Eligible studies were experimental or quasi-experimental study designs using quantitative, qualitative, or mixed methods that focused on the use of mHealth to address a health challenge with First Nations populations. For the purpose of this review, mHealth was defined as per the World Health Organization [3], as the delivery of health interventions via wireless devices (ie, mobile phone, smartphone, cell phone, tablet, personal digital assistant, or pocket personal computer) using wireless modalities (ie, app, short messaging service (SMS), multimedia messaging service (MMS), voice messages, email, and program or website designed for mobile or tablet use). Studies were excluded if they involved mHealth interventions delivered via websites designed for nonmobile devices, interventions with specific equipment (eg, telehealth systems), and intelligent assistive technology using cognitive devices, physiologic or environmental sensors, and tracking capabilities. Studies reporting at least 1 of the following outcomes were included: (1) use of the mHealth intervention (eg, clicks, SMS replies, and logins); (2) user perspectives of the mHealth intervention (eg, acceptability, barriers, or enablers to its use, functionality, and cultural responsiveness); and (3) clinical effectiveness (eg, impact on screening, clinic attendance, medication compliance, and knowledge). Eligible papers were full text, published in English and reported outcomes of studies conducted in Australia, Canada, New Zealand, and the United States. These countries were the focus of this review given the similarities in colonialist history and policies with intergenerational impacts for First Nations people [14,22]. Studies were included if participants from these 4 countries were either First Nations (or if part of a multicultural sample, First Nations results were specifically reported) or parents or service providers (either indigenous or nonindigenous) of First Nations people. Research protocols, editorials, abstracts, reviews, and case reports were excluded.

In October 2018, the following databases were searched systematically and without any limits to publication date: Cochrane Library, EMBASE, CINAHL, MEDLINE via PubMed, Scopus, Web of Science, and PsycINFO. A comprehensive list of search terms and strings for the key themes of mHealth and First Nations or Indigenous was developed with librarian assistance. Search strings included proximity operators, truncation, and phrase searching to explore possible iterations of both themes. To ensure capture of all relevant materials within the Australian Indigenous context, 3 additional Australia-specific searches were conducted in PubMed, CINAHL, and PsycINFO following recommendations of the Lowitja Institute [23]. A sample of search strings is available in Multimedia Appendix 1.

Selection of relevant scientific papers followed the PRISMA flowchart [21]. Citations and abstracts were exported, and duplicates were removed. Screening of both the title and abstract and then remaining full-text papers was conducted independently by 2 of 3 authors at each stage (GH, DL, and MN), with discussions to reach consensus. Reasons for exclusion of full-text papers were noted. Manual searches of reference lists were conducted to identify any further studies.

Data extracted from each full-text paper included the following: title, publication year, authors, country and region, conflicts of interest, database, study aims, intervention name, health challenge addressed, study design, setting, intervention focus, intervention modality, intervention content, control, key outcome measures and instruments, outcome assessment timing, outcome assessment modality, participants, recruitment method, response rate, loss to follow-up, and other notes. Intervention outcomes were noted under 3 main headings: use, user perspectives including perceived cultural responsiveness, and clinical effectiveness.

The extent to which First Nations expertise was present within or sought by research teams in the studies was assessed via data drawn from full-text papers using authorship lists or public author profiles online. Authors’ statements regarding their identity or heritage were noted. Where such statements were unavailable, First Nations authorship was noted as unknown. Affiliations with First Nations bodies were recorded from full-text papers and Web-based profiles where available. Study designs and methods were reviewed to identify the use of participatory design principles or involvement of First Nations stakeholders in the intervention development or study conduct.

Risk of bias was assessed for each study using the relevant Joanna Briggs Institute (JBI) Critical Appraisal Checklist. This checklist assesses the methodological quality of a study and the extent to which the possibility of bias has been addressed in study design, conduct, and analysis [24,25]. Two authors (GH and LC) independently rated the risk of a range of biases for each included study as yes, no, unclear, or not applicable and tallied the number of yes responses, with consensus achieved via discussion. To enable comparison across studies using checklists with different numbers of items (9 items for quasi-experimental studies and 13 items for randomized controlled trials), a percentage was calculated. Level of Evidence (LoE) for effectiveness was also determined via a JBI tool that provides a rank of study designs “based on the likely best available evidence” [25] (p. 4). Study designs were compared with the 5 overall levels (Level 1: Experimental through to Level 5: Expert Opinion and Bench Research), with the specific level selected from possible subsets.

