Analysis of gait kinematics to determine the effect of manipulating the appearance of stairs to improve safety

Participants

The characteristics of the participants for each experiment are presented in Table 1. The tenets of the Declaration of Helsinki54 were observed, the experiments gained institutional ethics approval and all participants gave written, informed consent. Community-dwelling, older participants were recruited for experiments 1 and 3 from a group of 50 retired volunteers who regularly sit as patients in undergraduate optometry clinics at the University of Bradford’s Eye Clinic and from a group of 200 older people on the mailing list of Bradford Older People’s Forum. The younger patients in experiment 2 were recruited from the student population of the University of Bradford and the sample size of eight was based on a previous study of stair descent (in older people and with a less dense cataract simulation). 41

All participants were independently mobile, able to follow simple instructions and, according to self-report, suffered no significant neurological, musculoskeletal or cardiovascular disorders that could interfere with balance control or stepping. Those with vestibular disturbances or diabetes, or a history of falling in the previous year, were excluded, as were those taking medications that could affect balance or vision. Participants had normal healthy eyes, determined by a full eye examination, including ocular screening using slit-lamp biomicroscopy, tonometry, indirect ophthalmoscopy, central visual field screening and binocular vision assessment. Corrected visual acuity was equal to or better than the logMAR score of 0.1 (Snellen score of 6/7.5 or decimal Snellen 0.80) in either eye. All participants engaged in light to moderate physical activities for at least 30 minutes, 5 days a week, including gardening, light housework and dancing. Participants wore their own shorts, t-shirts and low-heeled shoes during the trials and were asked to refrain from alcohol intake from the evening prior to testing. Participants wore their own spectacles if they were typically worn when walking outdoors for ecological validity reasons. This included bifocal or varifocal spectacles, which would lead to gait adaptations and increased variability of adaptive gait. 39,46 Three participants wore bifocals and three wore varifocals in experiment 1, and four participants wore varifocals and three wore bifocals in experiment 3.

Sample size calculations

To estimate required sample size, we used a paired t-test calculation. Although the main analysis will incorporate a repeated measures design, the effective power is likely to be similar to that using a simple t-test but will depend on unknown parameters that are difficult to estimate prospectively. A primary outcome measure was the increase in vertical toe clearance (step ascent) or horizontal heel clearance (step descent) with the illusion compared with no illusion. For experiments 1 and 3, using data from previous studies of step descent and ascent,38,41,42,55 and assuming a 50% increase in variance for older participants compared with younger ones for step ascent,55 a minimum effect of importance of 5%, an estimated standard deviation (SD) of 7% and using a two-sided alpha of 0.05, for a paired t-test a sample size of 21 would result in a power of 90% and a sample size of 16 would result in a power of 80%. Because of improved sensitivity and reliability of the motion analysis system provided by a newly developed marker-based event detection algorithms (see Appendix 1) plus the effect of increasing the number of cameras by 25% to 10, we estimated that the sample size could be reduced to 16 for experiment 1.

Vision measurements and cataract simulation

Binocular distance visual acuity was measured with the participant’s habitual correction for distance (distance spectacles if worn) using an Early Treatment Diabetic Retinopathy Study (ETDRS) logMAR chart at a working distance of 4 m and a chart luminance of 160 cd/m². A by-letter scoring rule and a four-letter termination rule were used. 56 The ETDRS chart is the current standard for visual acuity determination and provides substantially more reliable measurements than the traditionally used Snellen chart. 56

Visual acuity and contrast sensitivity were measured for each participant in experiment 2 while wearing cataract simulation glasses (Stereo Optical Co., Inc., Chicago, IL, USA) over the top of their spectacles or contact lenses, if worn. These glasses scatter light (blurred vision) in the same way as age-related cataract in proportion to the inverse of the squared glare angle. 38,41,42,57 They particularly reduce contrast sensitivity, which is a much better indicator of vision for everyday tasks than the more traditionally measured visual acuity. 57 Binocular contrast sensitivity was measured using the Pelli–Robson letter chart, which includes 16 triplets of letters that reduce in 0.15 log-steps of contrast sensitivity (the inverse of the contrast threshold). 58 The chart was used at 1 m with a luminance of 120 cd/m². A by-letter scoring rule was used, with the miscalling of the letters C and O being counted as correct identification to balance letter legibility. Participants were given at least 20 seconds at threshold to be able to identify additional letters. 58

Stair design and safety issues

Experiments 1 and 2

The stairs used, consisting of three steps, were 1 m wide with the top step consisting of a landing area measuring 1.5 m in length (Figure 2). Each tread/going measured 28.5 cm and the step risers ranged between 16.7 cm and 17.5 cm. The stairs, including step treads and risers, were all painted a uniform grey colour and a 94.5-cm-high (above going) handrail was attached to the right side of the stairs (as viewed during descent), and crash mats were positioned on the left side and in front of the participants for safety. In addition, an assistant was always present next to the stairway during all trials to help steady a participant should they stumble or trip (this did not occur across any of the trials and none of the participants used the handrail at any time). Safety concerns have meant that seven-step stairway studies have protected participants using a safety harness during trials. 35 However, these harnesses inevitably mean that a rather unnatural gait may be used. For these reasons, we used a three-step stairway, with safety features such as a handrail, crash mats and nearby assistant, as used by other gait laboratories. Safety concerns were also the reason why we recruited only participants who were fit and healthy and negotiated steps and stairs on a regular basis, typically in the home environment. Older participants were asked about any concerns with the three-step stairway during pilot studies and all responded that they felt perfectly safe, did not need the handrail and were used to using stairs on a regular basis. We therefore felt satisfied that negotiation of our custom-built stairs would be safe and relatively natural for all participants.

FIGURE 2.

(a) Staircase used in experiments 1 and 2; and (b) single-level raised surface used in experiment 3. The gait parameters assessed in each study are shown.

Tread edge highlighter

Repeated trials were undertaken in each experiment under four experimental tread edge highlighter conditions:

1. plain: no edge highlighter on the tread.

Three conditions involved a high-contrast, 5.5-cm-wide black strip being placed on the tread going in the following locations (Figure 3):

1. abutting: placed flush with the leading edge of the tread

2. away1: placed 1 cm from the leading edge of the tread

3. away3: placed 3 cm from the leading edge of the tread.

FIGURE 3.

The four stair edge highlighter conditions used for each experiment. (a) Plain; (b) abutting: strip placed flush with leading tread edge; (c) away1: strip placed 1 cm from leading edge; and (d) away3: strip placed 3 cm from leading edge.

For experiments 1 and 2, edge highlighters were placed on all three steps (i.e. across the entirety of the top, middle and bottom step edges), and in experiment 3 an edge highlighter was placed across the entirety of the edge of the walkway. The width of the black strip adhered to UK (building) standards for edge highlighters (between 5 cm and 6.5 cm). 48 The mean Weber contrast of the black strip against the grey tread background was 95% [luminance measured with a CS-100 Chroma meter (Minolta Camera Co., Ltd, Japan)], and the laboratory was well lit with an ambient illuminance of 400 lux.

General protocol

Older adults (experiments 1 and 3) and young adults (experiment 2) completed three and five trials of each highlighter condition, respectively. Highlighter conditions were presented in a randomised order (using a random number generator). Participants started from a standing position two-and-a-half walking steps away from the edge of the first step or the edge of the raised surface. The stairs were negotiated by all participants using a ‘step-over-step’ gait (i.e. alternative limb lead on each step). The same self-selected leading limb was used to begin each trial and participants were instructed to use their vision throughout the trial to help negotiate the stairs/walkway. Each experiment took between 1.5 hours and 2 hours and included 20–30 minutes for informed consent, initial vision measurements and marker attachment. Experiment 1 consisted of 15 trials (four conditions and three repetitions plus three ‘dummy’ trails); experiment 2 consisted of 26 trials (four conditions and five repetitions plus six ‘dummy’ trails) and experiment 3 consisted of 30 trials (four conditions, two step heights and three repetitions plus six ‘dummy’ trails).

