A systematic review design using qualitative methods was used. We adhered to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [25]. Our PRISMA checklist is provided in Multimedia Appendix 1. This review was registered with the PROSPERO (International Prospective Register for Systematic Reviews) on April 22, 2021 (CRD42021244562).

A search of studies published between January 1, 2016, and July 1, 2021, was performed using PubMed, EMBASE, Web of Science, Scopus, CINAHL, and Cochrane Central Register of Controlled Trials. Studies published before 2016 were excluded to reflect the rate of technological advancement in the research and development of wearables [26]. Our strategy was developed in consultation with a medical research librarian. A combination of Medical Subject Headings (MeSH) terms and free text keyword terms was used, including chronic disease (MeSH), wearable electronic devices (MeSH), health care OR outcome*, and select chronic conditions such as asthma. Our full search strategy is available in Multimedia Appendix 2.

Chronic disease is defined as any health condition lasting ≥3 months, which may lead to other health complications and may be associated with functional impairment or disability [27]. A health care outcome is defined as any parameter that demonstrates an effect on the patient experience, health care outcome (such as improvement in glycemic control in diabetes), clinician experience, or cost of health care delivery.

The inclusion criteria were (1) randomized controlled trials (RCTs) and observational studies, (2) feasibility studies observing the effect of wearables on a predefined health care outcome, and (3) studies published in peer-reviewed journals in English. Studies involving adults and children were included.

The exclusion criteria were (1) pregnant patient population, (2) studies demonstrating the validity or technological feasibility of wearables, (3) studies reporting the accuracy of wearables as their primary aim, (4) books or book chapters, (5) conference abstracts, and (6) systematic reviews.

Screening of potentially eligible studies was performed in 3 steps: duplicate removal, title and abstract screening, and full-text screening. Duplicates were removed using EndNote (version 20; Clarivate) and Covidence [28]. Additional duplicates that were not removed during this process were removed manually. A total of 2 review authors (GM and DF) independently screened the titles and abstracts for inclusion using the predefined inclusion and exclusion criteria specified previously. All studies not discarded through this process were then screened via a full-text review process by 2 review authors (GM and OC), from which studies were identified for inclusion. Conflicts that could not be resolved between the 2 review authors were resolved by a third investigator (CS). Full data extraction, categorization, and labeling of papers were performed by 1 author (GM) and validated by a second author (OC).

A risk of bias assessment was conducted for all RCTs by 1 author (GM). The Cochrane Collaboration’s tool for assessing the risk of bias [29] was used to assess each randomized study for bias from the randomization process, selection bias, bias because of missing outcome data, bias because of measurement of the outcome, and other biases otherwise not addressed. These are presented in Multimedia Appendix 3 [30-52]. For nonrandomized studies, the Risk of Bias in Nonrandomized Studies of Interventions tool [53] was used to assess bias because of confounding, selection bias, bias in classification of interventions, bias because of deviations from the intended interventions, bias because of missing data, bias in the measurement of outcomes, and selection of result bias. These are presented in Multimedia Appendix 4 [54-59].

All data were extracted from the identified studies under 6 key extraction themes that were most suited to address our original research question [60] (Textbox 1). A wide range of subheadings was selected, given the high variance in disease populations and outcome measures. A narrative qualitative synthesis of the included studies was conducted. The high heterogeneity of study designs, disease groups, patient populations, and outcome measures precluded the completion of the meta-analysis. Our results are based on the disease group, with relevant findings across all studies reported within the disease group in question. All outcomes were categorized as a component of patient experience, clinician experience, health care outcomes, or cost in alignment with the Quadruple Aim. A robust assessment of the synthesis is subsequently presented through critical reflection.

The wearables used in these studies were designed to be worn either continuously or intermittently on the body. Their capabilities and commercial availabilities are summarized in Table 1. Of the 30 publications, 19 (63%) studied wearables placed on the waist (n=10, 53%) and wrist (n=9, 47%), with 3 (10%) publications studying wearables placed on multiple sites of the body. Multiple types of wearables were studied, including pedometers, smartbands, virtual and augmented reality (AR) systems, flash glucose monitoring systems, and intelligent shoe insoles.

The study selection sequence is outlined in a PRISMA flowchart, which is presented in Figure 1. Our search yielded 2078 articles, with a further 7 articles identified through snowballing. From a total of 2085 articles, 409 (19.62%) were identified as duplicates and were thus removed, and a further 1576 (75.59%) studies were excluded after screening abstracts between July and August 2021, leaving 129 (6.19%) studies assessed for eligibility via full-text review, of which 99 (4.75%) were excluded. One study was identified through snowballing. A total of 31 studies met the inclusion criteria. Following peer review, 1 study was removed, leaving 30 studies included in our qualitative synthesis.

