Normally functioning human hands allow everyday participation in self-care, work, and leisure activities that involve precise grip and object manipulation [1]. Specifically, daily activities and fine motor tasks require finger dexterity, thumb-finger opposition, and hand opening-closing, which adapt to task requirements, including those needed to navigate the digital world. [2] Unfortunately, chronic disorders of the central nervous system (CNS) can impair hand function even during the early stages of the disease [3]. Damage to the CNS, including the spinal cord, can result in tremor, spasticity, sensory loss, weakness, and coordination loss in the upper limbs, which can negatively impact the ability to adapt to task requirements, thus limiting independence in activities of daily living (ADL) and quality of life [3]. For example, most individuals with Parkinson disease (PD) develop hand tremors over the course of the disorder, leading to difficulty with precise finger and hand movements [4]. In addition, ischemic strokes occur most commonly in the cortical regions supplied by the middle cerebral artery [5], affecting areas of the motor and sensory cortices responsible for the fine motor activity of the hands [6]. In these disorders and others, evaluating hand function at regular intervals can detect changes signaling neurological decline, or monitor response to disease-modifying therapies, symptomatic therapies, or rehabilitation.

Although assessments of hand function are routinely performed in clinics, clinicians have an increasing interest in deploying tools to measure hand function remotely. In-home remote monitoring of function, in general, provides benefits to patients by increasing convenience, reducing travel, and providing the ability to capture data more frequently. Over the past decade, many studies have examined remote monitoring devices in neurological and nonneurological populations [7,8]. For example, in multiple sclerosis (MS), studies have shown that continuous remote monitoring of ambulatory step count can capture—and even predict—changes in MS-related disability and can serve as a longitudinal outcome measure for targeted interventions [9,10]. To date, reviews have mainly focused on lower extremity function or overall physical activity [11]; in fact, the methodological discrepancies in remote device use and reporting regarding hand function have yielded conflicting results in terms of validity, reliability, and ease of clinical use.

In this systematic review, we evaluate the existing evidence regarding remote assessment devices for hand function in populations with chronic CNS disorders. We specifically examine evidence of validity, reliability, and feasibility for each domain of hand function and the stage of development of the assessments. Our findings are expected to facilitate ready implementation of remote assessment of hand function in prevalent neurological disorders.

This review was structured using the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [12] framework. Studies were included based on the following criteria: (1) participants had chronic neurological pathologies of the CNS, (2) participants were aged ≥18 years, (3) the studies were peer reviewed and original, (4) the studies were designed to objectively assess hand function, and (5) the assessments were deployable remotely (ie, outside of traditional in-person clinical settings). Studies were excluded if they were (1) conducted in participants with orthopedic impairments of the wrist or hand, (2) conducted in nonhuman primates, (3) designed as an intervention to improve an aspect of hand function (as the intent was to focus on assessment tools rather than a change of function), or (4) not published in English.

A literature search was performed using the following databases: PubMed and MEDLINE, CINAHL, Web of Science, and Embase. The search was conducted using both Medical Subject Heading terms and the following keywords independently and in combination: remote, assessment, outcome, test, measurement, hand, upper extremity, arm, and function. Independently, 2 researchers (AG and WYH) assessed articles for relevance and adherence to the eligibility criteria. Studies were recursively searched to identify cited and cited-by articles.

To evaluate the methodological quality of the included studies, we used the National Institutes of Health quality assessment for observational cohort and cross-sectional studies [13]. Each study was evaluated according to 8 criteria. The overall study quality was assessed as good (>5 criteria met), fair (4-5 criteria met), or poor (<5 criteria met).

The data were extracted (AG) and checked (WYH); discrepancies were resolved through discussion with the senior author (RB). The variables of interest included participant demographics, study design and duration, device type and modality, disease-specific severity levels, comparison assessments, and stage of development and implementation (to understand whether assessments were currently available for use). Participant satisfaction with the study protocol and assessment and time taken to complete the novel assessment were extracted when available. Extracted statistical data included concurrent validity (defined as the comparison between a new test and a well-established one [14]) and reliability (defined as a measure of stability or consistency [15]).