As per the University of Bradford’s Code of Practice, all participant identification information was locked in secure facilities and kept separate from the research data. All research data were anonymised by attributing a subject code (i.e. subject 1, subject 2, etc.) to all data collected from the participants, and these data were stored on a personal computer with controlled access.

Reducing the repetition or memorisation effect

As discussed previously, an energy-conservation strategy means that during stair descent, foot clearances become progressively smaller35,36 owing, in part, to somatosensory information providing increasingly accurate assessments of the step riser and tread dimensions. In research studies assessing the contribution of vision on gait and movement control, the multiple measures that are required to provide valid mean values of gait parameters, plus sensible assessments of variability, will mean that there is a trade-off between the numbers of repetitions and the likelihood of intrusion of somatosensory system inputs. Clearly, if a large number of trials were used and participants began to rely more on the somatosensory inputs, then any effects of visual influence on gait, although present in the initial few trials, might be lost/negated among the much larger number of somatosensory system-influenced data.

Several strategies were used to attempt to counter possible increase in reliance on somatosensory information. These included:

1. limiting the number of trials to three in most cases or five when a relatively small number of interventions are used (to keep the total number of trails typically below 30)

2. varying the participant’s starting location by ± 5 cm (in randomised order)59

3. using custom-built ‘stepping stones’ (square wooden blocks) of varying height, which participants step on (in random order as indicated by the experimenter) to return to the top of the stairs (Figure 4)

4. using ‘dummy trials’ after every third stair descent trial, with changes to the riser height (± 1 cm) and tread depth (1 cm or 2 cm) of step middle, bottom or both. 39,59

FIGURE 4.

Stepping stones used to return the participant to the top of the stairway. These were used in a random order to help limit somatosensory information regarding the height of the staircase steps.

In addition to providing participants with steps of a different height to perceive, and thus with inconsistent somatosensory information about riser height, this allowed us to inform participants that the height/tread depth and appearance of the steps would vary between some trials (this was done at regular intervals during the protocol). Data were not collected during dummy trials.

Kinematic data

A 10-camera motion-capture system (Vicon MX, Vicon, Oxford, UK) was used to record (at 100 Hz) segmental kinematics (in these studies: movement of pertinent parts of the body, particularly the lower limbs) as participants completed each trial. Cameras were wall- or ceiling-mounted at approximately 2.3 m above the floor. Participants were asked to wear sensible/comfortable flat-soled shoes and comfortable clothing, and used their habitual vision correction throughout each trial. Note that participants in experiment 2 wore the cataract simulation glasses over the top of any corrective lenses that they typically used. Reflective markers (1.4 cm in diameter) were placed on the lower body and thorax segments directly onto the skin, clothing, or shoes in accordance with the guidelines that are defined in Vicon’s ‘plug-in-gait’ full-body marker set. Markers were placed on the left and right greater trochanter, second metatarsal heads and distal phalange of the second toes, and a (non-rigid) cluster of four markers was placed on the sacrum. Virtual markers were created at each shoe’s heel and toe inferior tips (heel and toe tip), by constructing their positions relative to the heel and toe markers, respectively. 22 Markers were also placed on each tread edge or raised surface edge in order to determine its location within the laboratory co-ordinate system.

Vicon MX cameras were carefully positioned for each step/stair protocol. An important condition during data collection is that each marker must be seen by a minimum of two cameras in order to provide three-dimensional coordinates in the capture volume. The height of the stairway was much greater than the raised single-level surface (the landing area of the stairway was 54 cm above the ground; the raised surface was a maximum of 20 cm above the ground), which meant that a larger capture volume was required during stair negotiation compared with single-step descent. For stair negotiation, cameras were positioned so that they covered a volume of 2.5 m (height) × 1 m (width) × 4 m (length), which allowed all markers placed on the participant to be visible throughout the trial. For single-step descent, cameras were positioned so that they covered a volume of 2.2 m × 1 m × 4 m. The camera calibration process describes the intended data-capture volume to the Vicon system and was carried out prior to data collection. A 39-cm wand with three markers of known separation was moved around the intended movement-capture volume to calibrate the volume. A clinical L-frame was subsequently used to define the local co-ordinate system of the laboratory.

Data analysis

Labelling and gap filling of marker trajectories were undertaken within Vicon Nexus (Vicon, Oxford, UK) and the resultant co-ordinate three-dimensional (C3D) files were then uploaded to Visual 3D (C-Motion, Inc., Germantown, MD, USA) for further analysis. Marker trajectory data were filtered using a fourth-order, zero-lag Butterworth filter with a 6-Hz cut-off. Instants of foot touchdown and foot-off in each trial were determined using marker-based event detection algorithms (see Appendix 1). 60 The following variables were then exported:

• Penultimate foot placement: the horizontal distance from the leading-limb’s shoe tip to the edge of the top stair or edge of raised walkway when the foot was placed on the landing area during approach (see Figure 2).

• Final foot placement: the horizontal distance from the trailing limb’s shoe tip to the edge of the top stair or edge of raised walkway when the foot was placed on the landing area during approach (see Figure 2). Negative placement values indicated that the foot was placed before/prior to the edge of the stair/walkway, and positive values indicated that the shoe tip was beyond the edge of the top stair/walkway.

• Middle-step foot placement (experiments 1 and 2): the horizontal distance from the leading-limb’s shoe tip to the edge of the middle step when the foot was placed on the tread surface of the middle step (see Figure 2). Positive values indicated that the shoe tip was beyond the edge of the step and negative values meant it was placed before/prior to the edge of the step.

• Heel placement (experiment 3): the horizontal distance from the lead limb’s heel to edge of the raised walkway when the foot was placed on the ground (see Figure 2).

• Horizontal heel clearance: the horizontal distance from the leading limb’s heel to the edge of the top/middle step of the stairs or edge of raised walkway as the leading limb passed over it (swing phase) (see Figure 2).

• Vertical heel clearance: the vertical distance from the leading-limb’s heel to the edge of the top/middle step of the stairs or edge of raised walkway as the leading limb passed over it (swing phase) (see Figure 2).

• Descent duration: stairs/walkway descent duration was determined from the instant of leading limb foot-off prior to stepping over the top step of stairs or edge of raised walkway to the instant of leading limb touch-down on the ground (see Figure 2).

• Heel scuff/catches: the number of times a participant’s heel scuffed/caught the tread edge/going or riser during stair/walkway descent. Each heel scuff was only recorded if agreed on by the two experimenters present.

• Horizontal heel clearance < 5 mm: the number of times horizontal heel clearance fell below 5 mm. It has previously been determined that there is a greater risk of catching the heel on the tread edge/going if heel clearance falls below 5 mm, especially on flights of stairs where riser height varies between one riser and another. 36

It has previously been reported that stride length during the stair-to-floor transition is significantly increased when compared with mid-stair descent,61 suggesting that gait/stepping behaviour is significantly different when stepping onto the ground compared with stepping onto the stairs. Therefore, the effects of tread edge highlighter condition on foot placement, heel clearance and number of heel scuffs were only considered on/over the top and middle step.

