This systematic review (registered in PROSPERO [International Prospective Register of Systematic Reviews]: CRD420251074143) followed the guidelines outlined in the PRISMA-COSMIN (Preferred Reporting Items for Systematic Reviews and Meta-Analyses-Consensus-Based Standards for the Selection of Health Measurement Instruments) for Outcome Measures Instruments in Systematic Reviews (OMIs) (Checklist 1) [22]. The only amendments to the registered review protocol were the inclusion of additional searches from Google Scholar and citation lists to ensure a more comprehensive review of the relevant literature.

The literature search was reported according to PRISMA-S to promote transparency of the search process [23]. The search terms were constructed using the PICO (Population, Intervention, Control, and Outcomes) framework, as displayed in Table S1 in Multimedia Appendix 1. Six databases, including PubMed (National Library of Medicine), Web of Science (Clarivate), Cochrane Library (Wiley), IEEE Xplore (IEEE), Scopus (Elsevier), and Embase (Elsevier), were systematically searched on March 7, 2025, without any restrictions or search filters. Major terms related to older adults, digital technology, assessment, and locomotor capacity were combined for the initial search, according to the review by Honvo et al [24]. The search strategy was developed in consultation with an experienced librarian and experts in geriatric medicine and digital health. In addition, we searched citation lists from included studies by browsing reference lists and gray literature on Google Scholar on November 17, 2025, limiting the search to the first 100 publications, as search results beyond this number rapidly lost relevance to the topic of the review [25]. No other online resources, such as conference proceedings or print journals, were specifically browsed for this review. No study registries were searched, and no additional studies were sought by contacting authors, manufacturers, or others. The search strategy is detailed in Table S2 in Multimedia Appendix 1. The search process was not updated after the initial search date.

Following the PICO framework, the inclusion criteria were (1) study population (P): studies enrolling older adults as research subjects or separately reporting results on older people; (2) interventions (I): studies focusing on digital biomarkers (defined as objective, quantifiable physiological and behavioral measurements acquired through portable, wearable, implantable, or ingestible digital devices [15]), derived from digital technologies; (3) comparators (C): any or none; (4) outcomes (O): studies on the assessment of locomotor capacity (ie, endurance, balance, muscle strength, muscle function, muscle power, joint function) while examining the measurement properties of reliability and validity; and (5) studies written in English.

Reliability focused on 2 types [26,27]: (1) test‐retest reliability: the extent to which scores for patients who have not changed are the same for repeated measurements over time; and (2) measurement error: the systematic and random error of a patient’s score that is not attributed to true changes in the construct to be measured.

Validity prioritized 2 categories [27]: (1) criterion validity: the degree to which the scores of a device are an adequate reflection of a “gold standard”; and (2) hypothesis testing for construct validity: the degree to which the scores of a device are consistent with hypotheses, including convergent validity (measured by relationships to scores of other instruments) and discriminative validity (assessed through differences between relevant groups).

The exclusion criteria included the following: (1) systematic reviews and literature reviews; (2) books and other non–peer-reviewed literature; (3) studies conducted on populations with specific medical conditions (eg, Parkinson disease, stroke); (4) studies focusing on digital measures during rehabilitation exercises (eg, supervised therapeutic movements, postsurgical rehabilitation protocols, physical therapy interventions, and exercise programs designed for recovery or strength training); (5) studies lacking data on measurement properties; and (6) studies without full text.

After the literature search, all identified studies were imported into EndNote 21, and duplicates were removed. Two authors independently screened the remaining titles and abstracts for potential eligibility and then retrieved the full-text articles of the selected studies. Any discrepancies regarding study selection were resolved through consultation with a third reviewer.

Two authors independently extracted data using standardized forms (Table S3 in Multimedia Appendix 1), and discrepancies were resolved through discussion among the 3 authors. Data items comprised: (1) characteristics of studies (first author, year of publication, and study design); (2) participant demographic information (sample size, age, sex distribution, and setting); (3) digital technology- or device-related information (type, number, assessment tasks, and digital biomarkers); and (4) results of validity and reliability.

A total of 14,680 articles were identified from 6 databases, and 4,733 duplicates were removed. Following the initial screening of titles and abstracts, 123 articles were fully evaluated, with 9 of them eligible for inclusion in the review [31-39]. An additional 5 publications were obtained from reference lists [40,41] and gray literature [42-44]. The study selection process is presented in Figure 1.

The main characteristics of the included studies and populations are summarized in Table 1. These studies were published between 2013 and 2024 and were conducted across 10 countries, primarily in the United States (n=3). Thirteen studies employed a cross-sectional design, while 1 adopted a pretest-posttest design. Across all studies, balance was the predominant assessment target (n=13), with only 1 study evaluating muscle power. No eligible studies assessed endurance, muscle strength, or joint function, and the exclusion reasons included incorrect population [45], lack of reliability and/or validity data [46], studies not involving digital biomarkers or technologies [47,48], and conference abstracts [49]. The sample sizes ranged from 12 to 248, with the majority from community-dwelling populations.

