Qualitative and quantitative studies were included to answer the following two principal research questions.

1. How do wearables impact the health outcomes of NCDs in LMICs?

Wearables were defined as any device that can be worn by the user and included watches, wristbands, textiles, or accessories, such as rings and glasses. LMICs were determined using the 2022 World Bank Classification [11].

CVDs and their associated major metabolic risk factors, namely obesity, diabetes, and hypertension, were the primary focus due to their widespread prevalence and significant impact on mortality rates, particularly in LMICs [1, 2].

Health outcomes were defined as any reported changes in quantifiable parameters of the selected NCDs, such as body mass index (BMI), blood pressure (BP), waist circumference, or blood glucose and cholesterol levels. Other key indicators for evaluating effectiveness (such as calories burned or number of steps) were reported if mentioned.

Factors influencing or impeding the adoption of wearables were any quantitative or qualitative factors reported by consumers, healthcare professionals, or other relevant stakeholders, including non-governmental organizations (NGOs) or health ministries’ staff.

The detailed inclusion and exclusion criteria are summarized in Table 1.

The search strategy was developed and conducted by the main author (MS). The search terms comprised of three key concepts:

a) Wearables, smartwatches, and other similar terms

To ensure the inclusion of all relevant literature since the launch of the first mainstream wearable device, the Fitbit Classic^®, studies from January 2010 until June 2023 were searched in two databases, PubMed, and Web of Science because of their extensive use and global coverage. The complete search strategy and research queries can be found in Supplementary Material S1.

A secondary search using a backward and forward snowball strategy and hand-searching helped identify additional relevant studies.

If it was not possible to ascertain the study eligibility after the initial abstract and title screening, the manuscript was retrieved for a full-text reading.

The final search was conducted on 26th June 2023 and yielded a total of 890 records from the two databases (Figure 1). The publications were then exported to Rayyan, a web-based tool for systematic reviews, which allows for deduplication, documentation of inclusion and exclusion reasons, as well as blinding selection decisions between reviewers [12]. Out of the 890 publications identified, 277 duplicates were removed, and 580 articles were excluded after an initial title and abstract screening since they did not fully meet the inclusion criteria. The remaining 33 articles were then assessed for eligibility after a full-text reading by two reviewers (MS and VC).

Relevant information from all 17 studies was extracted from available full texts and reported into Excel 2023.

Four health outcomes commonly reported (BMI, glycated hemoglobin (HbA1c), systolic and diastolic blood pressure) were evaluated. Change-from-baseline and standard deviations were calculated from absolute values at baseline and at endpoint using the recommendations from Cochrane’s Handbook [13] unless already reported in the published results (Supplementary Table S2). To ensure a certain homogeneity, a 3-month endpoint was used for the present analysis as it was available for all, but one study [14]. Despite its shorter follow-up (2 months) this study by Shenoy et al. was nevertheless included in the analysis.

Standardized mean differences (SMD) and 95% confidence intervals were then calculated using RevMan (version 5.4.1) and forest plots were generated for each parameter. A SMD value of zero meant the changes in health outcomes were similar in both intervention and control groups. A negative SMD value suggested a stronger effect of the intervention on the changes in health outcomes compared to those observed in the control group. On the other hand, a positive SMD value indicated the changes observed in the intervention group were less than those observed in the control group. To interpret the strength of the effect, Cohen’s rules of thumb was used, where a SMD of ±0.2 represents a small effect, a SMD of ±0.5 represents a medium effect, and a SMD of ±0.8 represents a large effect [15, 16].

Risk of bias in individual RCTs was assessed using the Cochrane tools RoB 2 for randomized studies [17]. The likely extent of bias for each study was evaluated with a set of questions across different domains and reported following the GRADE guidelines (Supplementary Table S3) [18].

Factors evoked in the studies investigating barriers and facilitators impacting the adoption of wearables were summarized in a chart and described individually.