Statistical analysis

Penultimate and final foot placement, middle-step foot placement and descent duration for experiments 1 and 2 were analysed using two-way repeated measures analysis of variance (ANOVA; Statistica 5.5, StatSoft, Inc., Tulsa, OK, USA) with edge highlighter (plain, abutting, away1, away3) and repetition (three and five trials, respectively) as repeated factors. Horizontal heel clearance and vertical heel clearance for experiments 1 and 2 were analysed using three-way repeated measures ANOVA with step (top/middle step), edge highlighter condition (plain, abutting, away1, away3) and repetition (three and five trials, respectively) as repeated factors. Penultimate and final foot placement, heel placement, horizontal heel clearance and vertical heel clearance for experiment 3 were analysed using three-way repeated measures ANOVA with walkway height (16.5 cm/19.5 cm), edge highlighter (plain, abutting, away1, away3) and repetition (1, 2, 3) as repeated factors. All interactions between step number/walkway height, edge highlighter and repetition were found to be of no consequence to the findings of the study and thus are not reported in the results section. Post-hoc analyses were performed using Tukey’s honest significance difference test and the alpha level of significance was set at p = 0.05.

Main effect of edge highlighter

Experiment 1 (stair descent in older adults under habitual vision)

There was a statistically significant effect of highlighter condition on final foot placement [F(3,45) = 11.97; p-value = 0.001], but not on penultimate foot placement (p-value < 0.71). Final foot placement was further back from the tread edge for away1 and away3 than for abutting (post hoc; p-value = 0.005 and p-value < 0.001, respectively) or plain (post hoc; p-value = 0.003 and p-value < 0.001, respectively) (Figure 5). There was a statistically significant effect of highlighter condition on middle-step foot placement [F(3,45) = 15.67; p-value < 0.001]; placement was further back from the tread edge for away1 and away3 than for abutting (post hoc; p-value < 0.001 and p-value < 0.001, respectively) or plain (post hoc; p-value = 0.029 and p-value < 0.001, respectively).

FIGURE 5.

The effect of highlighter condition on (a) final foot placement relative to the top stair edge; and (b) horizontal heel clearance over the top stair edge in older adults during stair descent. Final foot placement was significantly further back from the stair edge when the edge highlighter was set back by 1 cm or 3 cm. Horizontal heel clearance was significantly closer to the stair edge when the edge highlighter was set back by 3 cm.

There was a statistically significant effect of highlighter condition on horizontal heel clearance [F(3,45) = 17.3, p-value < 0.001] and vertical heel clearance [F(3,45) = 8.38, p-value < 0.001 over the top step edge]; clearance values were smaller for away3 than for plain (post hoc; p-value < 0.001 and p-value = 0.001, respectively), abutting (post hoc; p-value < 0.001 and p-value = 0.001, respectively) or away1 (post hoc; p-value = 0.002 and p-value < 0.001, respectively; see Figure 5). There were no main effects of highlighter condition on the horizontal or vertical heel clearance over the middle-step edge (p-value > 0.13). There was no effect of highlighter condition on descent duration (p-value = 0.37). There were no statistically significant differences in within-subject variability between highlighter conditions in any of the outcome parameters analysed (penultimate and final foot placement, horizontal or vertical heel clearance over the top- and middle-step edge, descent duration).

Experiment 2 (stair descent in young adults with simulated age-related impaired vision)

There was a statistically significant effect of highlighter condition on final foot placement [F(3,21) = 5.88, p-value = 0.004], but not on penultimate foot placement (p-value = 0.60). Final foot placement was significantly further behind the tread edge for plain in comparison with abutting (p-value = 0.004) or away1 (p-value = 0.019) (Figure 6). There was a statistically significant main effect of highlighter condition on middle-step foot placement [F(3,21) = 3.09, p-value = 0.049], but a post-hoc analysis indicated there were no statistically significant differences between highlighter conditions. There was a statistically significant effect of highlighter condition on final foot placement and middle step foot placement within-subject variability [F(3,21) = 3.11, p-value = 0.048; and F(3,21) = 3.84, p-value = 0.025, respectively]. Final foot placement within-subject variability was reduced for the plain condition compared with away3 (post hoc; p-value = 0.038) (see Figure 6). Middle-step foot placement within-subject variability was reduced for the abutting condition, compared with away1 (p-value = 0.039) or plain (p-value = 0.035). There was a statistically significant effect of highlighter condition on horizontal and vertical heel clearance over the top step edge [F(3,21) = 3.8, p-value = 0.002; and F(3,21) = 4.14, p-value = 0.019, respectively]: clearance values were smaller for the away3 compared with abutting (post hoc; p-value = 0.001 and p-value = 0.026, respectively). There was a statistically significant effect of highlighter condition on horizontal heel clearance within-subject variability [F(3,21) = 6.25, p-value = 0.003]. Horizontal heel clearance within-subject variability was increased in the plain condition compared with abutting (p-value = 0.022), away1 (p-value = 0.024) or away3 (p-value = 0.003) (see Figure 6). There was a statistically significant effect of highlighter condition on vertical heel clearance over the middle-step edge [F(3,21) = 4.99, p-value = 0.009], but no effect on horizontal heel clearance (p-value = 0.06); vertical heel clearance was significantly closer to the edge in away3 compared with plain (post hoc; p-value = 0.005).

FIGURE 6.

The effect of highlighter condition on (a) final foot placement; (b) final foot placement within-subject variability; (c) descent duration; and (d) horizontal heel clearance within-subject variability in young adults with a simulated visual impairment. Final foot placement was significantly further behind the stair edge in the plain condition. Final foot placement within-subject variability and horizontal heel clearance within-subject variability increased significantly when presented with the plain condition, whereas descent duration was significantly longer in the plain condition.

There was a statistically significant effect of highlighter condition on descent duration [F(3,21) = 7.86, p-value = 0.001], and on descent duration within-subject variability [F(3,21) = 4.88, p-value = 0.001]. Descent duration was significantly longer for the plain condition compared with abutting (p-value = 0.009), away1 (p-value = 0.036) or away3 (p-value = 0.001) (see Figure 6). Descent duration within-subject variability was also significantly longer in the plain condition compared with away3 (post hoc; p-value = 0.006).

Experiment 3 [walkway descent (across both heights) by older adults under habitual vision conditions]

There was a statistically significant effect of highlighter condition on final foot placement [F(3,42) = 8.90, p-value < 0.001]; placement was further behind the tread edge for away1 than for plain (p-value = 0.002), and for away3 than for abutting (p-value = 0.025) or plain (p-value < 0.001) (Figure 7). There was a statistically significant effect of highlighter condition on heel placement [F(3,42) = 5.36, p-value = 0.003]; placement was closer to the tread edge in the away3 condition than in the abutting condition (p-value = 0.016) or plain (p-value = 0.015).

FIGURE 7.

The effect of highlighter condition on (a) final foot placement; and (b) horizontal heel clearance in older adults during walkway descent. Final foot placement was significantly further back from the stair edge when the edge highlighter was set back by 1 cm or 3 cm. Horizontal heel clearance was significantly closer to the stair edge when the edge highlighter was set back by 3 cm.

There was a statistically significant effect of highlighter condition on horizontal [F(3,42) = 17.70, p-value < 0.001] and vertical heel clearance [F(3,42) = 12.12, p-value < 001]. Clearance values (see Figure 7) were smaller for away3 than for plain (p-value < 0.001 and p-value < 0.001, respectively), abutting (p-value < 0.001 and p-value < 0.001, respectively) or away1 (p-value < 0.001 and p-value < 0.001, respectively).

There was a statistically significant effect of highlighter condition on descent duration [F(3,42) = 3.14, p-value = 0.035], but a post-hoc analysis indicated there were no statistically significant differences between highlighter conditions (p-value > 0.064).

There were no statistically significant differences in within-subject variability between highlighter conditions in any of the outcome parameters analysed (penultimate and final foot placement, horizontal or vertical heel clearance, heel placement or descent duration).

Main effect of repetition

In all three experiments, stair/walkway descent duration was significantly reduced in the last trial compared with the first trial. For example, in experiment 1, descent duration in trial 1 [2.10 seconds (SD 0.39 seconds)] was significantly longer than in trial 3 [2.01 seconds (SD 0.39 seconds); p-value = 0.019]. In experiments 1 and 2, final foot placement and middle-step foot placement were closer to the tread edge in the last trial than in the first trial (p-value < 0.05).