Multimedia Appendix 5 [30-52,54-59] presents the characteristics of all included studies. A total of 2446 participants across 9 countries were included. These participants had a mean age ranging from 10.1 to 74.4 years and 56.42% (1380/2446) were female. Of the 30 studies, 2 (7%) targeted a pediatric population (aged<18 years [30,31], 21 (70%) studied adults with a mean age between 40 and 65 years [32-49,54,55,61], and 6 (20%) evaluated individuals aged >65 years [50-52,56-58]. Approximately 50% (15/30) of studies recruited participants from specialist tertiary clinics, with the remaining 50% (15/30) recruiting participants in the community (eg, primary care settings and rehabilitation centers). Of the 30 studies, 24 (80%) were randomized, with the remaining 6 (20%) using a nonrandomized methodology. All randomized studies were subject to a risk of bias in the blinding of participants because of the nature of the wearable intervention being present or absent, with a notable exception being the use of sham virtual reality (VR) headsets in a study, which used identical hardware with different software with which the user interacted [35].

Within the literature synthesized, wearables and their influence on health care outcomes in 8 disease systems were studied across 18 chronic diseases. There were predominantly mixed findings within the studies included in this review, which are summarized in the following sections.

Of the 30 studies, 15 (50%) found a positive impact of wearables on the primary outcome studied. These findings were observed in studies involving multiple chronic disease systems and using multiple wearables. Wearables can be used to deliver nonpharmacological therapy to improve subjective pain scoring, as demonstrated by both Austin et al [37] and Garcia et al [35] in using 3D VR systems to deliver cognitive behavioral therapy and mindfulness practices to users with chronic pain. Tunur et al [59] found a lack of effect in delivering therapy through AR-based dance interventions in Parkinson disease, although the outcome studied was motor assessment, and only 7 participants were enrolled in the study.

Approximately 37% (11/30) of studies observed the capability of wearables to encourage and improve physical activity. There was no uniform measurement system to quantify physical activity across the included studies, with the 6MWT [31,39,47,48], 10-minute walk test [52], minutes of moderate to vigorous physical activity time [34,36], step count [45], MWD [51], VO₂ max [49], and adherence to prescribed walking therapy [56] used to measure exercise capacity. A positive impact was observed in 17% (5/30) of studies assessing the influence of wearables on physical activity. There did not appear to be a strong relationship between wearables and improvement in physical activity in these studies.

Approximately 20% (6/30) of studies [33,35,37,38,58,61] used subjective scoring systems to assess the influence of the studied wearables on primary outcome measures. Of these 6 studies, 5 (83%) [35,37,38,58,61] found a positive effect on the primary outcome across multiple chronic diseases. Given the potential for wearables to empower users with additional health-related data, subjective scoring systems may improve significantly using wearables, which may be reflected in health-related quality of life assessments. However, further research is required in this area.

Several associations were noted between the types of wearables used and the studied outcomes. One such observation was the positive effect of 3D VR systems on pain scoring in 2 distinct chronic pain syndromes [35,37], in which both studies used sham VR as the comparator. Both publications reported minimal side effects of using these immersive systems, with specialized programs used to counter the possible effects of cybersickness (motion sickness specific to the use of VR headsets). The use of either pedometers or smartbands (eg, Fitbit devices) had mixed effects on exercise capacity, with 10% (3/30) of studies [31,39,51] reporting improvements in walking distance using these devices and 7% (2/30) of studies [48,63] finding no influence on this outcome. Mixed results were also observed when studying the influence of either pedometers or smartbands on weight. Kim et al [32] found a significant reduction in BMI through the use of a Samsung Charm fitness tracker in people with obstructive sleep apnea. This is in contrast with the findings of both Lystrup et al [46] (using a Fitbit Charge) and Takahashi et al [63] (using a pedometer).

A total of 155 outcome measures were studied within the included studies. Of these 155 outcomes, 139 (89.7%) addressed the health care outcomes component of the Quadruple Aim, with 12 (7.7%) representing the patient experience of using wearables. Approximately 1.3% (2/155) of outcome measures evaluated the clinician experiences of wearables, with the remaining 1.3% (2/155) addressing cost. Within the health care outcomes, the most frequently studied included pain (11/155, 7.1%), quality of life (7/155, 4.5%), step count (7/155, 4.5%), and physical function (5/155, 3.2%). Approximately 7% (2/30) of studies explored the impact of wearables on patient experience as their primary outcome [51,57]. All outcome measures mapped to the patient experience included acceptance of the technology [57], adherence to wearing the device [36,44], compliance [50], engagement [35], presence (using VR) [37], satisfaction [31,35] and usability [50]. A summary of all outcomes in the included studies is presented in Multimedia Appendix 6 [30-52,54-59].