Out of the 17 studies included, 10 were randomized controlled trials (RCTs). All studies were conducted in Asia, six (35%) in China, two (12%) studies each in Thailand, Turkey, and Cambodia, and the five (29%) remaining studies were set up in India, Pakistan, Indonesia, Sri Lanka, and Malaysia. All were conducted in middle-income countries: 65% (11/17) in upper-middle-income countries (UMIC) and the rest in lower-middle-income countries (LMIC). Our search did not find eligible studies set up in Africa or South America.

Nine RCTs looked at health outcomes of wearables; five (56%) reported changes in BMI, four (44%) changes in blood pressure (BP), and four (44%) measured changes in HbA1c. Six studies followed patients for 3 months, the remaining three had a follow-up of two, six, and 12 months, respectively.

Although pedometers were the most frequent wearables evaluated (6/9, 67%), study designs and the type of wearables exhibited heterogeneity (Table 2). Four of these studies reported the number of steps done using a pedometer, of which 2 reported it only in the intervention group and none reported number of calories burned (Supplementary Table S2).

The forest plots (Figures 2A–D) summarize the impact of the interventions on the four key health outcomes: changes in HbA1c levels, diastolic and systolic blood pressure values and body-mass index were the most commonly measured health outcomes and are detailed below.

The risk of bias in individual RCT was considered low or with moderate concerns across all 5 domains except in one study (Supplementary Table S3). Although studies were not blinded, analysis was mostly performed by staff independent of the clinical team. Loss to follow-up was minimal in most studies. In one study, the risk of bias was high concerning the randomization process and raised some concerns about bias in outcome measurement.

A total of eight studies, including one RCT explored factors that influenced the adoption of wearables in LMICs. The studies are summarized in Table 3.

The barriers to adoption of wearables highlighted in the studies above have been summarized in Figure 3.

Liverani et al. conducted two studies in Cambodia involving national and international stakeholders, as well as the general population. They revealed significant disparities between urban and rural populations concerning the adoption of wearables. Practical inconveniences associated with the use of wearables posed a significant barrier, particularly for farmers. Their studies highlighted that wearables were more suitable for young, urban residents, as they commonly use email accounts or mobile applications. Additionally, among poorer populations, the cost of wearables was cited as a major obstacle. Furthermore, the lack of shared Application Programming Interfaces (APIs) by private companies made customization of wearables challenging for specific contexts and languages.

Limited human resources and health system capacities were identified as crucial barriers to the widespread adoption of wearables for patient monitoring. Extensive training for health workers and potential challenges in integrating the increased volume of data generated by wearables into health information systems were also noted. Moreover, inadequate funding and a lack of coordination between donors made scaling up beyond pilot projects difficult.

The authors also observed that, culturally in Cambodia, the elderly are often cared for by their children, which led to wearables being perceived as less relevant in the community. Similarly, Huang et al. found in their study in Chinese community dwellings that participants with more children were less willing to choose smart senior care, including the use of wearable sensors in self-monitoring.

In their study based in Pakistan, Nadeem et al. identified the most common barriers to wearable technology adoption in diabetes management. The survey respondents reported that their diabetes care team had never discussed the use of wearables as a management option with them. Other common barriers included the patients' comfort with wearing the device as well as concerns about drawing attention from others.

Arambepola et al. reported in their study in Sri Lanka that feedback from activity monitors was the most frequently cited motivation for changing activity behaviors. However, participants required some guidance in interpreting the feedback initially. Similarly, Lukkahatai et al. found in their study in Thailand that participants who could see their steps significantly improved their step count and sleep duration compared to those who could not see their results.

Finally, in Yao et al.’s study on factors predicting general practitioners (GPs)’ adoption of wearable ECG devices in China, privacy concerns, social influences from peers and senior physicians, and price perception were identified as key factors influencing the adoption of wearable ECG devices.