Heel scuffs and horizontal heel clearance ≤ 5 mm

The numbers of trials that resulted in a heel scuff or horizontal heel clearance falling below 5 mm in all three experiments are shown (as a percentage) in Table 3. In all three experiments the number of trials in which horizontal heel clearance fell below 5 mm was highest for the away3 highlighter.

Vision measurements

The mean binocular visual acuity for the older participants in experiments 1 and 3 was –0.02 (SD 0.06; Snellen score of ≈ 6/6, decimal 1.0) and the worst binocular visual acuity was 0.10 (Snellen score of 6/7.5, decimal acuity 0.80); as none of the participants had eye disease, these participants represent people of this age with excellent vision. However, many older people have some level of visual impairment. For experiment 2, mean binocular visual acuity was 0.16 (SD 0.16; Snellen score of ≈ 6/9, decimal 0.67), with contrast sensitivity being reduced to 0.75 log contrast sensitivity. Although visual acuity is good, the low level of contrast sensitivity means that vision is clearly impaired and represents the level of vision at the other end of the spectrum to that in experiments 1 and 3.

Presence of a tread edge highlighter

All stepping parameters in experiments 1 and 3 were unaffected by the presence of an edge highlighter that was placed flush with the leading edge of the tread in comparison with when there was no highlighter present. This finding agrees with previous research regarding the influence of edge highlighters on stairs,35,53 and is likely to be a consequence of the older adults who took part having very good binocular visual acuity, which suggests that they would have been able to delineate the edge of the treads when there was no edge highlighter present. However, the results of all three experiments indicate that the location of a tread edge highlighter relative to the tread edge can significantly affect foot placement and heel clearance when descending stairs or a raised surface and, notably, the location of the highlighter strip relative to the tread edge influences risk of tripping in visually normal adults and more so in adults with a visual impairment.

For young adults with simulated visual impairment, within-subject variability in final foot placement and in horizontal heel clearance over the top step was increased for the plain condition compared with all highlighter conditions. This, and the increased time it took participants to negotiate the stairs for the plain condition, suggests that there was uncertainty in determining the exact location of the top/first step edge when there was no edge highlighter present. There also appeared to be an increased risk of tripping in the plain condition, as evidenced by the high number of trials where participants caught their heel (15% of trials), or where horizontal heel clearance fell below 5 mm (7.5%), although no actual trips resulted from the scuff incidences. These results agree with a previous report indicating that minimum heel clearance within-subject variability increased when older adults were uncertain about the location of the tread edge, and that the percentage of heel clearances below 5 mm increased from 0.56% in good lighting conditions to 6% under poor lighting conditions. 36

Position of the tread edge highlighter

For older adults with habitual visual correction (experiment 1), final foot placement, middle-step foot placement, horizontal heel clearance and vertical heel clearance were significantly reduced for the away3 condition compared with the plain and abutting conditions. Final- and middle-step foot placements were also significantly reduced for the away1 condition compared with plain or abutting. Furthermore, for the away3 condition, horizontal heel clearance was reduced to 3.2 cm (SD 1.9 cm) from 4.7 cm (SD 1.4 cm), showing a large effect size of 1.07 [calculated as the difference between the intervention and control horizontal heel clearance divided by the SD of the control value (4.7 – 3.2)/1.4]. The away3 condition horizontal heel clearance was below 5 mm for 16.7% of all trials, suggesting that there was a greater risk of tripping when the definition of the edge of the tread was misleading. There was less risk of tripping associated with negotiating the walkway (experiment 3), although foot placement and heel clearance altered in a similar manner to that on the stairs. However, the magnitude of the change in both variables was much greater than evident on the stairs and, hence, this meant that there was less risk of participants catching their heel on the edge of the tread, as evidenced by the low percentage of horizontal heel clearances that fell below 5 mm.

Tread edge highlighter relative location was also seen to alter stepping parameters in young participants with simulated impaired vision (experiment 2). Horizontal heel clearance over the top step was significantly reduced for the away3 condition [3.8 cm (SD 1.8 cm)] compared with the control abutting condition [5.8 cm (SD 1.7 cm)], a reduction of 2 cm, giving a large effect size of 1.18 (5.8 – 3.8/1.7). This would have increased the risk of catching the heel on the step, and this is emphasised by the high percentage of trials for the away3 condition where scuffs occurred (10%) or where horizontal heel clearance fell below 5 mm (10%). In experiment 3, horizontal heel clearances from the raised-level walkway were much larger, but were significantly reduced for the away3 condition [7.8 cm (SD 2.6 cm) and 6.4 cm (SD 2.8 cm) compared with abutting 10.0 cm (SD 3.1 cm) and 8.2 cm (SD 2.9 cm)] for the 16.5-cm and 19.5-cm raised-level walkways, respectively, giving reductions of 2.2 cm and 1.8 cm and effect sizes of 0.71 and 0.62. These findings suggest that the location of an edge highlighter is important to consider for stair design and improving safety on stairs.

Comparison with previous work

It is known that foot placement and heel clearance when crossing obstacles or descending a kerb are determined/planned using visual information acquired during the approach to an obstacle or kerb edge. 21,63 The results from the present study suggest that the visual information provided by a tread highlighter during the approach to the stair/walkway contributes to determining foot placement and heel clearance. When the highlighter was placed further away from the tread edge, final foot placement was placed further behind the edge of the tread. Similarly, when the highlighter was set back by 1 cm (away1), final foot placement was placed behind the tread edge by a similar amount (i.e. ≈ 1 cm). Interestingly, for the away3 condition, foot placement was ≈ 1.5–2.0 cm behind the tread edge and, as a consequence of this, horizontal heel clearance was significantly reduced, thus reducing stair safety. Thus, increasing the distance between the highlighter and the actual edge of the tread will be likely to increase the risk of a trip incident occurring on stairs or when stepping down from a kerb, although further study is required to confirm this.

The findings presented here suggest that a high-contrast edge highlighter placed flush with the tread edge (abutting) rather than being placed back from the tread edge could reduce the risk of a trip incident occurring, particularly in high-risk older adults with uncorrected visual impairment or in those not using their corrective spectacles. Use of such highlighters could also improve stair safety for multifocal lens wearers, who are affected by blur in the lower visual field, which can lead to impaired contrast sensitivity and depth perception at critical distances required for detecting objects in the environment. 39,46

Although the number of low heel clearances and heel scuffs increased when the tread edge highlighters were positioned away from the tread edge in both young and older adults, the amount of foot overhang (when the most anterior portion of the foot is placed beyond the tread edge) on the middle step was decreased. Although there are no formal reports on the amount of foot overhang that would probably lead to a loss of balance, we estimate that stair users are at an increased risk if 50–60% of the shoe plantar surface (i.e. the part of foot distal to the metatarsophalangeal joint line) is hanging over the tread edge. In such a scenario, it is likely that controlling the body centre of mass in the forwards–downwards direction during the lowering phase would be difficult because of the reduced base of support. In the present study, the largest foot overhang experienced by younger and older adults accounted for, on average, 17% and 1.5% of the shoe plantar surface, respectively (based on an average shoe length of 28 cm and 27 cm, respectively), with a margin for error (95% CI) of < 29% and 23% respectively. This suggests that when the edge highlighter was set back from the tread edge there was less risk of a fall/trip as a result of foot overhang than there was of catching the heel on the step.