The selected studies evaluated many variables relating to hand function. To compare the most salient domains across studies, we classified assessments into the following hand function domains based on the Functional Repertoire of the Hand established by the American Journal of Occupational Therapy [16]: (1) finger tapping, which is the speed and accuracy of finger taps onto a prespecified target; (2) whole hand grasp, which is the range of motion and coordination of full hand movement; (3) pincer grasp, which is the range of motion and coordination of thumb to index finger movement; (4) hand tremor, which is the quantification of tremor distal to the wrist at rest; (5) reaction time, which is the time taken to respond to a predetermined stimulus using only fingers; (6) pinch and grip strength, which is the quantification of the maximum pinching and gripping strength; (7) finger dexterity, which is the in-hand manipulation of an object; (8) handwriting, which is the clarity and accuracy in drawing or writing; (9) ADL, encompassing tasks required for self-care independence [17]; and (10) instrumental ADL (IADL), encompassing tasks required for household or community-level independence [18].

A search of databases in June 2021 identified 1295 studies, and 33 additional studies were identified through recursive searches. After title and abstract screening and removal of duplicates, 9.42% (122/1295) of studies remained, and the full texts were assessed for eligibility based on the inclusion and exclusion criteria. Approximately 41% (50/122) of full-text studies were excluded for not meeting the inclusion criteria. The final 74 studies were confirmed by a second reviewer (WYH) to have met all eligibility criteria. The PRISMA diagram of the search process is outlined in Figure 1, and individual studies are summarized in Multimedia Appendix 1 [19-90]. Of the 74 studies reviewed, 49 (66%) were rated good in terms of overall methodological quality, 14 (19%) were rated fair, and 9 (12%) were rated poor. Study quality is summarized in Multimedia Appendix 2 [19-90].

Across the included studies, 3 different modalities of assessment devices were used, summarized in Multimedia Appendix 1. The most frequently used assessment was an external device specific to hand assessment, with the most common types being wrist-worn accelerometers [19-37] and specialized keyboards [38-47]. These designated external devices allowed the collection of information on reaction time, finger tapping speed, and finger dexterity. Although many study authors noted that their external devices were able to capture granular, specific data, many devices were developed under proprietary agreements and are not currently commercially available. The second most common type of assessment was generic smartphone- or tablet-based electronic devices adapted for hand assessment [48-59] or suites of assessments [60-66]. These assessments included an app designed to test finger tapping speed and the accuracy of drawing and tracing various shapes. Such apps facilitated the gathering of data on specific hand function domains at a relatively low cost for people who already had these electronic devices. Finally, 4% (3/74) of studies used telerehabilitation platforms to validate remote administration of well-established in-clinic assessments [67-69]. For example, Amano et al [69] validated the administration of the Fugl-Meyer Assessment and Action Research Arm Test (ARAT) via telehealth platforms, allowing clinical researchers to gather standardized outcome data through secure telehealth tools.

Most of the included studies (51/74, 69%) performed same-day, cross-sectional validation experiments where participants completed novel and comparative assessments at the same time point. However, 28% (21/74) of studies [23,25,36,37,42,46,47,61,63,64,66,70-80] remotely monitored participants’ hand function longitudinally. The duration of the remote monitoring period was 3 days [37] to 3 years [77]. Participant retention and adherence were reported by 5% (4/74) studies [61,66,75,76], all of which had >90% participant retention.

The included studies targeted 5 populations of patients with neurological conditions. Most studies (57/74, 77%) included individuals with PD [37,45,65,67,68,70,73,81]. Other populations evaluated were those with stroke (9/74, 12%) [71,72,82] and MS (6/74, 8%) [59,66,91]. Neurological conditions designated as spinal cord injury [83] and amyotrophic lateral sclerosis [47] were described in 1% (1/74) of studies each.

Most included studies evaluated individuals with mild to moderate disease severity on average, as graded by established disease-specific metrics (eg, the Movement Disorder Society–Unified Parkinson’s Disease Rating Scale [MDS-UPDRS] and the Expanded Disability Status Scale for people with MS) [37,45,59,65,67,69,70,73,82]. Only 8% (6/74) of studies specified the inclusion criteria to limit recruitment to participants with mild to moderate disease severity [37,41,53,58,63,71].

The sample sizes of studies varied between 1 (case study) [26] and 495 participants [66] in the experimental groups. Most studies (41/74, 55%) included control groups of healthy individuals or those with nonneurological conditions in determining the discriminant validity of the assessments (Multimedia Appendix 1).