Our results align with previous studies indicating that wearable devices tend to be effective in increasing physical activity and supporting weight loss. However, they often have limited benefits in managing chronic diseases, and for other health outcomes, their effects tend to be small, often insignificant, and inconsistent [9, 37].

Out of the four studies that investigated the impact of wearables on HbA1c levels, only one study reported a significant improvement compared to the control group. Regarding systolic and diastolic blood pressures, significant improvements were found in three out of four studies for SBP and two out of four studies for DBP. These results are surprising, since controlling SBP is generally considered more difficult than DBP [38]. Finally, four out of five studies reported significant improvements in BMI.

The factors evoked in the studies included in this scoping review are quite similar to those that have already been reported. In a systematic review analyzing the barriers to the sustainability of digital health interventions in LMICs, factors such as limited technological literacy, cost, and limited human resources were evoked [39], which were similar to those that were reported in the studies included in this review.

Self-monitoring is not well-established in many LMICs, particularly among the elderly who are often cared for by family members [28, 33]. To address this cultural barrier children or younger caretakers could be involved by helping the elderly learn to use wearables, allowing them to self-monitor. Younger caregivers could also use these tools to monitor the elderly from a distance, therefore reducing the burden on family caregivers.

Additionally, one study revealed that patients often lacked sufficient information and support from healthcare professionals concerning wearable usage, emphasizing the need to first educate healthcare providers on the benefits of these wearables, so they can effectively communicate this knowledge to patients [29].

The feedback on activity levels provided by many wearables was cited as a common motivator to increase activity levels. However, customizing these products to local contexts and languages would enhance their usability and overall effectiveness. Furthermore, it was emphasized that wearables should be designed to withstand local working conditions. Therefore, by incorporating these features, wearables could be better suited to meet the needs of users in diverse LMIC settings [30].

This scoping review has some limitations. Firstly, the search was conducted in two databases, PubMed and Web of Science; it was restricted to articles in French or English, with the search concluding in June 2023. Despite hosting a large body of literature worldwide, this restricted search may have left out relevant articles originating from countries underrepresented in the present manuscript, as well as studies with larger sample sizes. This field of research is rapidly expanding and future updates would certainly benefit from including regional databases and sources such as AJOL and LILACS. It could provide a more complete and a diverse perspective on factors influencing the adoption of wearables, and potentially could be challenging or refining our current conclusions. It would equally be important to consider the use of AI translation tools to include a larger body of manuscripts.

Secondly, despite a low to moderate risk of bias, the studies often had small sample sizes that could have limited the robustness of their results; and study designs were diverse, making direct comparisons and generalization challenging. In this context, subgroup analyses and meta regressions were not possible.

Another notable limitation was the relatively short study duration in most of the studies exploring the impact of wearables. Most of these studies had a study duration of only 3 months. Consequently, the potential long-term impact of wearables on cardiovascular health outcomes which take time to improve remains inadequately explored. Additionally, health outcomes such as number of calories burned or number of steps taken could not be evaluated: none of the studies included the number of calories burned, and only two included the number of steps in both intervention and control groups.

A strength of the studies included is that most studies included multi-component interventions, such as health education, personal fitness programs, supervised exercise, and low-calorie diets, providing a comprehensive approach to health improvement. The review also included studies with variable health technologies, offering insights into different types of devices.

The use of wearables in LMICs might be effective in improving health outcomes, such as BMI and DBP, especially when used in conjunction with other health interventions. However, to enhance adoption, there is a need for low-cost and customizable wearables that address specific local needs. It is paramount to better understand the long-term impact of wearables on health outcomes, especially in LMICs if one wants to foster their broader adoption in these regions. As a rapidly growing field of research, it is likely that more studies are being conducted in these regions. However, there is a need to harmonize study designs and study endpoints as well as increasing the sample size and the duration of follow-up so that the long-term impact of wearables with or without additional health interventions can be effectively evaluated.