Aging renders people vulnerable to gait deviations that impair efficient walking and limits the likelihood of achieving walking targets for health promotion. Physical activity guidelines for seniors recommend a target of 150 minutes of moderate intensity exercise accumulated over 1 week in bouts of 10 minutes [1]. Walking is the most practical exercise as it requires no equipment, or no specialized environment [2], and produces many physical and cognitive health benefits from the mental stimulation of exploring new avenues or neighborhoods [3]. Maintaining a level of physical activity is also critical to prevent secondary health conditions including cardiovascular disease, osteoporosis, obesity, and diabetes [4]. Despite capacity to walk at a health-promoting pace when tested clinically, it is rare for the North American seniors to do this in the real world for more than a few minutes a day [5,6]. It is hard to sustain walking without the capacity for an optimal stepping pattern indicating that quality drives quantity.

Reasons for failure to use walking capacity to achieve health-promoting walking targets include fear of falling or age-related gait abnormalities [6]. These are known to cascade into a slow, unstable, shuffling pattern that increases the work of walking, fatigue, and risk of falls and hip fracture [7]. There is a considerable evidence on how to improve seniors’ gait [8,9], and evidence shows that gait training is effective in improving gait pattern [10] but effects abate with cessation of training [11]. Hence, gait training alone will not translate into the sustained behavioral change needed for physical activity guidelines to be met.

During gait training, therapists use many of verbal and visual cues to emphasize stepping with heel first. This simple strategy changes posture from stooped to upright, lengthens the stride, stimulates pelvic and trunk rotation, and facilitates arm swing [12]. However, once verbal cueing ceases, patients frequently revert to an inefficient foot-flat gait.

For walking to become more normalized, people must relearn the motor sequences of good walking and develop the needed adjuncts to efficient walking: flexibility, strength, power, core stability, balance, and trunk rotation indicated by arm swing. Therapy can work on the adjuncts, but motor learning requires instruction, practice, and feedback. The 2013 review by Sigrist et al [13] frames motor learning as a lasting change of motor performance caused by training in which the parameters of a motor program are developed, and there is a gradual reduction of the variability in the newly developed motor program stimulated by sensory feedback loops. The phenomenon underlying motor learning is neural plasticity [14]. A 2014 review of this topic indicates that motor learning takes place with active practice of a skill and that this activity-dependent neural plasticity can be induced by both lengthy-extensive and brief-intensive practice [14]. The literature supports the benefit of augmented or extrinsic feedback for motor learning [14]. In particular, sonification for correct movement sequences has been shown to enhance motor learning in athletes [13,15].

It is well established that knowledge of performance is strongly associated with skill acquisition and motor learning compared with knowledge of results [16,17]. Technology is poised to provide this feedback. For walking, there are emerging technologies that use footwear-based gait monitoring systems [18]. None of the reviewed technologies provided real-time feedback, and all needed considerable data processing to produce usable information on walking performance. There is evidence that gait can be modified in response to real-time auditory feedback, but currently, no technology provides this type of feedback.

These principles guided the development of the Heel2Toe sensor, a biofeedback device that provides auditory feedback for each good step, in which the heel strikes first. The aim of this project was to bridge this feedback gap that exists outside clinical settings and equip seniors to practice correct gait at convenience. The hardware and algorithm underlying generation of auditory feedback from the Heel2Toe sensor are described elsewhere [19,20]. Briefly, Heel2Toe is a modification of an off-the-shelf device from the Shimmer Motion Development Kit. The sensor is a combination of three-axis accelerometer, a three-axis gyroscope, and a microcontroller. The algorithm detects the rate of angular velocity (AV) in sagittal plane at the ankle joint and provides an auditory beep when the rate of foot deceleration after heel strike crosses a threshold. Pilot work on Heel2Toe has demonstrated that it is highly accurate to detect good steps in clinical setting [19,20]. Starting the gait cycle with a strong heel strike lengthens the stride and changes posture from stooped to upright [12,21], indicating the value of focusing on AV as a treatment target.

The aim of this study was to contribute evidence toward the feasibility and efficacy potential for home use of theHeel2Toe sensor that provides real-time feedback for good heel strike when walking. Specifically, the objectives were (1) to identify the extent of the immediate response to the feedback, carry over when walking without feedback, and peak response to feedback and (2) to identify pleasures and challenges in using the feedback sensor.

A pre-post study design, with five sessions of training over 2 weeks, was employed to estimate the efficacy potential and identify feasibility issues of the Heel2Toe sensor when deployed for walking in the community.