Validity data were reported by 73% (54/74) of heterogeneous studies for comparison with well-established in-clinic assessments (Table 1). Approximately 12% (9/74) of studies examining external devices reported high, statistically significant correlations with well-established assessments [19,20,47,50,52,72,73,83,91]. In addition, 8% (6/74) of studies using smartphone assessments [28,49,52,66,79,84] and 1% (1/74) of studies using telerehabilitation [69] found moderate to high, statistically significant correlations with well-established assessments.

Of the 74 studies, 15 (20%) heterogeneous studies reported reliability statistics; 2 (3%) telerehabilitation assessments [68,69] revealed a high, statistically significant interrater reliability; and 1 (1%) external device assessment [76] revealed a high, although statistically insignificant reliability.

In populations with PD, 9% (7/74) of studies reported participant acceptability and usability of assessments. Albani et al [73] found that participants rated the hand gesture–based tracking system 5.9/7 on a poststudy usability questionnaire, indicating ease of use, high interface quality, and usefulness. In 4% (3/74) of studies [24,30,37], participants using wearable sensors to monitor hand tremors and finger tapping found the devices comfortable and easy to use. Both Goetz et al [75] and Ferreira et al [23] reported >80% of participant satisfaction with external devices to examine hand tremors. Mitsi et al [65] found that 76% of participants using a tablet-based assessment for finger tapping [65] and reaction time found it easy to use, with an additional 63% reporting willingness to use it long-term to monitor disease activity.

In populations with stroke, Burdea et al [71] asked both participants and caregivers to provide feedback on their video game–like assessment and intervention using a 5-point study-specific Likert scale (higher scores indicating statement agreement). Participants reported that the device was moderately easy to use (mean score 3.1/5.0), that they would encourage others to use it (mean score 4.3/5.0), and that they liked the system overall (mean score 4.2/5.0). However, participants encountered some technical difficulties during use (mean score 2.2/5.0). Caregivers also found the device setup appropriate for the home environment and easy to use (mean score 3.5/5.0).

In people with MS, Dubuisson et al [91] reported that 66.7% of participants preferred the portable in-home 9HPT in comparison with the standard in-clinic version.

Only 3% (2/74) of studies reported safety data [37,68]. Hoffman et al [68] reported that participants who received assessment via telerehabilitation were accompanied by a researcher to ensure safety. Boroojerdi et al [37] used a wearable patch and reported no adverse skin reactions at the application site or device malfunction. Adverse events were not reported in any of the included studies.

As the assessments in this review were novel, the availability for clinical implementation varied. Most studies (44/74, 59%) evaluated assessments requiring specialized equipment for implementation. These devices included specialized cameras, wearable devices, electromyography, and specialized keyboards. Although not an application, the cardboard 9HPT developed by Dubuisson et al [91] was designed specifically to be environmentally friendly, cost-effective, and used by patients at home. The remaining external devices evaluated in this review were designated as developmental, with a need for subsequent safety and prospective studies on usability before clinical use.

Approximately 3% (2/74) of studies using telerehabilitation methods required videoconferencing devices and a stable internet connection for both providers and patients for implementation. However, although Hoffman et al [68] similarly used telerehabilitation methods, their protocol required participants to use clinical equipment during in-home assessments (eg, a hand dynamometer and the 9HPT), potentially limiting widespread implementation.

A smartphone or tablet-based application was used in 27% (20/74) of studies to administer assessments. The FLOODLIGHT application studied by Creagh et al [59] is currently available for download for iOS and Android devices. The remaining applications were study-specific developments but, given compatible devices and secure broadband internet connection availability, have limited barriers to implementation.

The purpose of this review was to systematically gather available literature on remote assessments for monitoring hand function in people with central, chronic, and neurological diseases. The search yielded 74 studies that met the inclusion criteria, and 71 unique assessments were examined for validity, reliability, and clinical implementation. A wide variety of metrics were collected on a number of hand function domains, including the amplitude of finger tapping, finger dexterity, hand tremor, and ADL independence. Altogether, the studies provide a number of insights; however, to date, no single tool, or combination of tools, validly and reliably captures hand function across these major neurological conditions.