One study assessed muscle power using a smartphone during a sit-to-stand task [31]. The derived variables, including total movement duration, peak force, rate of force development, and peak power, were measured. Complete datasets were obtained, and the raw signals were filtered before analysis. Test-retest reliability of these variables ranged from 0.43 to 0.92. Measurement error, expressed as SEM, varied between 3.1 and 26.1. Convergent validity, evaluated by comparing smartphone measures with force plate data, showed a correlation coefficient ranging from 0.69 to 0.98, with peak power demonstrating a statistically significant correlation.

This systematic review summarized digital technologies and biomarkers for assessing locomotor capacity in general older adults, including their validity and reliability. Fourteen studies were included, identifying 5 types of digital technologies and 51 digital biomarkers (47 for balance and 4 for muscle power). Twelve studies reported generally acceptable test-retest reliability and validity for balance, indicating promising potential. One study evaluated muscle power and demonstrated good reliability and validity, though further research is needed to validate these findings. However, caution is warranted when interpreting these results, as reliability and validity vary by context, and the overall strength of evidence remains limited. Future research should address other locomotor components and focus on developing assessments that simultaneously ensure suitability, reliability, and validity [50,51].

The test-retest reliability reported across included studies was acceptable, suggesting that digital technologies can produce stable measurements of locomotor capacity among older adults. Nevertheless, there was considerable variability in reported reliability coefficients, ranging from 0.016 to 0.97. This broad range highlights the nascent stage of the field and could be attributed to differences in device modalities, task settings, and the specific digital biomarkers extracted. Specifically, force plates generally exhibited more consistent test-retest reliability than smartphone- and IMU-based systems, which is commonly attributed to their direct measurement of ground reaction forces and COP trajectories and the absence of variability introduced by wearable sensor placement and soft-tissue artifacts [52,53]. Consistent with previous literature [54], portable sensors yielded higher and more consistent reliability for static tasks than for dynamic tasks [43,44], likely because dynamic conditions introduce increased movement variability, sensor micro-slippage, and more complex signal processing requirements. Variable choice also markedly influenced reliability. For example, in a wearable sensor-based sit-to-stand assessment, linear acceleration and angular velocity variables demonstrated higher test-retest reliability than temporal variables [34], likely because they are directly derived from continuous sensor signals, whereas temporal measures rely on event detection. Overall, these findings underscore the need for standardized sensor fixation and cautious use of integrated spatial metrics when employing mobile devices and depth sensors in locomotor assessment of older adults.

Only about half of the studies provided measurement error, such as SEM and MDC, which limited the assessment of whether observed differences reflect true change or fall within the bounds of random measurement error. Among those that reported measurement error, none defined a MIC, which prevents evaluation of whether the reported MDC values correspond to clinically meaningful change [55]. Therefore, the rating results of measurement error for relevant studies remained indeterminate. Future studies should systematically report SEM, MDC, and MIC to enable robust evaluation of measurement performance and to support meaningful clinical interpretation.

Most studies showed acceptable validity, supporting the use of digital technologies to evaluate locomotor capacity. Regarding criterion validity, smartphone-derived linear acceleration measures showed strong correlations with 3D motion capture systems in static tasks but substantially weaker associations during dynamic conditions [44], highlighting the increased difficulty of accurately capturing complex movement patterns with smartphone sensors. In the studies comparing digital biomarkers with force plates, the correlation coefficients of linear acceleration variables derived from smartphones were generally lower than those of the energy variables from force plates and position variables from balance boards [33,36,41]. This could be attributed to inherent differences in the constructs measured; specifically, smartphones leveraged acceleration signals, whereas force plates and balance boards recorded ground reaction forces and COP.

As for convergent validity, the correlation between IMUs and Timed Up and Go tests was significant but low. This weak correlation may be because Timed Up and Go tests mainly assess rougher motor ability, whereas digital tools can measure balance dynamics and subtle changes more precisely [56]. Furthermore, the correlation between the force plate and clinical tests was generally low and nonsignificant in most cases. Clinical tests can measure functional balance abilities, whereas force plates can directly analyze balance related to reactive postural strategies and sensory integration [40,42].

The results of discriminative validity indicated that device-based measures were valid in differentiating balance performance between younger and older participants, which was consistent with a previous systematic review [57]. This not only reinforces the well-established physiological understanding of age-related declines in balance control, but more importantly, it positions digital biomarkers as sensitive tools for objectively quantifying these subtle yet significant changes associated with the aging process. The devices also distinguished between groups with impaired and normal balance. However, the criteria used to define balance groups varied across studies. This heterogeneity makes it difficult to compare the discriminative ability of different device-based measures.