It is important to consider the implications of the present study’s results in the context of stair safety. UK and US (building) standards clearly state what the desired thickness of the tread edge highlighter should be, but fail to specify the preferred positioning of the edge highlighter on the tread or the desired contrast between the highlighter and the rest of the tread surface. A high contrast (95%) existed between the black highlighter and grey tread surface in the present study. Clearly, if the contrast of the highlighter is reduced, it is less likely to be seen by older participants, which could be predicted from their contrast sensitivity and visual acuity scores. The results from the present study could, therefore, be used to inform existing building regulations regarding the use, positioning and contrast of edge highlighters on stairs, with the aim of improving safety on stairs and ultimately reducing the number of falls on stairs. This is likely to impact directly on the associated health-care costs of falls in older adults, which is considerable. 32,33

Study limitations

There were limitations with the present study. Most notably, the number of steps on the stairs could have been increased to replicate more representative real-world stair negotiation. However, it is possible that participants would receive increased somatosensory feedback from an increased number of steps, thus potentially confounding the actual effects of how edge highlighters influence stair descent gait characteristics. Increased somatosensory feedback is an unavoidable complication of adaptive gait studies59,64 that needs to be considered; a number of aspects of the protocol were manipulated in the present study which aimed at reducing such effects (see General protocol).

Experiment 4a: spatial frequency of the vertical lines

The first experiment aimed to determine the optimum spatial frequency of the vertical lines on the step riser that was required to increase perceived riser height. Images of a three-step sequence were produced and displayed on a Macintosh Cinema Display under the control of a G4 PowerMac (Apple Inc., Cupertino, CA, USA). When viewed from 33 cm, the geometry of the image was equivalent to that of actual steps viewed from a distance of 1.4 m (two steps) at an eye height of 160 cm, approximating to that of an average elderly person. Test stimuli had one of five different square-wave spatial frequencies (i.e. black and white gratings of the following thicknesses: 4, 8, 12, 16 and 20 cycles per step) placed on the bottom step riser whose height was equivalent to 190 mm (Figure 9).

FIGURE 9.

The varying square-wave spatial frequencies placed on the step riser when determining the optimum spatial frequency of vertical lines required to increase perceived riser height. (a) Four cycles per step; (b) eight cycles per step; (c) 12 cycles per step; (d) 16 cycles per step; and (e) 20 cycles per step.

On any single trial the perceived height of the chosen test stimulus was compared with a reference stimulus that contained no square-wave pattern, but in which the actual height of the bottom step could be chosen from one of seven heights spanning a range from physically smaller to physically taller (Figure 10). The height of riser on these seven reference stimuli ranged from 180 mm to 240 mm in 10-mm steps.

FIGURE 10.

Reference stimuli provided during the psychophysical determination of the optimum spatial frequency, which contain no square-wave pattern. The bottom step reference ranged between (a) 180 mm and (b) 240 mm.

The task of the observer on each trial was to decide whether the riser on the test (striped) step was smaller or taller than that on the reference step and to respond via the keyboard. Images were presented in succession for 500 milliseconds each and the order of the test and reference stimulus was randomised. A total of 20 responses were obtained (in pseudorandom order) for each test spatial frequency (thickness of the black and white gratings) at each of the seven reference step heights, making a total of 700 responses. Data were gathered across a number of sittings.

Psychometric functions were produced by plotting the percentage of ‘test step taller’ responses against reference step height (180 mm, 190 mm, etc.) and performing a least-squares logistic regression to find the reference step height that matched the perceived height of the test step (the point at which the test step was judged ‘taller’ on 50% of trials).

Following initial pilot data it was decided to introduce a nosing onto the landing portion of the first step. This was either 5.5 cm or 11 cm in width (Figure 11).

FIGURE 11.

The nosing provided on the landing portion of the first step. (a) 5.5 cm; and (b) 11 cm.

Experiment 4a: results

Figure 12 represents data from each the seven participants, showing perceived height of the patterned step as a function of the spatial frequency of the square wave.

FIGURE 12.

Perceived riser height as a function of square-wave spatial frequency. Error bars represent the SD of the estimates. Results are shown for the two step highlighters (5.5 cm in black or 11 cm in green). (a) Participant 1; (b) participant 2; (c) participant 3; (d) participant 4; (e) participant 5; (f) participant 6; and (g) participant 7.

All observers show statistically significant overestimations of the true height of the step (190 mm) and, for all but one observer, the magnitude of this overestimation increased with spatial frequency (i.e. the H–V effect was greater as the black and white gratings became thinner). Several of the observers were veridical in height judgement at the lowest spatial frequency (they judged the height of the step correctly when the black and white gratings were relatively thick) but overestimation of height increased rapidly with spatial frequency (as the gratings became thinner). The magnitude of this illusory percept is substantial, with overestimations approaching 25% for some observers. The width of the nosing appears to have little effect.

Experiment 4a: discussion

The results suggest that for the majority of participants, higher spatial frequencies (thinner and, thus, more frequent gratings) show the highest perceived height increase, and if we wish to obtain a range of foot clearances, a range of spatial frequencies on a step riser should be used, with lower frequencies (thicker gratings) likely to provide lower foot clearances and vice versa.

Experiment 4b: location and thickness of the nosing

In experiment 4a, we found little effect of the thickness of the nosing, but in experiment 4b we wanted to look more closely at the effect of its presence or position. To do this we chose a single spatial frequency on the step riser of 16 cycles per step on the basis of the results of experiment 1a. The horizontal line thickness of 5.5 cm was chosen as it is the guideline for stair nosings in UK Building Regulations. 50 We used an 11-cm-wide horizontal line thickness to determine whether or not changes in the thickness of the horizontal line would alter the size of the effect. We also used an illusion with the horizontal line set back from the tread edge, as provided on stairways with friction strips and as assessed previously (see Chapter 2). Four participants (aged 24, 28, 52 and 55 years) with healthy eyes and normal vision (visual acuity better than 0.0 logMAR or 6/6 or 20/20 on the Snellen scale) took part in the experiment.

Experiment 4b: results

All four observers show a consistent pattern of results, with the most significant overestimation of riser height (up to 20%) occurring for the abutting nosing condition (Figure 13). Absence of a nosing, or the presence of a gap between step edge and the nosing, reduced (but did not eliminate) the illusion.

FIGURE 13.

Perceived riser height increases in response to manipulations to the location and thickness of the nosing. (a–b) No nosing; (c–d) 5.5 cm abutting; (e–f) 5.5 cm gap; and (g–h) 11 cm abutting.

Experiment 4b: discussion

The results suggest that the presence of an abutting nosing reinforces the H–V illusion and produces the greatest magnitude of riser height overestimation. If a step edge highlighter is to be used as part of the H–V illusion, then it would seem to be important for the highlighter to be abutting the step edge. It is serendipitous that the data in Chapter 2 indicate that having the tread edge highlighter abutting the step edge also improves the safety of stair descent compared with highlighters set back from the tread edge (typically when used as friction strips).

Participants

Eleven participants were randomly recruited from the same pool as described previously (see Chapter 2) and we used the same inclusion and exclusion criteria. The mean age was 70 years (SD 7 years) and three participants were female. Mean height and mass were 1.73 m (SD 0.10 m) and 81.3 kg (SD 17.4 kg), respectively. Mean distance binocular visual acuity was –0.07 (SD 0.08; Snellen scale equivalent of approximately 6/5, decimal 0.83) and binocular Pelli–Robson letter contrast sensitivity was 1.85 log-units (SD 0.14 log-units). Some of the participants from the experiments discussed in Chapter 2 participated again, although they must have completed those experiments more than 3 months previously. Because of improved sensitivity and reliability of the motion analysis system as shown by the results of the Chapter 2 experiments and as provided by newly developed algorithms of detecting important events in the gait cycle (see Appendix 1),60 plus the effect of increasing the number of cameras from 8 to 10, we estimated that the sample size should be reduced to 11 for these experiments. Participants included four varifocal or progressive lens wearers and three bifocal wearers, two distance single-vision lens wearers and three people who did not wear spectacles when walking.

