A librarian (LO) collaboratively developed the highly sensitive medical subject headings (MeSH) term‒based search strategies with other review authors (SB, LK) and from July to September 2015 ran searches in the following databases: PubMed MEDLINE; Embase; Cochrane Central Register of Controlled Trials (CENTRAL) on the Wiley platform; the Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCO); PsycINFO (EBSCO); Web of Science; Center for Review and Dissemination; and Inspec (EBSCO). Additional searches were run in November 2015 using the following sources: Proquest Dissertations; Scopus; ClinicalTrials.gov; World Health Organization Clinical Trials; Controlled-Trials.org; Institute of Electrical and Electronics Engineers Explore; and Google Scholar. Search strategies for all databases except MEDLINE were adapted from the PubMed MEDLINE strategy. All databases were searched back to 1995, which is a point in time when access to mobile phones began to increase rapidly. No language limits were applied. The search strategy specified keywords, including text messaging, phones, mobile apps, and portable software combined with adherence or compliance, and search terms related to child, pediatric, adolescents, and youth. We also reviewed the search strategies of previous studies to include additional terms. See Multimedia Appendix 2 for complete search strategies in each database. An additional hand search of related themes in the Journal of Medical Internet Research was also conducted. We also attempted to identify additional studies by searching the reference lists of key studies and relevant systematic reviews. We contacted authors of included publications to obtain additional studies meeting the inclusion criteria.

The inclusion criteria were as follows: (1) adolescents (12-24 years old) [31] with a chronic illness requiring long-term daily or weekly medications for ≥12 months, (2) original research articles, (3) studies that were either randomized controlled trials, quasi-experimental studies, or pilot/feasibility studies (including single arm, pre-posttest), (4) text messaging or mobile phone‒based interventions (app or mobile intervention), and (5) medication adherence as the primary or secondary outcome. The exclusion criteria included (1) mean or median age of entire study cohort in the study was either <12 years old, >24 years old, or not specified, (2) adolescent participants were not the target of the intervention (eg, intervention targets babies born to adolescent mothers with human immunodeficiency virus or targets parents of adolescents), (3) text messaging and mobile phone apps as interventions focused on disease monitoring or ecological momentary assessment, but not meant to improve medication adherence, or (4) technology-based interventions other than text messaging and mobile phone apps, such as Web- or Internet-based interventions, personal digital assistant, etc.

We used a standardized form for data extraction. Data items in the extraction form included the following: first author’s name; publication year; country; condition or disease focus of the study; participants’ age; study design; duration of intervention and follow-up; components of text messaging or mobile phone app interventions; control group (if applicable); adherence measures; adherence rates; other outcome measures such as disease-related outcomes of morbidity and mortality, HRQOL, and self-efficacy or self-management skills; and theoretical framework. Two authors coded all included articles individually, and then the lead author (SB) independently reviewed all codes. Disagreements were resolved by discussion or by consultation with a third author, if needed.

Studies described in each article were evaluated for the quality of evidence using the GRADE approach (Grades of Recommendation, Assessment, Development, and Evaluation) [32]. This method evaluates four different key domains including consistency, directness, risk for bias, and precision of the evidence. Two authors graded all included articles individually, and then the senior author (LK) independently reviewed all grades. Disagreements were resolved by discussion or by consultation with a third author, if needed.

Data were analyzed quantitatively and qualitatively. Our primary outcome measure was mean change in medication adherence rate, and data were analyzed for those who had baseline and follow-up values. We also analyzed mean change in adherence-related laboratory markers. Standardized mean differences (SMD) with 95% confidence intervals were calculated—using means and standard deviations—to evaluate the efficacy of text messaging and mobile phone‒based interventions in improving subjective and objective measures of adherence, including adherence rates and adherence-related laboratory markers [33]. Data were analyzed using Stata 13.

A total of 1423 citations were retrieved; 1137 in the July-August 2015 search and 286 in the search in November 2015. After removing duplicates, 1118 original articles remained (see Figure 1). Two authors (SB and LK) independently screened the article titles and abstracts of 1118 records against inclusion criteria, and a total of 156 records met all predefined inclusion criteria. Two authors (SB and LK) then independently reviewed the full text of these articles in detail against the exclusion criteria, and 141 articles were excluded. A total of 15 articles met all predefined criteria to be included in this review [34-48]. We did not identify any non-English articles that met our inclusion criteria. The study flowchart and reasons for exclusion of full text papers were documented in an adapted PRISMA study flowchart (see Figure 1) [49].