In this systematic review, we identified several digital biomarkers that demonstrated excellent reliability and validity for assessing locomotor capacity in older adults. Aligning with previous studies [58,59], COP parameters, including COP displacement, mean frequency anteroposterior (AP), COP sway AP, and COP sway mediolateral (ML), measured from balance boards and force plates during static standing tasks, exhibited strong reliability and validity for assessing balance. Furthermore, RMS acceleration, derived from smartphones, was also identified as a reliable and valid indicator of static balance. It has been frequently utilized in portable sensor-based balance assessments [33,37,39]. Beyond static balance measures, dynamic balance indicators such as walking speed, mean step time, and mean step length showed great reliability and validity for assessing dynamic balance. These gait-related biomarkers were also used to enhance the discrimination of gait disturbances and disease severity in individuals with Parkinson disease [56]. Additionally, total movement distance and step number in the 20-second stepping test measured by infrared depth sensors exhibited strong test-retest reliability and differentially discriminated between high- and low-balance subgroups. These 2 variables appeared to contribute to the differentiation of disordered balance and to suggest directions for future research.

Among studies focused on muscle power, Cerrito et al [31] demonstrated that smartphone-based peak power during sit-to-stand was a robust digital biomarker for evaluating lower-body muscle power in older adults. The use of peak power as a valid indicator of muscle power was further supported by studies employing linear transducers [60]. Furthermore, sit-to-stand muscle power was found to be a stronger predictor of falls, fractures, and physical independence than grip strength [61,62].

Integrating multiple variables can address multidimensional deficits that single measures miss [63] and has shown utility in frailty classification [64]. There is a growing trend toward constructing composite indices for balance impairment. Although 2 studies [34,37] integrating IMU-derived variables reported encouraging discriminative performance, these indices remain exploratory given the absence of rigorous validation and external replication. To establish their clinical applicability in balance assessment, future studies should focus on systematic performance evaluation and external validation.

Eleven studies were rated as having doubtful or inadequate methodological quality, leading to inflated or underestimated reliability and validity. Specifically, many studies lacked details about the familiarization process or the rest periods between repeated measurements. These omissions may introduce variation due to participant fatigue, learning effects, or other extraneous factors.

In addition, only 3 included studies reported a sample size greater than 50, and none used a sample size calculation. Small sample sizes may inflate the impact of individual variability, reduce statistical power, and result in unstable estimates of reliability and validity. In heterogeneous older adult populations, insufficient sample sizes further limit the generalizability of findings and increase uncertainty in measurement performance. Future research should prioritize adequately powered studies with predefined sample size considerations tailored to the intended measurement properties.

Only De Groote et al [33] met both good methodological quality and an adequate sample size, but the selected indicators, including RMS acceleration ML, RMS acceleration AP, mean acceleration ML, and mean acceleration AP, did not demonstrate satisfactory reliability and validity. Although many studies have reported good reliability and validity for other parameters, the methodological flaws in these studies necessitate caution when interpreting these findings.

The digital transformation of healthcare in recent years has significantly advanced digital technologies, facilitating the transition of digital biomarkers from research contexts to clinical settings. Nevertheless, incorporating digital technology into healthcare to assess locomotor capacity in older adults comes with its own set of challenges.

One major challenge lies in balancing cost-effectiveness and measurement accuracy when selecting appropriate technologies to assess locomotor capacity. Force plates that measure postural sway by calculating COP are considered the gold standard for assessing balance [65]. However, their high cost and limited portability restrict their widespread clinical application [36]. More affordable and available alternatives, such as smartphones, IMUs, balance boards, and infrared depth sensors, have been suggested to demonstrate acceptable reliability and validity for assessing balance and locomotor performance [66]. In addition, emerging technologies are providing new perspectives for monitoring locomotor function. For example, Teel et al [67] explored the VR-based balance module for detecting persistent balance deficits in clinical concussion care, reporting a sensitivity of 85.7% and a specificity of 87.8%. As the field advances, hybrid technology integration may enable the capture of different aspects of locomotor function, thereby offering a more nuanced and holistic understanding of an individual’s physical health [61]. Manupibul et al [68] demonstrated that the integration of force-sensitive resistance sensors and IMUs provided complementary data, high accuracy, and enhanced performance to estimate the gait parameters.

The successful clinical translation of digital technologies for locomotor capacity requires careful consideration of user experience and usability, particularly ease of use among older adults with digital exclusion [69,70]. Tools or systems with complex interfaces or that require frequent manual adjustments may pose substantial barriers to adoption, compliance, and long-term use in older populations. Consistent findings from qualitative studies highlight ease of use as a key facilitator of technology adoption in this population [71,72]. Current research primarily involved participants performing specific, structured tasks under controlled conditions, resulting in potential missing data [33,39]. To achieve clinical effectiveness, future developments in locomotor capacity assessment tools should emphasize unobtrusive sensing, voice-based guidance, and minimal user interaction [73].

