Mobile health (mHealth) apps, defined by the World Health Organization as “medical and public health practice supported by mobile devices, such as mobile phones, patient monitoring devices, personal digital assistants, and other wireless devices” [1], have become increasingly relevant in the health arena since the introduction of smartphones in 2007 [2]. With >6 billion smartphone users, indicating a penetration rate of >78% by the end of year 2020 [3], the number of mHealth apps has been increasing exponentially over time, leading to >351,000 mHealth apps available in the market in 2021 [4]. The COVID-19 outbreak accelerated this pattern, and mHealth provided a valid opportunity to deliver care remotely [5-7].

In oncology, mHealth apps have shown to provide benefits to patients throughout the care pathway [8-10]. Cancer treatments are complex, and mHealth apps can help patients manage their therapy more effectively and efficiently [11] by enabling better collection of patient data, remote monitoring by clinicians, patient education, and user-friendly communication tools [12]. In addition, apps have been shown to increase medication adherence, leading to reduced adverse events and increased quality of life [13,14]. This is particularly helpful for patients undergoing oral anticancer treatments, often performed in outpatient settings, whose success relies heavily on patients’ treatment compliance [15]. Overall, mHealth apps have the potential to increase patient empowerment by enhancing self-efficacy and improving patient-physician interaction [16].

Not only do individual patients benefit from using mHealth solutions, but also the broader health care system. There is a growing interest in the uptake of mHealth solutions in clinical practice because they have the potential to offer more accessible and cost-effective health care solutions [17]. Compared with conventional in-person therapies, mHealth can reduce health care costs while maintaining the same treatment quality by allowing the patient to attend follow-up appointments remotely [18,19]. By reducing commuting to and from the hospital, mHealth also holds great promise in mitigating the environmental impact of health care delivery [20], as commented by a recent study that appraised the potential environmental impact as a distinguished outcome domain of mobile medical apps [21].

The potential of mHealth is also reflected at the policy level, with an increasing number of countries gradually adopting regulatory frameworks [22]. For instance, the mHealthBelgium framework allows systemized recognition of mHealth apps as a medical device [23] using 3 validation levels depending on the safety level and socioeconomic value. Apps labeled with a level 3 status can be refunded by the National Institute for Health and Disability Insurance [24]. Similarly, in Germany, patients can apply for reimbursement of an mHealth app as a part of their statutory health insurance scheme if it is certified under Digital Health Applications (DiGA) regulation [25]. As of May 2022, the only DiGa-certified mHealth app for cancer care is CANKADO PRO-React Onco, which provides digital support to patients undergoing cancer treatment by facilitating communication with physicians and promoting patient education and empowerment [26]. In France, although some apps already receive reimbursement (eg, MOOVCARE POUMON for lung cancer telemonitoring) [27,28], the government is working on an assessment framework similar to that of the German DiGa [29]. In England, the National Institute for Health and Care Excellence developed an evidence-based standards framework for digital health technologies (DHTs), which is intended to be used by both technology developers and decision makers to inform the evidence development plans of the technology developers and commissioning of DHTs from the decision makers [30]. In this context, the European Union has recently launched a task force with the mission of harmonizing the evaluation of digital medical devices [31].

Increased interest in mHealth in cancer care has been observed in the fast-growing number of scientific publications in the past few years. However, most studies have investigated the impact of mHealth apps on patient outcomes. For instance, recent literature reviews have assessed the effect of mHealth apps on pain management in patients with cancer [32-35]. Other studies have investigated the impact of mHealth apps on patients’ quality of life, satisfaction with care, and user acceptance. However, there is limited evidence on the impact of the uptake and use of mHealth apps within the clinical setting. DHTs, including mHealth, are challenged by the phenomenon of early dropouts and abandonment [36]. To date, the implementation of mHealth apps has been analyzed less extensively. Does mHealth guarantee time and monetary savings for both patients and health care providers? Are mHealth apps used beyond the controlled study settings? In this context, implementation science is defined as “applied research that aims to develop the critical evidence base that informs the effective, sustained, and embedded adoption of interventions by health systems and communities” [37]. Through an extensive set of validated frameworks, tools, and strategies, this study investigates barriers and enablers to implementation that, respectively halt or facilitate the actual uptake of clinically proven interventions.