There were no studies in which the primary outcome could be classified as addressing either the clinician experience or cost of health care provision, with only 4 secondary or exploratory outcomes representing these facets of the Quadruple Aim. In the management of Parkinson disease, it was noted that when using data generated by a finger-worn wearable, the consultation time spent with a neurologist was significantly reduced when compared with an in-person clinical assessment, although this was reported with a statistical statement of inequality (range 29-45 minutes vs 45-60 minutes; P<.05) [50]. Widyastuti et al [48] performed a cost-effectiveness analysis of pedometer-guided PR in COPD compared with conventional center-based PR and found that the cost of establishing pedometer-guided care was significantly cheaper than that of conventional PR, with a mean cost-saving of €76.3 (US $80.30) per patient. Maddison et al [49] also performed a cost-effectiveness analysis of telerehabilitation versus center-based rehabilitation, reporting a per capita program delivery difference of £1185 (US $1247.10; P=.02); however, hospital service use costs were not significantly different (P=.20).

A critical reflection of the synthesis process was performed [64]. The methodology used for this synthesis was aligned with the PRISMA guidelines and registered with PROSPERO, and data extraction was performed based on 6 main themes. It was believed that this enabled researchers to answer the predefined research question. Although a wide range of chronic diseases and wearables were studied and included in our synthesis, certain patient populations were not represented, including those with underlying mental health conditions, chronic skin conditions, or malignancy. This may reflect a limitation of our search strategy or may equally reflect a lack of understanding of the use of wearables for certain disease groups.

The evidence used was not subject to major bias. Most publications involved a relatively small number of participants, and only 80% (24/30) of the studies were randomized. No assumptions were made regarding adherence, acceptability, intended use, or anticipated outcomes in the included studies. No discrepancies were identified within the included studies, and focus areas of future research are highlighted in the Discussion section. The strict search criteria precluded the identification of ongoing studies. However, by reviewing the many trial protocols in our abstract screening and review of the gray literature, it is anticipated that the rate of technological advancement and interest in the field will strengthen the existing evidence base in wearables and health care outcomes in chronic disease management. The aspects that may influence the implementation of wearables in existing health care platforms are highlighted in the Discussion section.

A total of 30 studies published between 2016 and 2021 investigated the effectiveness of wearables in improving health care outcomes in individuals with chronic diseases. Our systematic review found both positive and neutral results when studying the influence of wearables on health care outcomes in chronic disease. Encouragingly, none of the identified studies demonstrated a negative impact of wearables on outcomes, such as adverse effects, treatment burden, or the inability of study participants to effectively use the prescribed technology because of poor inadequate digital literacy.

We identified several studies demonstrating the positive influence of the studied wearables on primary health care outcomes. One such example is the positive influence of guided exercise prescription using the Nike+ FuelBand on peripheral vascular disease with intermittent claudication [51]. Approximately 7% (2/30) of RCTs found a significant improvement in quality of life indices [51,61], which correlated with either greater self-efficacy in chronic disease management [61] or functional capacity [51].

A key finding was the lack of a clear association between the use of a particular wearable and its acceptability within a chronic disease population. Most health apps synchronizing with wearables, including those studied in this review, are disease focused and provide information on the studied conditions, such as pressure points in diabetic foot ulceration. Although this does provide clinically useful information, it is unclear whether patients use this information effectively to guide their self-management or even find it useful. We also have limited knowledge of the acceptability of wearables to patients with chronic disease or whether there are any barriers to implementation within specific chronic disease groups. Supporting mobile apps should be designed using a patient-centered approach, incorporating personalized advice and recommendations from the health data provided by the accompanying wearable. This may unlock the potential of wearables in chronic diseases.

To rationalize the incorporation of wearables into existing health care models, it is important to demonstrate cost-effectiveness; however, economic outcomes were not incorporated into the publications identified in this review. Wearables enable remote monitoring, which has its own economic advantages. One of the studies involving patients with implantable cardiac defibrillators found that remote monitoring reduced health care costs by 25% [65]. Transferring even a small proportion of center-based monitoring services established for people with chronic disease remotely through the use of wearables has the potential to substantially reduce the cost of care delivery in many chronic disease settings. Robust economic analyses incorporating the number needed to treat analyses should be inherent to future studies on wearable technologies.