Although our search included studies published since 1995, no eligible studies were identified before 2005, with most studies (12/15, 80%) published since 2010 (Table 1) [34-39,41-44,46,48]. Most studies were performed in the United States (in whole or in part; 11/15, 73%) [34,36-38,41,42,44-48]. Studies targeted eight different CHCs, including those of adolescents with acne (n=2) [34,39], asthma (n=1) [47], depression (n=1) [42], diabetes mellitus (n=4) [35,40,43,46], human immunodeficiency virus (n=2) [37,41], liver transplant (n=2) [44,45], sickle cell disease (n=2) [36,38], and systematic lupus erythematosus (n=1) [48]. Most studies were small in size (ie, ˂50 participants; 11/15, 73%) [34-37,42-48], and just over half (8/15, 53%) included samples of young adults (with a mean or median age ≥18 and ≤24 years old) [34,37,39,41-43,47,48]. In terms of study design, more than half were RCTs (8/15, 53%) [34,39-43,47,48], and the remainder were primarily single-arm nonrandomized trials (6/15, 40%) [35-37,44-46] or retrospective chart reviews (1/15, 7%) [38]. The duration of the studies varied: 2-4 weeks (2/15, 13%) [42,43], 12-16 weeks (5/15, 33%) [34,35,39,46,47], 24 weeks (3/15, 20%) [36,37,41], and 12 months or more (5/15, 33%) [38,40,44,45,48]. Only one study evaluated the durability of intervention effects in a crossover design with 6-month follow-up in one of the study arms after the intervention was discontinued [41]. Measures of medication adherence included self-report (9/15, 60%) or biomarkers (9/15, 60%), as well as other forms of monitoring (7/15, 47%), such as Medication Event Monitoring System caps, directly observed therapy, pill count, and pharmacy records abstraction. Three studies (20%) measured additional adherence behaviors, including completion of laboratory visits [44], clinic visits [48], and monitoring of peak expiratory flow values in patients with asthma [47]. Only four studies (27%) incorporated explicit theoretical approaches or frameworks into the model of intervention effects, including Theory of Planned Behavior [43], Gamification theory [35], and Social Cognitive Theory [40,41]. Most studies (12/15, 80%) were rated “low” or “moderate” according to the GRADE criteria (Table 1) [32], with low ratings primarily due to limitations in design as well as imprecision of results.

The majority of interventions used text messaging to promote medication adherence via reminders or motivational messages (12/15, 80%) [34,37-41,43-46,48]. Of these 12 studies, only one combined text messages with other in-person intervention components (educational sessions) [47]. Additional interventions used other mobile phone‒based approaches or apps (3/15, 20%), including multifunction mobile phone apps [35,42] and mobile phone‒based directly observed therapy [36].

Of the 15 studies reviewed, 7 (47%) demonstrated statistically significant differences in medication adherence or related health outcomes [36,37,41,43-45,47]. The majority of the studies included in this review provided enough information to calculate standardized outcomes, such as SMDs. Most studies reported overall moderate to large SMDs of subjective and objective markers of adherence; however, most SMDs had wide 95% confidence intervals (Table 4). Several studies combined data related to the assessment of the efficacy or the usability/feasibility of different interventions reporting high levels of satisfaction and few technical or feasibility problems [35-37,40,41,44-46]. Additional results of each individual study are summarized in Multimedia Appendix 3 for text message interventions, and Multimedia Appendix 4 for mobile phone app interventions.

Medication nonadherence is a widespread problem in pediatric CHCs, and among adolescents in particular. In this systematic review, we identified 15 studies that met all our inclusion criteria. Our results suggest that there is modest evidence to support the efficacy of text messaging and mobile phone apps as interventions to improve medication adherence in adolescents with CHCs. Most of the included studies were of low to moderate quality because of methodological limitations, imprecision of results, or both. The included studies showed evidence of the acceptability and feasibility of text messaging and mobile phone app interventions, suggesting a potentially promising area of intervention development and further investigations in the near future to better understand their efficacy and cost-effectiveness.

Our findings suggest moderate SMDs for most included interventions, which is consistent with earlier reports of adherence-enhancing interventions (ie, non-technology specific). However, given the heterogeneity of the included studies, the observed moderate effect size should be interpreted with caution [18,23]. In contrast, Pai and McGrady reported small effect sizes in a systematic review of adherence-promoting interventions [20]. In our review, the quality of the included studies was low to moderate, which was similar to a recent review of findings for technology-mediated interventions for treatment adherence across all age groups [19], and more specifically for adolescents with chronic physical conditions [28].