Many of the studies were of good quality, and several study characteristics were found to enhance their quality. Including controls with nonneurological conditions as a comparison, when available, helped demonstrate the discriminant validity of the novel assessments examined. Most studies included participants with lower disability status, which likely allowed for more dynamic testing of hand function domains. Unfortunately, most of the included studies reported statistically insignificant associations with standard in-clinic metrics. As prior literature suggests that traditional in-clinic assessments have limited granularity for upper limb function in populations with neurological conditions, differences between the novel assessments and these traditional in-clinic tests could indicate that the new tools capture additional aspects of function (eg, quantifying pincer grasp) relative to the traditional in-clinic assessments or vice versa. In addition, few studies reported reliability, especially interrater reliability, suggesting the need for more research and that the included tools remain primarily in the development phase.

The most commonly assessed hand function domain was finger tapping speed, with moderate to high agreement across comparison assessments. The finger tapping test is a valid and reliable measure of bradykinesia in PD [94] and a predictor of ADL independence in acute stroke [95]. It is relatively simple to quantify finger tapping in-clinic or via a smartphone or tablet app by counting the number of finger taps within a specific time frame. Although overall construct validity and participant satisfaction were high, further work in other hand function domains will help determine the most salient predictors of ADL independence and response to treatment and intervention.

This review highlights important aspects of the feasibility of remote evaluations. Participant and caregiver satisfaction, when reported, were moderate to high for these technologically innovative assessments. This suggests that participants found the novel assessments easy to use and effective in evaluating their hand function despite being nontraditional. Further, 28% (21/74) of the included studies demonstrated the feasibility of remotely monitoring hand function over multiple days. This is a key finding, as long-term monitoring of hand function in a patient’s natural environment has the potential to identify changes in real time, allowing for timely intervention modifications.

Regarding patient safety, although the included assessments were noninvasive and posed a relatively low safety risk, ensuring the secure transfer of data, especially with internet-based communication (eg, telerehabilitation and smartphone or tablet-based apps) between patient and clinician, is critical to confidentiality and Health Insurance Portability Accountability Act compliance. Future studies should report on data storage and encryption methodologies.

The assessments evaluated were in varying stages of development and implementation. The most readily implementable types of assessment were those using telerehabilitation or smartphone- or tablet-based apps. According to 2019 data, 85% of Americans own a smartphone, and 93% use the internet regularly, of whom 75% use a home high-speed broadband network [96]. Given these statistics, telerehabilitation and application-based assessments, if interoperable across devices, might be relatively accessible for most patients. Lower costs could make clinical implementation less of a challenge. Furthermore, with no specialized devices to purchase or distribute to patients, clinics could similarly benefit from these cost-effective measures.

A major limitation of this review is the heterogeneity of hand function domains evaluated, which, when compounded with the methodological variability (in comparison assessments, inclusion criteria, and statistical approaches), made it difficult to compare the various tools. Future studies that include more homogeneous patient populations and standardized reporting of correlation coefficients with comparison assessments will facilitate analysis across domains and assessment types. A second limitation was the paucity of studies conducting repeated trials of the assessments, limiting the identification of any practice effects with use of a new device. In repeated trials of smartphone-based assessments, performance improved in the first 10 trials because of a practice effect, followed by a narrowing of variance as the practice effect waned and familiarity with the assessment increased [97]. Follow-up studies should include repeated trials, preferably over multiple days, to capture these effects and fluctuations in disease progression. Third, the effect of confounding variables (eg, disease-modifying therapies, age, and disease duration) was infrequently described in validity statistics; the generalizability of this review should proceed with caution. Fourth, all tools included require active participant engagement as opposed to passive monitoring (eg, collecting data on dexterity as a participant types to complete a survey). Passive monitoring may be able to capture similar metrics with a reduced participant time burden. Finally, we may have missed relevant studies published in non-English languages.

This review suggests that remote assessments can be valid and reliable tools for measuring hand function impairments in chronic neurological diseases and that doing so is clinically feasible and acceptable to patients. In the past decade, personal smartphone and computer ownership have become commonplace; with it, patients and health care providers are able to communicate in real time, opening new avenues for care delivery and disease monitoring. We highlight the current potential to implement remote assessments via telerehabilitation and smartphone- or tablet-based apps. As interventions for ambulation and lower extremity function become increasingly robust, these methods will allow clinicians to reliably assess multiple domains of hand function to monitor disease progression and response to interventions.