There are several challenges in the implementation of wearables in health care. Wearables were initially designed for the health and fitness industry and are not subject to the regulatory standards required for medical equipment. Only a handful of wearable technologies have been approved by the US Food and Drug Administration, such as the Apple Watch Series 6 electrocardiogram app for detecting atrial fibrillation [66]. Most commercially available wearables are classified as nonmedical devices, and although some software within these are accurate, they cannot support clinical decision-making without legal regulation [67]. We noted that 73% (22/30) of studies included in this systematic review used commercially available wearable technology that is not currently legislated under either the US Food and Drug Administration or the Therapeutic Goods Administration (Australia); therefore, this may represent a limitation in applicability to health care settings. Wearables can also collect many different aspects of user information using sensor technology, which requires stringent data security and privacy processes. The accuracy of recently implemented software in wearables is variable and must be recognized to prevent avoidable misdiagnosis and unnecessary patient-clinician interfaces and minimize patient anxiety [68]. Until these challenges are clearly addressed, it is difficult to envision the safe implementation of wearables in existing health care models.

There is a high degree of variance in the chronic diseases studied and in the wearables used, which limits the ability to provide a strong evidence base to support the use of a specific wearable to treat a specific cause. Further research needs to focus on the impact of a specific wearable for a specific chronic disease to generate evidence for its use, especially given the multiple capabilities of most wearables. In addition, there may be several applications within the studied wearables for which only one wearable is particularly effective for a chronic disease group. Understanding these capabilities through targeted research is essential for the implementation of wearables in chronic disease management.

A finding from this systematic review was the lack of studies focusing on the clinician experience of wearables in health care. Without a deep understanding of the perceived benefits and risks of wearables in chronic disease management from the clinician’s perspective, it is difficult to envision their integration into clinical practice. Challenges to more established digital health transformative initiatives, such as introducing electronic health records, likely exist for the use of wearables in health care. These include changes to clinical workflows, patient safety concerns, learning curves of practitioners in understanding new technology, and challenges in integrating systems [69]. Interdisciplinary teams, including clinicians, data scientists, and experts in artificial intelligence, will be required to shape the future of wearables in health care because of the limited familiarity of clinicians with big data and artificial intelligence [70]. Alongside observational and randomized controlled quantitative analyses, qualitative research into the clinician’s perspective will prove invaluable in optimizing the use of wearables in health care.

There were several strengths of our review. Given the rapid shift to care decentralization amidst a global pandemic [71], this review provides a foundational evidence base for the effect of wearables in improving health care outcomes for individuals with chronic diseases. A large number of participants was obtained, encompassing a range of chronic diseases, with a wide mean age range and reasonably equal gender distribution. Approximately 80% (24/30) of studies implemented a randomized controlled design, generating a comparator to assess the effect of the wearable intervention studied, which was predominantly the gold standard of care for the studied chronic disease. Although pedometers have been outdated commercially by newer technology, their inclusion as wearables in this study remains a strength because of the long-term evidence of their ability to encourage physical activity [72].

Our review has several limitations, such as the use of now outdated wearables. All smartwatch brands used in these studies have since been superseded by more advanced devices with greater technical capabilities. Although this may simply reflect the rate of technological advancement, it may also indicate that the studies published in this review may not reflect the capabilities of wearable technology from 2021 onward. It is unknown whether advanced information, including heart rate variability, Firstbeat algorithms such as stress levels and Garmin’s Body Battery, and advanced sleep stage estimation may infer real-time information on disease status. The rate of increasing technical capability of wearables justifies the regular systematic review of the literature, given the increasing publication outputs and commercial availability of more advanced wearables.

The interventions used as comparators differed greatly between studies, which made the comparison of the results challenging. The heterogeneity of the participants in each study precluded the ability to perform a meta-analysis, leading to an inability to assess the strength of evidence for a particular wearable on a predefined health care outcome. Although there were no apparent major risks of bias in our risk of bias assessments, this process was conducted by only 1 researcher. In addition, within the included studies, there was a lack of multicenter trials, which we propose should be conducted to increase the statistical robustness. All outcome measures reported were quantitative in nature, which represents a limitation, as qualitative research strengthens the existing evidence base. Given the increasing number of publications in this field, it is highly likely that a meta-analysis will be feasible for future systematic reviews of this nature. In future studies, all 4 end points of the Quadruple Aim should be represented to aid implementation in health care systems. We believe that further research is worth conducting to strengthen the evidence regarding wearables in chronic disease management.

Our systematic review did not find a clear role of wearables in improving health care outcomes in chronic disease. However, wearables are becoming increasingly popular within the community, and as research and development in wearable technology progress, it is anticipated that these devices will play an increasing role in supporting healthy lifestyle modifications for their users. More research is required to ascertain a clear causality between wearables and health care outcomes, as defined by the Quadruple Aim, for people with chronic disease. As the evidence base for the use of wearables in chronic disease management is strengthened, further challenges will need to be overcome to allow widespread adoption in the health care setting, including stringent regulatory approval, data privacy and confidentiality, and software accuracy.