Experimental details

The tenets of the Declaration of Helsinki were observed,54 the experiments gained institutional ethical approval and all participants gave written, informed consent. Visual acuity and contrast sensitivity was measured as previously described (see Chapter 2). A raised walkway of uniform grey colour, with a width of 1 m, a height of 16.5 cm and a landing area measuring 2 m in length was used (Figure 14). This represents a surface-level change typically encountered when stepping up onto a kerb. Crash mats were positioned on the right and left sides of the walkways for safety, and during all experiments a research team member was positioned close to the walkways to aid participants if they lost balance or stumbled during the trial (although this did not occur across any of the trials).

FIGURE 14.

Single-level raised surfaced used in experiment 5. The gait parameters assessed are shown in the respective parts of the figure.

Adaptive gait parameters were assessed as 11 older subjects took two walking paces before stepping up onto the raised walkway. Participants came to a standstill after stepping up onto the raised walkway. The step riser and tread was either plain (control); plain apart from an abutting 5.5-cm, high-contrast, black step edge highlighter (see Chapter 2); or included one of three versions of the H–V illusion superimposed onto the step (Figure 15). The mean Weber contrast of the black strips against the grey tread background was 95% (luminance measured with a CS-100 Chroma meter), and the laboratory was well lit with an ambient illuminance of 400 lux. Each of the five riser conditions was repeated three times and the 15 trials were presented in a randomised order.

FIGURE 15.

The three sets of gratings placed on the step riser as part of the H–V illusion. The gratings had a spatial frequency of (a) four cycles per step; (b) 12 cycles per step; or (c) 20 cycles per step. They were all accompanied by a 5.5-cm, horizontal, black, high-contrast strip along the tread edge (a tread edge highlighter) to complete the H–V illusion.

Because of the many repetitions of stair ascent involved in the experiment, subjects can become familiar with the step heights owing to somatosensory information from the feet. This is not the same situation as when ascending stairs outside the home and is an experimental problem that may dilute the effect of the visual illusion. Several strategies were used to attempt to counter possible increase in reliance on somatosensory information, including limiting the number of repetitions for each condition to three; varying the participant’s starting location by ± 5 cm (in randomised order); using custom-built ‘stepping stones’ (square wooden blocks) of varying height, which participants step on (in random order as indicated by the experimenter) to return to the floor and using ‘dummy trials’ after every third stair ascent trial, with changes to the riser height (± 1 cm or 2 cm). 59 Participants were asked to wear their own, comfortable, flat-soled shoes and were allowed to lead with whichever foot they felt most comfortable with. They were then asked to lead with that foot in all subsequent trials. The kinematic data were collected using the same system described previously (see Chapter 2), except that a digitising wand (C-Motion, Inc., Germantown, MD, USA) determined virtual landmarks at the anterior inferior point of each shoe (shoe-tip), and also determined the tread edge of the raised walkway. All measurements, including 19 trials (five-step appearance variables with three repetitions each, plus four ‘dummy’ trials),were captured within a 1.5-hour time period.

Data analysis

Labelling and gap filling of marker trajectories were undertaken within Vicon Nexus and the resultant C3D files were then uploaded to Visual 3D for further analysis. Marker trajectory data were filtered using a fourth-order, zero-lag Butterworth filter with a 6-Hz cut-off. Instants of foot touchdown and foot-off in each trial were determined using marker-based event detection algorithms (see Appendix 1). 60 The following variables were then exported:

• Penultimate and final foot placements: the horizontal distance from the leading limb’s (penultimate) or trailing limb’s (final) shoe tip to the edge of the bottom stair or edge of the raised walkway when the foot was placed on the ground during approach (see Figure 14).

• Vertical toe clearance: the vertical distance from the leading limb’s shoe-tip to the edge of the raised walkway as the leading limb passed over it (swing phase; see Figure 14).

• Ascent duration: walkway ascent duration was determined from the instant of leading limb foot-off prior to stepping over the edge of the raised walkway to the instant of the leading limb’s touch-down on the raised walkway. 60

In addition, to determine whether or not the H-V illusions may have disrupted balance owing to a mismatch of information from the visual and somatosensory systems, the variation in mediolateral displacement of the swing-limb foot and trunk during single-limb support were assessed.

All variables were analysed using a two-way repeated measures ANOVA with illusion (plain, abutting, spatial frequency four cycles per step, spatial frequency 12 cycles per step, spatial frequency 20 cycles per step) and repetition (three trials) as repeated factors.

Results

Vertical toe clearance was significantly increased by the H–V illusions [F(4,40) = 13.74, p-value < 0.0001] (Figure 16). There was an increase in vertical toe clearance for each H–V illusion in comparison with plain and abutting (post hoc; p-value < 0.0007 and p-value < 0.004 respectively). There were no statistically significant differences between each H–V illusion, or between plain and abutting conditions (p-value > 0.05). Between-subject variability was reduced for the 12 cycles per step spatial frequency illusion (SD 1.9 cm) in comparison with the four cycles per step spatial frequency illusion (SD 2.5 cm) and the 20 cycles per step spatial frequency illusion (SD 2.4 cm), although differences in within-subject variability did not reach statistical significance (p-value = 0.41).

FIGURE 16.

Vertical toe clearance (VTC) was significantly increased by each H–V illusion in comparison to the plain and abutting visual conditions. Spfr4, four cycles per step spatial frequency; spfr12, 12 cycles per step spatial frequency; spfr20, 20 cycles per step spatial frequency.

The presence of the H–V illusions did not appear to disrupt other aspects of the gait cycle. The H–V illusions had no statistically significant effect on penultimate or final foot placement and step ascent negotiation time (p-value > 0.05) or mediolateral displacement of the foot and trunk during single-limb support (p-value > 0.05). There were also no statistically significant differences for within-subject variability across all dependent variables (p-value > 0.05).

Discussion

All three H–V illusions led to highly statistically significant increases in vertical toe clearance of the walkway edge from approximately 7.04 cm to 8.48 cm (20% increase) for the four cycles per step and 12 cycles per step spatial frequency illusions and to 8.93 cm (27% increase) for the 20 cycles per step spatial frequency illusion. The increases in toe clearance are much larger than those found in the pilot study of 0.52 cm (3.4% increase) and this is most likely to be the result of the multifactorial nature of the illusory effect used in the pilot study (see Figure 8), in which the effects of the Helmholtz square illusion, which are probably included in the H–V illusion used in that study, resulted in objects appearing to expand in a direction orthogonal to the striped texture within them and the opposite effect to that which we attempted to create. 55 The level of toe clearance increase at 1.4–1.9 cm in the current study may seem small, but, given that these represent 20%+ levels of increase and dangerous levels of toe clearance are reported when they are within 0.5 cm of the step edge,36 these figures actually represent relatively large changes in adaptive gait. They certainly suggest that the use of this illusion on surface-level changes such as kerbs would lead to increased foot clearances and much less chance of tripping.

There was no indication that the H–V illusions lead to changes in other measures of adaptive gait, and there was no change in penultimate or final foot placement, step ascent negotiation time or within-subject variability across all dependent variables. In addition, the mismatch of information from the ‘tricked’ visual system and the somatosensory system did not appear to cause any problems, with no statistically significant mediolateral displacement of the foot and trunk during single-limb support for example. It is possible, that such effects may be seen with a multiple step system, but there did not appear to be any problems with the one-step walkway used in this study.