A narrative approach was used for data synthesis because of the mix of study designs and research approaches. Data were summarized under 7 headings as follows: (1) study characteristics, (2) design and content of mHealth Interventions, (3) risk of bias, (4) First Nations expertise, (5) intervention use, (6) user perspectives, and (7) clinical effectiveness .

All 13 studies included in this review were published between 2005 and 2018 (Multimedia Appendix 2): 5 studies were randomized controlled trials, where control groups were waitlisted for intervention (n=1) [26], received usual care (n=1) [27], had limited contact via SMS (n=2) [28,29], or received a combination of SMS, a paper resource, and usual care (n=1) [30]; 8 studies were single-arm quasi-experimental designs with a mix of qualitative and quantitative outcomes. Most studies were conducted in Australia (n=9), others in New Zealand (n=2) and the United States (n=2). No eligible studies were identified from Canada. All mHealth interventions were carried out in community settings or a combination of community and clinic and hospital settings. Participant demographics across the studies varied owing to the nature of the interventions. Adolescents and younger adults (15-30 years) were participants in 3 interventions focused on sexual health [31], new fatherhood [32], and suicide [26], whereas middle to older adults (40-75 years) were participants in studies relating to colorectal screening [27] and heart failure patient support [33]. In total, 2 studies recruited parents of younger children for interventions around child health [28,34], while other studies sought a wider sample of age for the intervention [29,30] or did not state participants’ age (n=2) [35,36]; 8 studies included a higher proportion of female participants, while 2 studies had more males and 3 did not report gender. Most studies identified participants as either First Nations (n=6) [27,31-33,36,37] or service providers to First Nations clients, or parents and carers of First Nations children (n=4) [28,34,35,38]. In studies where participants were not solely First Nations (n=3), specific outcomes were reported for First Nations people [26,29,30].

A wide range of mHealth designs, content, and intensities were used in the included studies (Multimedia Appendix 2). A total of 5 studies addressed mental health and suicide, while single studies focused on postpartum blood glucose screening, smoking, heart failure, colorectal cancer screening, chronic otitis media, hazardous drinking, infant feeding, and sexual health. Interventions were aimed at risk reduction, prevention or screening (n=7) [26,27,29-32,34], treatment and disease management (n=3) [28,33,36], or targeted multiple disease phases (n=3) [35,37,38]. The most common intervention modality was SMS (n=5) [27,29-32], followed by mobile apps (n=4) [26,35,37,38], MMS and SMS (n=1) [28], and via tablet software (not an app) (n=1) [33]. Two interventions used a combination of SMS and mobile phone or tablet apps [34,36].

Using the JBI tool, LoE of the 13 studies were determined as either Level 1.c—Experimental, randomized controlled trial or Level 2.d—Quasi-experimental, Pretest—Posttest or retrospective control group study [25]. There was variability in the quality of the 5 randomized controlled trials (JBI Level 1.c), with a median quality score of 7.0 and a large range (6-11) out of 13 criteria to address risk of bias (Table 1). The 8 quasi-experimental studies (JBI Level 2.d) had a median quality score of 3.0 (range 2-6) out of 9 criteria for managing risk of bias. Once converted to percentages to allow comparison across all study designs, the risk of bias score varied across studies from 22% through to 85%. The randomized controlled trials met a higher number of quality criteria, and thus exhibited a higher mean percentage of 62%, whereas the quasi-experimental studies met fewer criteria with a mean percentage of 34%. The most common risks of bias identified in the randomized controlled trials were lack of clarity in allocation and or blinding protocols (n=5) [26-30] and absence of reliable outcome measures (n=3) [26,28,29]. Risk of bias commonly identified in the quasi-experimental studies were absence of control group (n=8) [31-38] and lack of clarity about treatment or care other than the intervention (n=7) [32-38].