Therefore, this study aimed to investigate the determinants of mHealth uptake using a theory-guided framework from implementation science, the Consolidated Framework for Implementation Research (CFIR). The CFIR was intended as a practical tool to map and interpret empirical evidence regarding factors (ie, barriers and facilitators) that could affect the implementation of mHealth in cancer care.

This review follows the updated methodological guidance for scoping reviews [38] and the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines [39]. Scoping reviews aim to identify the main concepts, theories, sources, and knowledge gaps regarding a given topic of interest. The study protocol has not been registered. The 22-item PRISMA-ScR checklist for scoping reviews is provided in Multimedia Appendix 1.

Web of Science, PubMed (MEDLINE) and ScienceDirect were consulted. The search was extended to the papers published from January 2017 to March 2022. A 5-year timeframe was deemed appropriate considering the sharp increase in the number of studies on the topic and the rapid obsolescence of previous studies. Additional relevant studies were identified by screening the bibliographies of other published reviews (snowballing).

The search strategy was defined jointly by the research team and ultimately built around 2 broad content areas, cancer and mHealth. The exact keyword string used was as follows: (cancer OR tumor OR tumour OR oncolog*) AND (mHealth OR “mobile health” OR phone OR smartphone OR app). The search was restricted to titles and abstracts in PubMed, and to titles, abstracts, and keywords in Web of Science and ScienceDirect.

RefWorks [40] was used to retrieve relevant information from articles that were later exported in Microsoft Excel form for articles screening and data extraction. All papers selected for full-text reading were handled by the bibliographic reference manager, Zotero [41].

Only empirical study designs describing the development, evaluation (including testing), and implementation of an mHealth intervention were included. Other study types, including literature reviews, meta-analyses, conference abstracts, and clinical guidelines, were excluded. Studies were included if they focused on mHealth apps used as support for ongoing cancer therapies or management of related adverse events. Typical app functionalities included, but not limited to, enhancing patient self-monitoring, self-efficacy, or education, as well as fostering patient-clinician communication. Conversely, studies assessing mHealth apps used in other phases of the care pathway (eg, screening, diagnosis, and palliative care) were excluded. mHealth apps exclusively delivering noncore ancillary services for patients with cancer (eg, mental health, physical activity, and smoking cessation) were also out of scope. As for the target mHealth users, only adult patients undergoing cancer treatment were considered, whereas studies on cancer survivors, pediatric populations, or other targets with risky conditions or behaviors (eg, comorbidities) were excluded. Finally, studies not published in English were excluded. A detailed illustration of the inclusion and exclusion criteria is provided in Textboxes 1 and 2.

After double-checking a sample with a second reviewer (VA), the researcher GG screened all retrieved articles based on title and abstract, whereas full-text reading was performed by GG and VA. Disagreements regarding the inclusion of a given article were resolved by a third researcher (RT). All researchers agreed on the final selection of the studies selected for data synthesis. Owing to the variety of included studies in terms of design, objectives, and sources of evidence, no assessment of the risk of bias or methodological quality was undertaken.

Data extraction was performed in a Microsoft Excel grid. The extracted data included a general overview of the studies (eg, publication country, study objective, design, and duration), information on study participants (eg, number of participants, age, cancer type and stage, and cancer treatment), information on mHealth apps (eg, use time, app name, and main functionalities), study outcomes, and related metrics. The taxonomy by Dodd et al [42] that classifies the outcomes in medical research, was used to cluster the apps in the selected studies based on the investigated outcomes. In addition, CFIR was used to guide data collection and analysis of enablers and barriers to mHealth implementation, as well as strategies to overcome them. CFIR encompasses 5 domains and 39 constructs associated with effective implementation [43]. CFIR acts as a practical guide for systematically assessing potential barriers and facilitators when implementing innovation. CFIR integrates perspectives from different stakeholders and settings without inferring assumptions or drawing conclusions about the mechanisms of implementation, which is well suited to the heterogeneous literature to be synthesized [44]. A comprehensive explanation of the CFIR variables is provided in Multimedia Appendix 2.

The results were summarized using mainly a narrative synthesis and organized into 2 major sections. First, an overview of the selected studies and underlying app functionalities was provided, including key statistics (eg, count and proportions) and summary characteristics when relevant. Evidence on barriers and enablers specific to mHealth implementation was then analyzed following the CFIR framework. We did not expect to find evidence on every CFIR subdomain in each selected study; therefore, data analysis was conceived as a synthesis of subsets of relevant, available observations.