In terms of which spatial frequency illusion to take forward into future studies, it may be that the 12 cycles per step spatial frequency illusion is preferred. The four cycles per step and 12 cycles per step spatial frequency illusions, providing 20% increases in toe clearance, are preferred to the 20 cycles per step spatial frequency illusion (27% increase) for further study as the larger increased toe clearance has more potential to provide mismatches with the information from the somatosensory system that could potentially disrupt the balance and gait controls systems. In addition, there may be an advantage that the 12 cycles per step illusion would be seen by more visually impaired people given that when it is viewed from two walking steps away it would require a lower resolution (≈ 1.80 logMAR, Snellen score of 6/380) compared with the 20 cycles per step illusion (1.60 logMAR, Snellen score of 6/240). The 12 cycles per step spatial frequency illusion may be preferred to the four cycles per step illusion given the reduced intersubject variability in vertical toe clearance (SD 1.9 cm compared with SD 2.5 cm, respectively; a 24% decrease), plus the apparent lack of an effect for the four cycles per step illusion for a small number of participants in the psychophysical experiments.

Participants

Fourteen participants were randomly recruited from the same pool as described previously (see Chapter 2), and we used the same inclusion and exclusion criteria. Some of the participants from Chapter 2 experiments participated again, although they must have completed those experiments more than 3 months previously. There were nine females, with characteristics as follows: age (mean 69 years, SD 7 years), height (mean 1.66 m, SD 0.09 m), mass (mean 68.8 kg, SD 14.3 kg) and binocular visual acuity (mean –0.08, SD 0.07 logMAR; Snellen scale equivalent 6/5, decimal 1.2).

Experimental details

The tenets of the Declaration of Helsinki were observed, the experiments gained institutional ethical approval and all participants gave written, informed consent. 54 Visual acuity and contrast sensitivity was measured as previously described (see Chapter 2). A three-step stairway was used, which was 1 m wide, with the top step consisting of a landing area measuring 1.5 m long (Figure 17). Each tread/going measured 28.5 cm and the step risers ranged between 16.7 cm and 17.5 cm.

FIGURE 17.

Dimensions of the stairs used in experiment 6. The gait parameters assessed are shown in the respective parts of the figure.

The stairs, including step treads and risers, were all painted a uniform grey colour and a 95-cm-high handrail was attached to the left side of the stairs (as viewed during ascent), and crash mats were positioned on the right side of the stairway. In addition, an assistant was always present next to the stairway during all trials to help steady a participant should they stumble or trip (this did not occur across any of the trials and none of the participants used the handrail at any time). Adaptive gait parameters were assessed as subjects took three walking paces before ascending the three-step stairway. The step conditions assessed were:

1. plain with an abutting 5.5-cm, high-contrast, black step edge highlighter on each tread edge (control condition, Figure 18a)

2. plain with an abutting 5.5-cm, high-contrast, black step edge highlighter on each tread edge plus the H–V illusion superimposed onto the bottom step riser (the first step, Figure 18b)

3. plain with an abutting 5.5-cm, high-contrast, black step edge highlighter on each tread edge plus the H–V illusion superimposed onto the top step riser (the last step, Figure 18c)

4. plain with an abutting 5.5-cm, high-contrast, black step edge highlighter on each tread edge plus the H–V illusion superimposed onto the bottom and top step risers simultaneously (Figure 18d).

FIGURE 18.

Four visual conditions were provided in experiment 6. (a) Abutting: 5.5-cm edge highlighter placed flush with the leading tread edge; (b) H–V illusion superimposed on the bottom step riser only; (c) H–V illusion superimposed on the top step riser only; and (d) H–V illusion superimposed on the bottom and top step riser.

The H–V illusion consisted of black vertical gratings of 12 cycles per step (see Figure 15) on the step riser plus the 5.5-cm black tread edge highlighter previously described (see Chapter 2). The mean Weber contrast of the black strips against the grey tread background was 95% (luminance measured with a CS-100 Chroma meter), and the laboratory was well lit with an ambient illuminance of 400 lux. Each of the four step conditions was repeated three times and the 12 trials were presented in a randomised order. Several strategies were used to attempt to counter possible increase in reliance on somatosensory information, as described previously (see Chapter 2). 59 Participants were asked to wear their own, comfortable, flat-soled shoes and were allowed to lead with whichever foot they felt most comfortable with; they were then asked to lead with that foot in all subsequent trials. The kinematic data were collected using the same system described previously (see Chapter 2), except that a digitising wand determined virtual landmarks at the anterior inferior and posterior inferior point of each shoe (shoe tip/heel tip), and also determined the tread edge for each stair. All measurements, including 15 trials, were made within a 1.0-hour time period.

Data analysis

Labelling and gap filling of marker trajectories were undertaken within Vicon Nexus and the resultant C3D files were then uploaded to Visual 3D for further analysis. Marker trajectory data were filtered using a fourth-order, zero-lag Butterworth filter with a 6-Hz cut-off. Instants of foot touchdown and foot-off in each trial were determined using marker-based event detection algorithms (see Appendix 1). 60 The following variables were then exported for each step (bottom, middle and top):

• Penultimate and final foot placements: the horizontal distance from the leading limb (penultimate) or trailing limb (final) shoe tip to the edge of the bottom stair edge when the foot was placed on the ground during approach (see Figure 17).

• Vertical toe clearance: the vertical distance from the leading-limb shoe tip to the edge of the bottom/middle/top step as the leading limb passed over it (swing phase; see Figure 17).

The following variables were chosen to determine whether or not any changes in adaptive gait owing to the H–V illusion lead to increases in instability:

• Ascent duration: walkway ascent duration was determined from the instant of leading limb foot-off prior to stepping over the bottom step of the stairs to the instant of leading limb touch-down on the landing of the stairs. 60

• Single-limb support: single-limb support was defined from the instant of leading limb foot-off prior to stepping to the instant of leading limb touch-down on the raised walkway or step above.

• Variability in mediolateral displacement of the foot: the variation in mediolateral leading limb foot displacement during single-limb support on the trailing limb for single-step ascent. Variation in mediolateral leading limb foot displacement during single-limb support on the ground, bottom and middle step for three-step ascent.

• Variability in mediolateral displacement of the trunk: the variation in mediolateral trunk displacement during single-limb support on the trailing limb for single-step ascent. The variation in mediolateral trunk displacement during single-limb support on the ground, bottom and middle step for three-step ascent.

Data were analysed using a random-effects regression model with maximum likelihood estimating, using Stata Release 13.0 (StataCorp LP, College Station, TX, USA). Owing to the exploratory nature of the study, a ‘type I’ error adjustment of the alpha level was not deemed necessary and the level of significance was set at a p-value of < 0.05. Factors of interest were incorporated sequentially and their statistical significance were tested using a likelihood ratio (LR) test. Factors with a p-value of < 0.1 were provisionally retained, whereas those above 0.1 were dropped. The final model adopted was the most parsimonious one that was felt to explain the data adequately. The p-values quoted in the text of the paper are those associated with the specific terms in the final regression model, which were:

1. step appearance: a fixed factor with four levels: plain (the control condition), or the H–V illusion placed on the top and bottom steps, bottom step only, or top step only

2. step position: a fixed factor with three levels, (bottom, middle and top step)

3. repetition: a fixed factor with three levels, (trials 1, 2 and 3).

Results

Vertical toe clearance data for each visual condition and for each step are shown in Figure 19. Vertical toe clearances varied with the four different appearances of the steps for the bottom [LR χ² = 53.6, degrees of freedom (df) = 3; p-value < 0.0001] and top steps (LR χ² = 41.0, df = 3; p-value < 0.0001), but were the same for the middle step (LR χ² = 1.4, df = 3; p-value = 0.71). For the bottom step, vertical toe clearance was increased when the illusion was placed on the bottom step only (z-value = 4.2; p-value < 0.0001) or on both the top and bottom step (z-value = 4.9; p-value < 0.0001), but was similar to the control (showing a trend to be slightly reduced) (z-value = –1.9; p-value = 0.063) when on the top step only. For the top step, vertical toe clearance was increased when the illusion was placed on the top step only (z-value = 5.3; p-value < 0.0001) or on both the top and bottom step (z-value = 4.2; p-value < 0.0001), but was similar to the control (z-value = –0.1; p-value = 0.92) when on the bottom step only.

