Background
Description of the condition
Description of the intervention
Why it is important to do this review
Aims
Methods
Registry of umbrella review protocol
Literature search
Inclusion criteria
Paper selection and data extraction
Author | Number of and year of publication of included studies | Databases Searched | Study Objective | Population | Sample Size | Type of Device | Main Results |
---|---|---|---|---|---|---|---|
Pang et al 2019 [17] | N = 9 (2010–2015) | CINAHL Embase MEDLINE Compendex | To summarise and critically examine evidence regarding the detection of near falls (slips, trips, stumbles, missteps, incorrect weight transfer, or temporary loss of balance) using wearable devices. | Adults (aged > = 18 years of age) | Average per study = 21 participants Total = 192 participants | N = 3 (accelerometer) N = 4 (accelerometer and gyroscope) N = 1 (accelerometer and an Android mobile phone) N = 1 (multiple sensors) | N = 5 (Accuracy/sensitivity and specificity of 97% or greater) N = 3 (Accuracy was improved by increasing the number of wearable devices) N = 2 (Chest and right thigh most accurate location for single device placement) |
Nguyen et al 2018 [13] | N = 24 (2015–2017) | Springerlink Elsevier IEE Xplore Digital Library Multidisciplinary Digital Publishing Institute (MDPI) | To systematically evaluate the use of Internet of Things (IoT) technology, especially in terms of sensing techniques and data processing techniques in performing falls management for supporting older adults to live independently and safely. | Adults (aged > = 18 years of age) | Average per study = 7 participants Total = 170 participants | N = 5 (accelerometer) N = 2 (accelerometer and gyroscope) N = 3 (smartphone) N = 6 (camera or laser) N = 2 (“wearable sensor”) N = 3 (multiple devices) N = 2 (wireless networks) | Wearable devices are effective for falls detection - achieving high specificity, sensitivity, and accuracy. Heterogenous methodology in the included studies make quantitative interpretation difficult. |
Montesinos et al 2018 [10] | N = 13 (2008–2014) | PubMed Embase IEEE Xplore Cochrane Central Registry of Controlled Trials (CENTRAL) World Health Organisation International Clinical Trials Registry Platform | To synthetize the empirical evidence regarding inertial sensor-based falls risk assessment and prediction to identify optimal combination of sensor placement, task and features aiming to support evidence-based design of new studies and real-life applications. | At least 10 participants with an average age of 60 years old or over with no severe cognitive or motor impairment. Studies in which participants were labelled as fallers and non-fallers. | Average per study = 93 Total = 1211 participants | N = 9 (accelerometer) N = 3 (accelerometer and gyroscope) N = 1 (gyroscope) | The statistical analysis of features reported in the 13 shortlisted studies revealed significant, very strong, positive associations in 3 different triads of feature category, task, and sensor placement: • Angular velocity – Walking – Shins • Linear acceleration – Quiet standing – Lower back • Linear acceleration – Stand to sit/Sit to stand – Lower back |
Chaudhuri et al 2014 [16] | N = 57 (2007–2013) | PubMed CINAHL Embase PsycINFO | To systematically assess the current state of design and implementation of fall detection devices. This review also examines the extent to which these devices have been tested in the real world as well as the acceptability of these devices to older adults. | Adults (aged > = 18 years of age) | Information not available | N = 57 (wearable systems) | Most common types of devices: • Systems with device on trunk. Median sensitivity = 97.5% (range 81–100). Median specificity = 96.9% (range 77–100) • Systems involving multiple sensors. Median sensitivity = 93.4% (range 92.5–94.2) and a median specificity of 99.8% (range 99.3–100). • Systems involving devices around arms, hands, ears, or feet had a lower median sensitivity and specificity [81.5% (range 70.4–100) and 83% (range 80–95.7) respectively]. |
Silva de Lima et al 2017 [15] | N = 4 (2005–2015) | PubMed Web of Science databases | To provide an overview of the use of wearable systems to assess freezing of gait (FOG) and falls in Parkinson’s disease with emphasis on device setup and results from validation procedures. | Parkinson disease patients (aged > = 18 years of age) | Average per study = 44 participants Total = 177 participants | N = 2 (accelerometer) N = 1 accelerometer and gyroscope) N = 1 (accelerometer and force sensor) | High specificity (86.4–98.6%) and sensitivity (93.1% only one study) for wearable device detection of falls. |