A purposive sample of six people, four women and two men, over the age of 70 years, was identified from geriatric services at the Montreal General Hospital from September to October 2017. Participants were identified by a geriatrician or other health care professionals and included if they reported no limitation in walking without an aid and no cognitive impairments. The participants were selected to have a range of walking capacity from very limited to functional. Ethical approval was obtained from the Ethics Review Board of McGill University, Health Centre Research Institute.

Participants were assessed on physical performance tests and self-report measures. Physical performance tests included gait speed, 30-second chair stand, and 2-minute walk without and with auditory feedback. The self-report questionnaires included single item on perceived walking speed, lower extremity function scale (LEFS), life space mobility scale, and activity-specific balance confidence scale. LEFS scoring is based on fit to the Rasch Model, and therefore, not all items have to be administered to derive a legitimate total score [22]. Posttraining outcomes additionally included questions about the system usability and a semistructured interview on challenges and pleasures of using the Heel2Toe sensor. The interview was conducted separately with each participant.

The intervention involved a therapist visiting a participant’s residence to provide walking training with the Heel2Toe sensor, for five sessions over 2 weeks (Figure 1). The training involved walking in the participant’s neighborhood with the sensor. Care was taken that they walked on an obstacle-free path. The duration of the training was determined by the participants themselves based on interest and tolerance. On each training day, participants were instructed to walk for at least 15 minutes with the sensor at a comfortable pace and taking rests when needed. The training was accompanied with home exercises targeting flexibility and strength at ankle, knee, and hip joints with a particular focus on core strength and trunk rotation. At the end of the training, a semistructured interview was conducted with all participants.

The gait signals recorded with the Heel2Toe sensor were analyzed using MATLAB (MATLAB and Statistics Toolbox Release 2017b, The MathWorks, Inc, Natick, Massachusetts). The gait parameters extracted for each person over the entire walking period were proportion of good steps (%), total walking time (seconds), and average cadence (steps per minute). AV (degrees per second) in sagittal plane at ankle joint during heel strike was extracted for each step and averaged over the walking duration yielding mean, SD, and coefficient of variation (an indicator of consistency of stepping).

We found that the Heel2Toe sensor was feasible to use in the community setting with older adults and that they improved on gait quality after the planned five training sessions, averaging 73 minutes (range: 43-114 minutes) in total. The proportion of good steps and AV improved without any detriment to cadence. All six participants showed longer duration of time spent in walking from the initial training days. However, the most dramatic effect was seen for duration of walking bouts which frequently exceeded 10 minutes (Figure 2) such that most (five of six; Table 2) participants would now be capable of meeting the Canadian Physical Activity Guidelines of 150 minutes of moderate to vigorous activity (required walking cadence ≥100 steps per minute) per week in bouts of 10 min.

Posttraining, five of six participants showed a reduction in the coefficient of variation of AV, a parameter indicating inconsistency of stepping pattern. Before training, the coefficient of variation ranged from 23% to 59%. Previous studies have shown a higher coefficient of variation in step width, and stance and stride time among older adults is associated with increased occurrence of falls [26-28], with the suggestion that a treatment target is to reduce the coefficient of variation with exercise interventions. After 5 days of training, the range was 9% to 49%.

We purposely chose a sample of people diverse in physical function. In all, two people were quite frail (A and B). Participant A was severely limited in mobility (Table 1), yet he improved on the proportion of good steps and degree and consistency of AV (Table 2 and Figure 2). Participant B also improved on these parameters. The most functional walker, participant D, showed no change as she was high on all parameters but enjoyed the experience of the sensor and could see how it would prevent deterioration. In a definitive trial, these data can be used to optimally select people for intervention.

How did the sensor achieve these outcomes? One hypothesis is that the auditory feedback acts as a positive reinforcement to a rhythmic stepping pattern. With symmetrical walking, each good step produces a beat that is repeated with periodicity. To produce the rhythmic pattern (the beat), the participants modified their stepping pattern to maintain the rhythm. In the long run, auditory cues could enhance cortical motor excitability. This has previously been studied with upper limb movements and walking tasks that required persons synchronizing to an external auditory cue [29]. The underlying basis of auditory motor synchronization is that brain poses anticipatory tendency for a rhythm, and this anticipation guides subsequent movements [29].