Context-specific performance also represents a key challenge for clinical application. Most of the studies included in this review collected data during the performance of predefined tasks under controlled conditions [35,40,44]. While certain digital biomarkers may demonstrate strong reliability and validity in specific tasks, their performance may vary across different tasks or settings. For example, RMS acceleration derived from smartphones in static tasks exhibited excellent criterion validity, but its performance deteriorated when applied to dynamic tasks [44]. Its performance can also be influenced by factors such as sensor placement, sampling frequency, and configured range [74]. Furthermore, the lack of standardized definitions and measurement protocols for these biomarkers complicates their reproducibility and generalization across studies.

To facilitate clinical application, interpretability represents an important consideration for future research. Interpretability can be considered from two perspectives: the clinical meaningfulness of digital biomarkers and the transparency of algorithms. The interpretability of digital biomarkers depends on how they reflect locomotor capacity, including the establishment of meaningful cut-off points that correspond to declines or improvements in mobility. Minimal clinically important difference is defined as the smallest change in a clinical outcome measure that is considered beneficial and meaningful to patients. By establishing minimal clinically important difference thresholds, clinicians can determine whether changes in digital biomarkers are clinically significant, thereby supporting decision-making [75]. With the integration of digital biomarkers, artificial intelligence-based algorithms may be employed, raising concerns about “black-box” systems, and clinical implementation should prioritize explainable artificial intelligence approaches to improve healthcare professionals’ and patients’ understanding of the measurement mechanism [76].

This systematic review offers a comprehensive overview of digital technologies for assessing locomotor capacity in older adults, providing integrated insights into digital health and physical function assessment. This systematic review has several limitations to note. First, we included only articles published in English, which may introduce publication bias. Second, although our search strategy was intentionally broad, the included studies predominantly focused on balance assessment, with other key components of locomotor capacity remaining underrepresented, reflecting a gap in the existing literature on other locomotor capacity components. Third, the included studies exhibited considerable heterogeneity, with variable study quality and assessment protocols. This variability is due to the emerging and expanding nature of this research field, which has yet to establish optimal assessment protocols and highlights a valuable strength of this review. Fourth, this systematic review predominantly focused on general older adults and excluded populations with specific diseases, such as Parkinson disease. The scope of our findings may be limited, and the generalizability to populations with specific diseases requires further investigation. Finally, this review may not have captured all emerging technologies, such as virtual reality and artificial intelligence, which are rapidly evolving and hold considerable potential for locomotor capacity assessment.

Despite these limitations, the sufficient test-retest reliability and validity of the device-based assessment support its potential use in examining older adults’ balance and muscle power. Several recommendations need to be formulated. First, future studies should enhance methodological quality and adhere to COSMIN terminologies and recommendations [27,77]. Second, researchers should thoroughly evaluate both reliability and validity and compare the relevant measurement properties for various digital tools to facilitate the development of the optimal system. Third, alongside rigorous validation of measurement properties, future research should prioritize developing user-friendly digital tools. Current studies often employ task-specific protocols, limiting their widespread adoption in clinical practice and community settings. Fourth, variability in reliability and validity results indicated that no single device or method can be universally recommended for locomotor capacity assessment at this stage. Further studies need to standardize protocols and assess the robustness of these technologies in diverse real-world settings. Lastly, given the current diversity of technologies, the studies included in this review do not cover all existing tools, such as virtual reality and artificial intelligence applications. Future research is expected to explore the integration of these emerging technologies for a more precise and fine-grained assessment of locomotor capacity.

This review offers an integrated understanding of the potential and limitations of digital technologies for assessing locomotor capacity in older adults, summarizing 51 digital biomarkers with generally acceptable reliability and validity. Unlike previous studies that predominantly targeted specific sensor types or disease-specific populations, our review incorporates diverse technologies and biomarkers across general older populations, offering a novel perspective on the potential for scalable, objective, and remote monitoring of changes in locomotor capacity related to aging. Our findings pave the way for developing digital biomarkers that enable the early detection of subtle declines in locomotor capacity, which is critical for timely intervention. However, the current evidence reveals several limitations, particularly task dependency and limited investigation of key locomotor components. The methodological quality of the included studies was generally doubtful, limiting the immediate clinical applicability of digital assessments. Future studies should expand the evaluation to a broader range of locomotor components and improve methodological quality with predefined sample size calculations. Furthermore, age-friendly assessments with enhanced clinical interpretability should be explored to facilitate clinical implementation. By addressing these challenges, digital biomarkers could become essential tools for enhancing the monitoring and management of older adults’ locomotor capacity, contributing to healthy aging strategies.