A majority of the included studies (n=8) were subject to loss to follow-up. Studies acknowledged these modest (5%-12%) [26,27,29,33] or greater losses to follow-up (19%-49%) [28,30,31,36] and where appropriate, the losses to follow-up were largely accounted for through intention-to-treat analysis.

A notable level of First Nations expertise was embedded in the mHealth studies through authorship, consultation, and research methodology (Table 1). A variety of resources were used to identify First Nations authorship including the following: full-text research papers, author profiles, or other Web-based sources [39-50]. The majority of studies (n=8), included at least 1 author who identified as First Nations in the country in which the study was conducted [27,29,30,32,33,36-38]. A total of 3 research teams based in Australia [26,34] and the United States [31] had close affiliations with First Nations bodies, some of which served as a source for participant recruitment. First Nations expertise was also sought through participatory design principles in 5 research studies [28,31-33,35] or through key reference groups composed of First Nations stakeholders (n=5) [26,27,33,35,37].

Cultural expertise was used to ensure appropriate language, imagery, and literacy from health interventions. In 3 studies, the importance of images, design, and audio were highlighted to ensure authentic representation of First Nations identity and assist users who face literacy challenges [26,33,38]. For example, participants using a mental health app highlighted the relevance of imagery given the First Nations significance of art for knowledge transfer throughout generations [35]. Language was a central reference point in 8 studies with consideration of word content for literacy key words, cultural relevance, and local dialects [27-33,38]. Consultation with First Nations stakeholders regarding cultural relevancy helped tailor SMS content for young Aboriginal fathers in Australia [32] but did not result in local language use for SMS prompts for colorectal cancer screening in Alaska [27]. Content of mHealth interventions which drew on cultural assets such as family and relationships, and a strength-based approach to First Nations health were noted in 3 studies [29,35,38], including the AIMhi Stay Strong App [38].

Three studies reported participant use of the mHealth interventions with mixed results (Table 2) [26,32,36]. In a randomized controlled trial of a self-guided mental health app (iBobbly), 85% of participants with data available completed all 6 activities during the 6-week trial; however, use data was only available for 66% of trial participants [26]. A quasi-experimental study investigating the use of SMS to support young Aboriginal fathers in Australia also reported use by the majority of participants, with more than half clicking the link provided in 1 particular SMS (56%). The overall response rate to mood-tracking messages was 75% (106/141) [32]. In contrast, a quasi-experimental study [36] reported low response rates to clinic appointment prompts sent by SMS, phone call, email, and social media via mobile app. Although participants had nominated mobile phones as their preferred contact method, only 14% of contact attempts through the various mHealth modalities (including only 1 of 29 SMS) received a response.

A total of 6 studies reported participants’ views of acceptability, cultural safety, or relevance for mHealth to deliver health interventions (Table 3). In that, 1 study used phone messaging [28] and 5 studies involved apps (Table 3) [33-35,37,38]. User perspectives were reported using quantitative outcomes from surveys or interviews [28,33] and qualitative data from interviews and focus groups [34,35,37].

User perspectives on phone messaging: All participants of a randomized controlled trial reported SMS and MMS communications as culturally appropriate [28]. The majority (76%) indicated they would be happy to receive health messages in the future, although 3 of 9 participants preferred SMS clinic reminders to educational MMS [28].

User perspectives on tablet or phone apps: An intervention delivered via tablet software in a quasi-experimental study employed a survey with a 5-point Likert scale to assess aspects of navigation and usability. Participants responded to positively phrased statements with 1=strongly disagree to 5=strongly agree, and where statements were phrased negatively, scores were reversed [33]. The mean participant satisfaction score for the heart failure education program was calculated at 4.15/5 (n=5). Participants also answered open-ended questions relating to acceptance, comprehension, and impact of the app. User satisfaction scores and open-ended feedback indicated user comprehension, positive impact, and overall acceptance of the program and its modality, despite being the first occasion each participant had used a tablet [33].