A total of 6190 papers were identified through the search (2564 records from PubMed, 3626 from Web of Science, and 506 from ScienceDirect). After duplicate removal, 3915 records remained for screening based on the title and abstract. A final number of 91 studies were included for analysis. Figure 1 describes the PRISMA flowchart [36,37].

Of the 91 studies, 78 (86%) [8,9,12,16,45-118] were research articles, whereas 13 (14%) [119-131] were study protocols. From 2017 to 2022, the number of published articles increased steadily over time. Almost half of the studies (43/91, 47%) were published in Europe, with Sweden (9/91, 10%), the United Kingdom (7/91, 8%), and Germany (6/91, 7%) having the highest number of publications. Outside Europe, relevant studies on mHealth in oncology were conducted in the United States (18/91, 20%), China (9/91, 10%), and South Korea (6/91, 7%).

In terms of study designs, randomized controlled trials (RCTs), including secondary analyses of RCT data, were the most common (26/91, 29%), followed by mixed-methods studies (24/91, 26%), qualitative design studies (12/91, 13%), pilot studies (11/91, 12%), other non-RCTs (7/91, 8%), pre-post studies (3/91, 3%), quasi-experimental studies (3/91, 3%), and other study designs (5/91, 6%). The majority (52/91, 57%) were single-arm studies, whereas 43% (39/91) of the studies were comparative, with 2 or multiple arms. Most of the included studies had a prospective design (84/91, 92%), 3 were retrospective, and others were combined retrospective and prospective branches (4/91, 4%).

Owing to their heterogeneous nature, the selected articles had different study durations, ranging from 2 weeks for small-scale trials to up to 2 years for larger-scale RCTs. The median sample size of the study participants was 51, ranging from a minimum of 5 to a maximum of 4475 patients.

Multimedia Appendices 3 and 4 provide an overview of the descriptive statistics and detailed study characteristics in a tabular format.

The 91 studies included for analysis describe 73 mHealth apps, of which 29 (40%) were designed for supporting any cancer types [9,12,47,48,50,54,56-58,65-69,71,73,74,76,81-83,85-88,​90,91,93,105-109,111-113,116,117,121,122,125,128], followed by 17 (23%) on breast cancer [49,64,70,72,77,​78,80,95-98,103,114,115,118,124,126,129-131], 5 (7%) [52,53,63,99,101,102] on gastric and colon cancer types, 3 (4%) [75,79,89,119] on lung cancer, 3 (4%) on thyroid cancer [84,100,123] type, and 2 (3%) on hematological cancer types [9,55,56,94,121]. The remaining apps (15/73, 21%) [8,16,45,46,51,59-62,92,104,110,118,120,127] covered other types of cancer, such as pancreatic, bone marrow, prostate, brain, and gynecological cancers.

Many apps did not support a specific cancer treatment (23/73, 32%) [46,51,53,62,71-74,83,84,86,88,93,95,98,99,104,105,108,​112,117,118,121,124,126]. The most frequent treatment specifications were chemotherapy (15/73, 21%) [47,50,55,63,64,70,78,85,87,101,110,114-116,119,127,129,131], oral anticancer treatments (13/73, 18%) [9,48,52,​56-58,65,76,80,82,90,91,94,111,122,128], radiotherapy (3/73, 4%) [54,92,120], and others (8/73, 11%) [12,16,45,59,60,66-69,75,81,89,102,106,107,113,125], which included several treatment types, such as a combination of chemotherapy and radiotherapy. Nonpharmacological treatments include surgery (8/73, 11%) [8,49,61,79,96,109,123,130] and transplantation (3/73, 4%) [77,97,100,103].

mHealth users can be patients, clinicians, a broader pool of health care professionals (HCPs), or different combinations of users. Most commonly, apps are designed for both patients and clinicians (42/73, 58%) [12,16,48-55,58,59,61,63,66,​67,69,75,77,84-86,89,91-94,96,98,100,101,104-106,108,110,​112-116,119,120,122-125,127-129,131], who typically access different interfaces and functionalities (eg, self-reporting function for patients, web-based dashboards with overview of patient activity for the clinicians). Only 32% (23/73) apps [8,47,60,64,65,68,70-73,76,78-83,87,88,95,97,99,102,103,107,111,​121,126,130] were designed for exclusive patient use. This is the case for certain medication adherence apps that focus mainly on providing reminders to patients [65,76,79,80,82,99,111]. The remaining apps (8/73, 11%) [9,45,46,56,​57,62,74,90,109,118] had diverse combinations of end users with patients, clinicians, caregivers, and pharmacologists. The app functionalities are listed in Table 1.