While text messaging and mobile phone app interventions offer a straightforward approach to address adherence to medications among adolescents, with broad application across CHCs, some challenges exist. None of the included studies measured the long-term durability of intervention effects across randomized conditions; thus, there is a need to establish the exposure or dosage needed to sustain long-term effects. In addition, the characteristics of the included studies also suggest that there is a need for improvement in intervention design; only four included theoretical models or approaches to target specific mechanisms of action to optimize efficacy [35,40,41,43]. McGrady and colleagues have recommended that articulation of the underlying mechanism of action for adolescent-specific adherence interventions is an advancement much needed to bring developmental and behavioral specificity to this growing area of research [50]. Furthermore, only two studies measured potential moderators of the intervention effect [41,43].

Several of the studies measured feasibility and acceptability outcomes and found high levels of satisfaction and few feasibility issues, which are promising for advancements in these technologies and consistent with earlier reports [19,21,28]. The high level of satisfaction is noteworthy given the frequency of messaging, which for most studies included at least daily messages, reflecting a relatively high tolerance in this group for intervention. The evidence of feasibility in these studies suggests that adolescent-specific text messaging and mobile phone‒based approaches may be an important and promising area of future intervention development. Given methodological limitations in the studies reviewed, larger studies with long-term outcomes are warranted, particularly with sufficient power for clinically important outcomes. Recent evidence supports the efficacy of text message and mobile phone app interventions to promote preventive behavior among adolescents [51]. Furthermore, in addition to efficacy data, cost-effectiveness is another important aspect of intervention evaluation [9,22,52]. The cost to develop and maintain each intervention can be a barrier to widespread use of these interventions. Cost also impacts the variability in patients’ access to technologies. Formal economic evaluation of different interventions will inform health care authorities to decide whether adoption of such interventions would be economically efficient [22,52]. In a related systematic review, we found insufficient evidence regarding the cost-effectiveness of text messaging and mobile phone apps to promote adherence in adolescents with CHCs [52]. Our findings highlight the need for further investigation of cost-effectiveness to inform the scalability, sustainability, and future cost savings of such approaches [52].

Our systematic review has a number of strengths. First, we conducted our review following the recommendations for rigorous systematic reviews methodology [32,49,53]. Second, we used a highly sensitive search strategy guided by a librarian information specialist and had no language restrictions in order to minimize publication bias by identifying as many relevant studies as possible. Additional resources were also searched including published systematic reviews, clinical trial registries, and different electronic databases. Third, although we limited our search to studies published since 1995, there were no eligible studies identified before 2005. Therefore, the possibility that we have missed earlier studies is very small. Finally, 2 authors completed the review process independently at all stages.

Our systematic review of the literature should be considered within the context of some potential methodological limitations. First, similar to any other systematic literature review, although our search criteria were designed to be comprehensive, it is possible that we missed relevant articles. Second, we included only articles published in peer-reviewed journals, and publication bias with the tendency to report positive study results cannot be excluded [54]. Third, a number of studies reported insufficient information, the definition of adherence varied, and the study sample size and ages as well as methods of adherence assessment used in the included studies were heterogeneous. These limitations prohibited a meta-analysis from being performed [55]. Finally, many of the included studies had relatively small sample sizes.

The number of adolescents with CHCs continues to increase and medication nonadherence is a clear challenge. The use of text messaging and mobile phone app interventions to improve medication adherence among adolescents with CHCs has shown promising feasibility and acceptability; however, there is modest evidence to support their efficacy. Further evaluation of short- and long-term efficacy and cost-effectiveness of these interventions is warranted. In addition, better understanding of barriers to medication adherence would inform further development of text message and/or mobile phone app interventions to improve adherence and health outcomes in adolescents with CHCs, such as sickle cell disease [56,57]. Adolescents are frequent users of text messaging and mobile phone apps, and engaging adolescents with CHCs in their self-management is critical for improving long-term outcomes. Seeking adolescents’ perspectives could enhance uptake and long-term engagement, while minimizing patient fatigue. The currently available data from medication adherence studies using text messaging and mobile phone app interventions are heterogeneous, particularly in relation to process and outcomes measures, which limit the evidence generated and conclusions that can be drawn from those studies. Nevertheless, consistent use of Web-based and mobile health interventions reporting guidelines [58] would enhance comparative research across studies [59]. The functionalities of different mobile technologies continue to improve with gradually decreasing cost suggesting potential economies of scale where interventions could be delivered to large populations.