The other 3 quasi-experimental studies reported qualitative data for user perspectives from interviews and focus groups. Mental health service providers for Aboriginal and Torres Strait Islander people expressed positive feedback for the AIMhi Stay Strong App through themes of acceptability, building relationships, and applicability [35]. Constructive feedback regarding barriers to practical use of the app were also recorded. Povey et al [37] gathered user perspective data on factors that would impact acceptability of the AIMhi Stay Strong and iBobbly apps. Focus group themes included characteristics of the person, the environment, and the app. Participants expressed enthusiasm for the progressive nature of the apps in supporting Australia’s First Peoples with mental health challenges. Local language was referred to as a key element for acceptability and transferability [35,37,38].

Qualitative data from parents who used an infant feeding app indicated the intervention was well received and culturally appropriate, despite no cultural tailoring [34]. In contrast, service providers (staff) felt the lack of cultural consideration could be a barrier to use. However, parent participants viewed the app as a viable delivery method for quality content. Parent participants also expressed their familiarity with mobile phone technology and a preference for the app versus a website. Some technology frustrations were experienced by participants; however, this did not inhibit their use of the app. Parents suggested embedding personal stories in the app to further connect them with program content and lift cultural relevance [34].

A total of 10 studies reported data relating to the impact of an mHealth intervention, although the outcomes varied significantly owing to the nature of interventions, study designs, and outcome measures (Table 4). Of the 10 studies, 6 were delivered using phone messaging (SMS and MMS), including 4 randomized controlled trials [27-30] and 2 quasi-experimental studies [31,36]. The remaining 4 studies involved apps, tested through randomized controlled trial (n=1) [26] or quasi-experimental designs (n=3) [33,34,38].

The ever-evolving nature and terminology of mobile health technology presents some challenges. To ensure consistent inclusion and exclusion of intervention studies, this paper adhered to the World Health Organization definition of mHealth [3], which includes text messaging, mobile apps, and other wireless modalities. Although this definition facilitates inclusion of a wide range of interventions, thus strengthening the review, it is acknowledged that some limits could have been set by the authors in an effort to manage the breadth of mHealth interventions. Publication bias may also have impacted the identification of mHealth studies with First Nations populations, whereby studies with inconclusive or negative findings may not have been published [60]. In addition, this review did not include unpublished (gray) literature, which may have yielded further studies examining mHealth interventions. Quality assessment was conducted using 2 JBI tools to allow risk of bias comparison across studies using differing designs, with 2 authors independently assessing each study to minimize error in the quality assessment. However, it is recognized that quality assessments may reflect limitations in reporting (eg, owing to publication space) rather than study conduct. Similarly, First Nations expertise was assessed using desktop research only. Information on authorship, key reference groups, participatory design, and affiliations was not always clear or available. The authors acknowledge that studies may have drawn on additional features of First Nations expertise.

This is the first systematic review to document the characteristics, use, user perspectives, and clinical effectiveness of mHealth from peer-review scientific studies, with unrestricted health topics, where participants are First Nations people, or parents or service providers to First Peoples from Australia, New Zealand, or the United States. Encouragingly, mHealth interventions have been conducted with considerable First Nations expertise, rather than mere consultation, to guide cultural usability and meaning for community. This is a meaningful finding given that First Nations knowledge is the most valuable asset available to address challenges of First Nations health and community [61]. It is a feasible conclusion that embedding First Nations expertise into the projects helped augment favorable reports and feedback for use and user perspectives in the mHealth interventions. Difficulty remains, however, in the ability to draw conclusive statements about clinical effectiveness given the heterogeneity of mHealth interventions with First Nations populations. Clinically significant outcomes have been reported for both messaging and app modalities. While messaging appears to have a more mature evidence base, apps are acceptable, and they are being used effectively with First Nations Peoples. Trials are underway to validate previous findings of significance and to explore the transferability of mHealth across the diversity of First Nations populations. Importantly, this ongoing research, particularly in the area of mental health and suicide prevention, holds potential for significant impact with First Nations communities.