The selected studies assessed mHealth impact using a wide range of outcome metrics analyzed using the taxonomy by Dodd et al [42]. Outcomes most recurrently fall under the Life impact area, with 73 outcomes in the Delivery of Care outcome domain [8,9,12,16,45,46,48-55,57-63,65,67,70,71,73-87,90-92,​94-104,107-116,119-122,128-131], 37 in Global quality of life [8,57,60,62,64,65,68,72,75,78,84,86,88-91,93,97,101,103,​105,106,112,114,115,119-121,123-131], 16 in Emotional functioning and well-being [66,70,77,83,89,90,93,106,​115,119,125-128,130,131], 8 in Physical functioning [75-77,89,101,102,105,130], and 7 in Social functioning [66,69,84,115,125,126,131]. Within this core area, recurring metrics were the acceptability, usability, and feasibility of mHealth apps, which could be assessed either using validated questionnaires, or qualitatively, through study-specific questionnaires or interviews. Specifically, feasibility was assessed in 41% (37/91) studies, usability in 40% (36/91) studies, and acceptability in 35% (32/91) studies.

As for the Physiological or clinical area, 12 outcomes are General outcomes [65,70,76,77,89,96,103,115,125,127,130,131] and 4 relate to Neoplasms: benign, malignant, and unspecified [4,77,101,112]. As for the Resource use area, outcomes fall under Hospital (n=10) [49,59,64,65,79,89,97,120,124,125], Societal burden (n=7) [65,77,115,119,127,128,131], and Economic (n=1) [126] domains. Adverse events related outcomes were recorded 9 times [49,53,56,57,60,78,93,108,112] and Mortality or survival [97] once. The outcome core areas and domains are summarized in Table 2.

The overarching aim of this study was to investigate the determinants of mHealth uptake to inform the translation efforts of mHealth interventions in routine care. Studies illustrating the development, evaluation, and implementation of mHealth apps for cancer patients were considered, and information on barriers and enablers of app uptake was extracted following the CFIR scheme.

Many facilitators of app implementation in clinical settings have been identified. The involvement of patients and HCPs in app development has frequently been observed. Codevelopment was presented as a way to include desired mHealth features in early design efforts, to prevent unnecessary shortcomings, and activate a sense of ownership. These findings corroborated the idea that users should be intimately involved in the identification, design, and conduct phases of research, and not just be targets for the dissemination of study results [136]. An iterative development approach was often mentioned, as it ensured extensive usability testing during the development process.

As for implementation barriers, gradual rollouts and subsequent app version releases could be perceived as burdensome. From the provider’s perspective, mHealth could be referred to as a source of extra workload for the clinical staff. Conversely, factors characterizing providers, such as organizational leaders and management, staff, and culture, which can influence their ability to adapt and successfully use an intervention, were not systematically observed. From the user’s perspective, the fear of poorer patient-clinician interactions (eg, through remote monitoring) can diminish the sense of trust in the organization, in line with what was observed in prior works [137]. Although references to the outer setting (eg, laws and guidelines) were reported, mHealth was presented more as a way to address new or existing patient needs than as a way to respond to a given external pressure.