The Heel2Toe sensor provides direct positive auditory feedback, which could be perceived as rewarding stimulating neural plasticity and increasing the pleasure in walking, stimulating behavior change. Ultimately, the aim is to improve health-promoting walking rather than just functional walking, so that older people can derive pleasure and health benefit from walking. The sensor is not designed to be worn all the time but to be worn to practice optimal walking with the aim that this would carry over into other walking activities. As it is linked to a smartphone and the sensor is very small (size of a matchbox), it could be worn for longer periods of time.

Fear of falling and age- or illness-related changes co-occur in most seniors and can induce an inefficient and dangerous gait pattern [30,31]. To normalize walking, people must relearn motor sequences of good walking and develop needed adjuncts to efficient walking: flexibility, strength, power, core stability, balance, and arm swing. Therapy targets adjuncts but motor learning requires instruction, practice, and feedback. Motor learning is framed as a lasting change of performance occurring with training in which parameters of a motor program are developed and consolidated. Early on, formation of the motor program of the to-be-learned task can occur rapidly but demands high levels of attention. Later, the motor program is refined, improving error detection or correction mechanisms, reducing movement variability. Finally, movements become highly automatized, skilled, and consistent, and the motor program is now relatively permanent [32].

The phenomenon underlying motor learning is mostly because of neural plasticity [33]. A review of this topic [33] indicates that motor learning takes place with active practice of a skill and that this activity-dependent neural plasticity can be induced by both lengthy-extensive and brief-intensive practice. The literature supports the benefit of augmented feedback for motor learning. In particular, sonification for correct movement sequences has been shown to enhance motor learning in elite athletes [34] but is less useful for novices who have no idea of the correct movement. Walking is a natural way to get about [35], and as older persons are not novices to walking but have lost the expertise with age, their walking pattern should respond to auditory feedback. This type of positive feedback has been shown effective in the short term to improve gait pattern in people poststroke [36]. It is superior to auditory alarms signaling incorrect movements as feedback because good movement is more motivating [34].

This solution to poor gait is unique in that there is positive reinforcement, in real time, which stimulates motor learning of correct gait. The Heel2Toe sensor provides information in real time, in other words, knowledge of performance and not just knowledge of after-the-fact results, which is provided by most other technologies in the field today. This is a completely novel and original approach to gait enhancement. There have been other approaches to monitor step counts, but these have not attempted to improve gait quality. The review of the literature conducted by our team did not find any study focusing on feedback related to gait quality and ankle kinematics.

Finally, debriefing interviews suggested readiness of seniors to adopt technology as long as it is simple and user friendly. This project is timely and relevant to increasing the proportion of older population and builds upon the potential of technology to stimulate innovation, thereby advancing Canadian economic and social development. An increasing proportion of older adults use smartphones [37,38], and this proportion is likely to increase as technologically savvy cohorts age.

This sensor could be on the foot of every person who needs to maintain or improve optimal gait. By formally practicing gait improvement with positive auditory feedback, people could develop the habit of walking better leading to walking more often and for longer.

Through the use of the Heel2Toe device, every step becomes therapeutic, engaging large muscle groups, which improves peripheral and core muscle strength and through this improves balance, allows the person to walk at a faster pace. Our data also support changes to gait consistency (lower coefficient of variation with training), making walking more rhythmical, which, in the long run, is more sustainable [35].

The sensor is in development, and refinements to the algorithm will be made, such as to provide different thresholds for the feedback to occur (low, medium, and high AV). An instructional manual and video are in production to optimize the participants’ capacity to use the Heel2Toe sensor. The plan is to develop a full-scale trial, now that there are some data that people can change their gait with the device.

This was a very small study focusing on proof-of-concept only. On the basis of the results that short-term intensive training with positive auditory feedback produced changes in gait quality, a full pilot study is warranted including the second motor learning phase and longer-term practice to estimate sustainability.

The results of this study have future implications in exploring the neural basis of auditory-motor synchronization during walking, application of motor learning principles to enhance walking performance, and technology design of wearable sensors for older adults. Understanding the neural basis of auditory motor synchronization will help design interventions to use auditory feedback to improve walking symmetry. The application of motor learning principles to enhance walking performance based on movement-generated auditory feedback and long-term effects on skill acquisition is an area yet to be explored. Debriefing interviews conducted after the intervention concluded that an optimal wearable device for seniors needs to be simple and easy to use, provide real-time meaningful feedback, have a software program that requires minimal preprocessing (zero effort) before use, and have the option for technical support or supervision from a rehabilitation professional [39].