FIGURE 19.

Box-and-whisker plot of vertical toe clearance data for each visual condition. 1, control condition with horizontal step edge highlighter on step tread only; 2, illusion on top and bottom steps; 3, illusion on bottom step only; and 4, illusion on top step only (for bottom, middle and top step).

The most parsimonious model for vertical toe clearance (LR χ² = 313.8, df = 17; p-value < 0.0001) indicated statistically significant effects of step number, appearance and repetition, with statistically significant interaction terms of step × appearance and step × repetition. The full model is displayed below.

There was no statistically significant appearance × repetition effect (LR χ² = 2.1, df = 6; p-value = 0.91). Vertical toe clearance was significantly reduced on the middle [–1.75 cm, standard error (SE) 0.27 cm; z-value = –6.4; p-value < 0.001] and top steps (–1.64 cm, SE 0.27 cm; z-value = –6.0; p-value < 0.0001) compared with the bottom step across all conditions.

Penultimate and final foot placements were unaffected by step appearance or repetition (LR χ² = 3.1, df = 5; p-value = 0.68; LR χ² = 3.9; p-value = 0.56). All measures of potentially increased instability did not change with step appearance. Single-limb support (LR χ² = 4.0, df = 3; p-value = 0.26), ascent duration (LR χ² = 5.3, df = 3; p-value = 0.15), medial–lateral variability of foot movement (LR χ² = 2.7, df = 3; p-value = 0.44), medial–lateral variability of trunk movement (LR χ² = 0.7, df = 3; p-value = 0.86), cumulative distance of the foot (LR χ² = 2.9, df = 3; p-value = 0.41) and cumulative distance of the trunk (LR χ² = 2.2, df = 3; p-value = 0.53) did not show any change with the appearance of the steps. The variability of vertical toe clearance is shown in Figure 19. Inspection of the box plot suggests that there was no systematic difference in variation across steps and conditions. Inspection of the box plots for penultimate foot position, final foot position, single-limb support, ascent duration, medial–lateral variability of foot movement, medial–lateral variability of trunk movement, cumulative distance of the foot and cumulative distance of the trunk all showed no systematic difference in variation across steps and conditions.

Discussion

The H–V illusions led to highly statistically significant increases in vertical toe clearance of the pertinent step edge, with average increases of 1.0 cm representing an increase of 17.5% over the control step mean toe clearance of 5.8 cm. Despite these increases in vertical toe clearance with the illusions on the bottom only, top only or top and bottom steps, there were no changes in vertical toe clearance for the middle step, which remained at about 5.0 cm for all visual illusion conditions. Thus, with the illusion on the bottom step only, vertical toe clearance increased over the bottom step to 7.3 cm (SD 1.6 cm) [control 6.3 cm (SD 2.1 cm)], yet returned to near control values of 5.0 cm (1.3 cm) and 5.3 cm (1.9 cm) for the middle and top steps [control 5.2 cm (SD 1.4 cm) and 5.3 cm (SD 2.0 cm)]. This provided an effect size of 0.48 [(7.3 – 6.3)/2.1]. With the illusion on the top step only, vertical toe clearance was similar to control values at 5.8 cm (SD 1.9 cm) and 5.0 cm (SD 1.4 cm) for the bottom and middle steps [control 6.3 cm (SD 2.1 cm) and 5.2 cm (SD 1.4 cm)], but then increased over the top step to 6.3 cm (SD 1.9 cm) (control 5.3 cm, SD 2.0 cm). This provided an effect size of 0.53 [(6.3 – 5.3)/1.9]. With the illusion on both the bottom and top steps, vertical toe clearance was increased over the bottom step [7.5 cm (SD 1.9 cm) vs. 6.3 cm (SD 2.1 cm) control], then returned to near control levels of vertical toe clearance of 5.0 cm (SD 1.4 cm) [control 5.2 cm (SD 1.4 cm)], then increased again over the top step to 6.1 cm (SD 1.9 cm) [control 5.3 cm (SD 2.0 cm)]. These provide effect sizes of 0.57 [(7.5 – 6.3)/2.1] for the bottom step, which is reduced to 0.40 [(6.1 – 5.3/2.0] for the top step.

There was a statistically significant repetition effect during the study, with vertical toe clearance becoming smaller with increased repetition, despite several aspects of the study design being aimed at reducing this effect. This is probably caused by the increased familiarity with the step heights provided by the somatosensory system leading to reduced vertical toe clearance to conserve energy and, to some degree, this is unavoidable in these adaptive gait studies. Certainly, the repetition effect did not obscure the effect of the H–V illusions on vertical toe clearance, which was relatively large. As previously discussed, the level of toe clearance increase at 1.0 cm in the current study represents a 17.5% level of increase and dangerous levels of toe clearance are reported when they are within 0.5 cm of the step edge,36 so that these figures actually represent relatively large changes in adaptive gait. They certainly suggest that the use of this illusion on stairways could lead to increased foot clearances and much less chance of tripping.

There was no indication that the H–V illusions lead to changes in other measures of adaptive gait and there was no change in penultimate or final foot placement, step ascent duration or within-subject variability across all dependent variables. In addition, the mismatch of information from the ‘tricked’ visual system and the somatosensory system did not appear to cause any problems, with no statistically significant mediolateral displacement of the foot and trunk during single-limb support for example. This suggests that the addition of the illusions to the steps does not appear to have in any way caused unsafe gait adaptations.

The results are very positive as they indicate that these relatively simple modifications to step appearance can significantly increase foot clearance over step edges during stair ascent by about 1.0 cm (≈17.5%), which could possibly make the difference between clearing a step edge or hitting it and potentially tripping. In addition, there appeared to be no other changes in significant aspects of adaptive gait plus no obvious indication of problems caused by different inputs from the somatosensory system and the ‘tricked’ visual system. Given that falls typically occur on the first and last few steps of a stairway, the use of the H–V illusion on the top and bottom step or steps seems the most useful.

Limitations of the study include the small sample size (although the very clear effect of the illusion has somewhat overcome this), the fact that the study only used a three-step stairway, whereas most stairways include significantly more steps, and that all participants were fit and healthy older adults, whereas most fallers are frail with several risk factors for falls. Although the results are very encouraging, to determine whether or not the illusions actually prevent falls on stairways, an assessment of falls rate on stairways with and without the illusions would need to be performed.

Full papers in peer-reviewed journals

Foster RJ, De Asha AR, Reeves ND, Maganaris CN, Buckley JG. Stair-specific algorithms for identification of touch-down and foot-off when descending or ascending a non-instrumented staircase. Gait Posture 2014;39:816–21.

Foster RJ, Hotchkiss J, Buckley JG, Elliott DB. Safety on stairs: Influence of a tread edge highlighter and its position. Exp Gerontol 2014;55:152–8.

Elliott DB, Foster RJ, Whitaker D, Scally A. Buckley JG. What you see is what you step: The horizontal-vertical illusion increases toe clearance in older adults during stair ascent. Invest Ophthalmol Visual Sci 2015;56:2950–7.

Conference presentations

Foster RJ, Hotchkiss J, Buckley JG, Elliott DB. Safety on stairs: influence of a tread edge highlighter and its position. World Conference of the International Society for Posture & Gait Research (ISPGR), Vancouver, BC, 29 June–3 July 2014.

Foster RJ, De Asha AR, Reeves ND, Maganaris CN, Buckley JG. Stair-specific algorithms for identification of touch-down and foot-off when descending or ascending a non-instrumented staircase. World Conference of ISPGR, Vancouver, BC, 29 June–3 July 2014.