Rucco et al 2018 [11] | N = 42 (2002–2017) | IEEE Xplore SpringerLink Science Direct PubMed | To provide an overview of the most adopted sensing technologies in these fields, with a focus on the type of sensors (rather than algorithms), their position on the body and the kind of tasks they are used in. | Healthy “aged” population | Average per study = 32 participants Total = 1331 participants | N = 12 (accelerometer) N = 7 (accelerometer and gyroscope) N = 6 (accelerometer and pressure sensors) N = 3 (accelerometer + another device) N = 1 (gyroscope) N = 4 (camera or radar or console) N = 9 (three or more devices) | • Single sensor = 70% use accelerometer • Two sensors = 1) Approaches that combine accelerometer with a pressure sensor (usually in shoes). 2) Approaches that use accelerometer and gyroscope sensors (usually on same electronic board). • Three or more sensors = other sensing technology used (magnetometer, camera, EMG). • Sensor placement = mainly on the trunk. Second most likely position is foot or leg (about 30%). |
Sun et al 2018 [14] | N = 22 (2011–2017) | PubMed Web of Science Cochrane Library CINAHL | To systematically evaluate the use of technology in performing fall risk assessments, and more specifically, to evaluate the test, sensor, and algorithm effectiveness on predicting and/or discriminating older adult fallers from non-fallers. | Older adults (Aged > 60 years of age) | Average per study = 86 participants Total = 1896 participants | N = 11 (accelerometer) N = 4 (accelerometer and gyroscope) N = 4 (console) N = 1 (laser) N = 2 (accelerometer and pressure sensor) | A diverse range of diagnostic performance was observed (Accuracy: 47.9–100%, Sensitivity: 16.7–100%, Specificity: 40–100%, AUC 0.65–0.89) for wearable device detection of falls. |
Data synthesis
Risk of Bias and relative quality assessment
Study | Q1 | Q2 | Q3 | Q4 | Q5 | Q6 | Q7 | Q8 | Q9 | Q10 | Q11 | Q12 | Q13 | Q14 | Q15 | Q16 | Overall Quality of Study |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Pang et al. 2019 [17] | Y | Y | N | Y | N | Y | N | Y | PY | N | N/A | N/A | Y | Y | N/A | Y | Moderate |
Nguyen et al. 2018 [13] | Y | N | N | Y | N | Y | N | PY | N | N | N/A | N/A | N | N | N/A | Y | Critically low |
Montesinos et al. 2018 [10] | Y | N | Y | Y | N | Y | Y | Y | Y | N | Y | Y | Y | Y | Y | Y | Moderate |
Chaudhuri et al. 2014 [16] | Y | N | N | Y | Y | N | N | N | Y | N | N/A | N/A | N | N | N/A | N | Low |
Silva de Lima et al. 2017 [15] | Y | N | N | Y | N | N | N | N | N | N | N/A | N/A | N | N | N/A | Y | Critically low |
Rucco et al. 2018 [11] | Y | N | N | Y | N | N | N | Y | N | N | N/A | N/A | N | Y | N/A | Y | Critically low |
Sun et al. 2018 [14] | Y | N | Y | Y | N | N | N | PY | N | N | N/A | N/A | N | Y | N/A | Y | Critically low |
Results
Studies included
Characteristics of included systematic reviews
Types of wearable devices in included systematic reviews
Wearable devices for falls detection and their effectiveness
Quality appraisal methods of studies included within included systematic reviews
Quality appraisal of included systematic reviews
Discussion
Summary of evidence
-
Angular velocity – walking – shins.
-
Linear acceleration – quiet standing – lower back
-
Linear acceleration – stand to sit/sit to stand – lower back
Strengths of this review
Limitations of this review
Implications for future research
-
Are wearable devices as effective as proven in previous studies if tested in “real world” settings with a large sample size of older adults?
-
What is the most effective system design that older adults will accept for use in daily living?
-
Can wearable devices be used to enable alerts of deteriorating balance control?
-
How, practically, could wearable devices be integrated with a comprehensive falls risk assessment?
-
How can the gap between clinical functionality and user experience of these devices be improved?
-
An effective, validated, tool for evaluating wearable devices for falls detection that can be replicated in future high-quality studies.
-
What are the most effective algorithms to use combined with these wearable technologies?
-
Is there potential for these devices to be used in different types of falls experienced by people with stroke, MS, age-related frailty, and other conditions associated with ageing?