Although some of the findings discussed above are supported in previous research [138], and more broadly in the implementation science literature applied to DHTs, mHealth-specific dimensions resonated in this analysis. The peculiarities of mHealth, including the iterative nature of the corresponding interventions, frequent user interactions, a nonlinear relationship between technology use, engagement, and outcomes, implications at the organizational level, and challenges associated with genericization, distinguish apps from other DHTs [139]. For instance, compared with medical devices, typically evaluated through comparative evidence, studies on mHealth are often single-arm, noncomparative. Implementation hurdles related to system interoperability, data management, and patient privacy could appear to be more intricate for mHealth. Although these factors are reflective of the implementation challenges of DHTs, the distinctive features of mHealth seem to exacerbate their complexity.

mHealth will become increasingly important. On one hand smartphones are becoming increasingly prevalent and provide augmented functionalities (eg, cameras to capture high definition images of body parts). In contrast, demographic and epidemiological trends report a boom in chronic conditions, whose needs can be addressed by mHealth. Digitalization of the health care sector is a key priority in the political agenda, as confirmed by the expected massive capital injection in response to the COVID-19 pandemic. With more than €750 billion (US $798.38 billion), the next-generation European Union fund will invest a relevant share in promoting digital health, further boosting the development of mHealth apps. Although a stronger financial commitment is advocated [131], even in contexts where governments are directing huge health care spending to mHealth (eg, German DiGA), reimbursement policies do not always translate into actual clinicians’ prescriptions and are not a guarantee for users’ uptake [140]. Therefore, there is a need to adopt assessment frameworks for DHTs, including mHealth apps. Guidance on how to operationalize later implementation efforts is strongly advocated to avoid investing in technologies that are likely to be abandoned.

To our knowledge, this is the first review of the literature that uses a theory-guided framework to explore the determinants of mHealth implementation using a comprehensive approach in the area of cancer care. Other syntheses of primary studies mostly investigate the distinguishing features of mHealth [141,142] or their effectiveness in improving patient outcomes [143,144]. Studies illustrating the implementation initiatives in the area of mHealth are still limited and mainly document individual case studies. Although the field of implementation science has been growing, there is still a need to expand the use of implementation research to contribute to more effective public health and clinical practices [136]. Evidence suggests that theory-informed approaches to implementation science can enhance the translation and use of digital technologies in daily practice [145,146]. Under the lens of implementation science, Bardosh et al [138] conducted a qualitative evaluation of a single mHealth intervention addressing medication adherence and patient engagement. Heinsch et al [147] conducted a review of the theories that inform the implementation of eHealth interventions, and concluded that these are focused predominantly on predicting or explaining end user acceptance, and suggested that future research should test models that reflect the multidimensional, dynamic, and relational nature of the implementation process. Our work adds to the available literature by conducting a multidomain, multiple-stakeholder assessment of the determinants of mHealth implementation using the CFIR model. Rather than focusing more on a limited set of studies describing prevailing implementation research, our findings provide an integrated perspective on the factors that could influence the uptake and implementation of mHealth in clinical settings.

This study has several limitations. First, the papers selected for analysis were heterogeneous in terms of study characteristics (eg, purposes, study setting, design, duration, number, and types of participants). The decision to include a diverse range of studies was justified by the exploratory nature of scoping reviews [148] and stemmed from the observation that evidence from implementation research on mHealth solutions remains scarce. This was reflected in the search string, where implementation-related terms had to be. In addition, elaborating on implementation strategies, such as those described by the ERIC taxonomy Powell et al [149], seemed premature and was not performed. Although 29% (26/91) of the studies were RCTs, a proxy for evidence strength or quality, 24% (22/91) of the selected records had a sample size smaller than 20 patients, and 57% (52/91) were single-arm studies. Given this heterogeneity, a risk of bias assessment was not performed, although this is not unusual in scoping reviews [39,150]. Study heterogeneity also limits the possibility of performing meta-analyses on comparable outcomes. Finally, limiting the search to studies in English published since 2017 excluded a priori other potentially relevant earlier studies written in different languages.

This review sheds light on the determinants of mHealth uptake in clinical practice, exploring the barriers and enablers of the implementation of cancer care apps using an established implementation science framework. It contributes to filling the knowledge gap by systematizing the dimensions that should be factored into when designing an implementation strategy for mHealth apps.

Future studies should investigate whether and how specific dimensions such as app development and deployment platforms could affect implementation-related elements. In addition, a core set of outcomes associated with successful implementation, measured in studies that discuss implementation initiatives including hybrid designs, should be developed [151]. Finally, future studies should complement the organizational perspective from the current work with a patient-oriented (user) view and investigate the relationship between patient-reported measures and implementation outcomes. In this regard, technology adoption models such as the Technology Acceptance Model [152] or the Unified Theory of Acceptance and Use of Technology [153] could be relevant theoretical starting points.