforce-oriented strategies such as provider-provider communication. Several completed studies have not published results which raises concerns about publication bias. For high-income countries, the mHealth research landscape is certainly changing. In 2012, Labrique and colleagues identified 215 mHealth intervention studies registered in clinicatrials.gov with 176 involving an RCT design [7]. Publications arising from these trials will result in major shifts in the quantity and quality of research evidence becoming available and may cure mHealth of its chronic “pilotitis.” Although this will be useful, it may take up to a decade before there is more clarity on the role of mHealth in strengthening health systems in LMIC settings.

Limitations

There are a number of limitations to this review. Although the classification framework we used was very useful, many studies lacked sufficient detail to characterize them in finer detail. For example, there may be considerable variation in the design and delivery of SMS behavior change interventions, and hence, it is difficult to appreciate differences between interventions within any particular mHealth domain. Similarly, we had limited ability to analyze specific technical approaches used and were unable to make conclusions regarding the similarities and difference between platforms used. This is important because there is a clear need for standardized and approved architectures for mHealth tools. Recent US Food and Drug Administration guidance on regulatory requirements may facilitate this [43]. Another limitation was that owing to the paucity and heterogeneity of RCTs in the review, we were unable to conduct a quantitative meta-analysis of the outcomes. This may be addressed in coming years as more trial results in this field are published. Although we examined the leading contributors to NCD mortality, we did not examine other NCD areas such as musculoskeletal conditions which are a major contributor to disability. It is important that research on both disease-specific and non-disease primary health care strategies are conducted to enable a more nuanced understanding of both the disease management and systemic challenges that mHealth may be able to address.

Conclusion

On balance, despite the promising findings demonstrated by some mHealth interventions in this review, we conclude that the current evidence base is insufficient to guide decisions on policy and practice. There is a lack of research on end-to-end health care systems where multifaceted strategies are taken to improve patient care. Restricting mHealth to patient-level behavior change initiatives on its own will not be adequate to promote reductions in NCD burden in LMICs. Mechael et al. recommend that mHealth move from single-solution-focused approaches to become an integrator of health information across the entire continuum of care [8]. In particular, the development of mHealth tools to strengthen workforce capacity, communication, and workflows is of particular importance. This would lead to better alignment with WHO-PEN and support practical links between the use of mHealth tools and national and international policy frameworks. It should be noted, however, that we are not advocating for a departure from “grassroots” approaches to intervention development. Such approaches are critical in maximizing user-responsiveness and sensitivity to context, and when companioned with other measures to address system gaps, their likelihood of lasting success is increased.

We recommend four priority areas to improve the mHealth research agenda: (1) comparative effectiveness studies examining mHealth versus other “traditional” health care improvement strategies; (2) large, multinational studies powered on “hard” clinical endpoints such as mortality and hospitalizations that enable cross-country comparisons; and (3) process and economic evaluations of effective and failed interventions to determine contextual opportunities and constraints for scale-up. A fourth more complex research priority area is the need to examine policy-level barriers to large-scale adoption of promising mHealth interventions. Factors such as mobile network coverage, data governance and consumer rights, patient identifiers, inter-operability and standards, regulatory approval, medical advice liability, and sustainable business models tend not to be considered in traditional research studies and yet are of crucial relevance to delivery of promising interventions at scale. Greater engagement with policy makers in study design and implementation is needed to ensure that research does not occur in a policy vacuum and that interventions can be integrated with existing national and local initiatives. Similarly, greater engagement with the private health care sector, including insurance providers, is needed. Given private investors are likely to be major payers for mHealth systems and that a large proportion of health care in LMICs is sought privately, it is critical that research and business agendas are better aligned [6]. While there is every reason to be optimistic about the transformative power of mHealth to reduce NCDs in LMICs, there is much work to do to convert the rhetoric into reality.