Kinematic analysis of video captured falls experienced by older adults in long term care

On the one hand, real-life falls in older adults may generate impact velocities that are higher than those observed in laboratory-based falls in healthy young adults, due to age-related

Trang 1

Kinematic analysis of video-captured falls experienced by older adults

in long-term care

W.J Choia,n, J.M Wakelingb, S.N Robinovitchb,c

a

Department of Physical Therapy, Chapman University, 9401 Jeronimo Rd, Irvine, CA 92618, USA

b Department of Biomedical Physiology and Kinesiology, Simon Fraser University, Burnaby, BC, Canada

c

School of Engineering Science, Simon Fraser University, Burnaby, BC, Canada

a r t i c l e i n f o

Article history:

Accepted 15 February 2015

Keywords:

Falls

Biomechanics

Older adults

Hip fracture

Impact velocity

Injury

Kinematic analysis

Video

a b s t r a c t

Falls cause 95% of hip and wrist fractures and 60% of head injuries in older adults Risk for such injuries depends in part on velocity at contact, and the time available during the fall to generate protective responses However, we have no information on the impact velocities and durations of falls in older adults We addressed this barrier through kinematic analysis of 25 real-life falls (experienced by 23 individuals of mean age 80 years (SD¼9.8)) captured on video in two long-term facilities

All 25 falls involved impact to the pelvis, 12 involved head impact, and 21 involved hand impact We determined time-varying positions by digitizing each video, using direct linear transformations calibrated for each fall, and impact velocities through differentiation

The vertical impact velocity averaged 2.14 m/s (SD¼0.63) for the pelvis, 2.91 m/s (SD¼0.86) for the head, and 2.87 m/s (SD¼1.60) for the hand These values are 38%, 28%, and 4% lower, respectively, than predictions from an inverted pendulum model Furthermore, the average pelvis impact velocity was 16% lower than values reported previously for young individuals in laboratory falling experiments The average fall duration was 1271 ms (SD¼648) from the initiation of imbalance to pelvis impact, and

583 ms (SD¼255) from the start of descent to pelvis impact

Theseﬁrst measures of the kinematics of falls in older adults can inform the design and testing of fall injury prevention interventions (e.g., hip protectors, helmets, andﬂooring)

1 Introduction

Falls are the number one cause of injuries in older adults,

including at least 90% of hip fractures and wrist fractures, and 60%

of head injuries (Grisso et al., 1990; Harvey and Close, 2012; Palvanen

et al., 2000) The risk for injury during a fall depends in part on the

velocity at contact (or“impact velocity”) of the impacting body parts

(Majumder et al., 2008; Robinovitch et al., 1991) Accordingly, impact

velocity is a key input parameter for biomechanical testing of fall

injury prevention technology (e.g., hip protectors (Mills, 1996; Minns

et al., 2004a; Robinovitch et al., 2009), helmets (ASTM, 2007), and

compliantﬂooring (Knoefel et al., 2013;Laing and Robinovitch, 2009;

Minns et al., 2004b)) Risk for injury during a fall may also depend on

the time duration of the fall, which governs the faller's ability to

initiate and execute protective responses, such as arresting the fall

with the upper limbs (DeGoede et al., 2001; Robinovitch et al., 2005)

Our current knowledge of the impact velocities and durations associated with falls is limited to the results of laboratory studies with young adults falling (from standing height) onto gym mats (Feldman and Robinovitch, 2007; Hsiao and Robinovitch, 1998; Robinovitch et al., 2003) However, the kinematic patterns observed in lab-based falling experiments with young adults may differ substantially from real-life falls in older adults, due to differences in the situational and environmental context of falls,

or age-related changes in physiological factors such as physical and cognitive function, medication use, and disease On the one hand, real-life falls in older adults may generate impact velocities that are higher than those observed in laboratory-based falls in healthy young adults, due to age-related declines in sensorimotor and cognitive function, and a corresponding absence or declines in balance recovery responses (e.g., stepping or grasping) and fall protective responses (e.g., upper limb fall arrest) On the other hand, the perturbation conditions associated with real-life falls, which tend to be caused by internal versus external perturbations (Robinovitch et al., 2013), may cause them to less severe than laboratory-based falls which have used large external perturba-tions to overcome participants' ability to recover balance

Contents lists available atScienceDirect

journal homepage:www.elsevier.com/locate/jbiomech

www.JBiomech.com Journal of Biomechanics

http://dx.doi.org/10.1016/j.jbiomech.2015.02.025

n Corresponding author.

E-mail address: wchoi@chapman.edu (W.J Choi).

Trang 2

Our purpose in the current study was to document the impact

velocities of key body sites (hip, head and hand) during real-life falls

in older adults We collected and conducted kinematic analysis of

video footage of real-life falls experienced by older adults in

long-term care (LTC) facilities We then delong-termined the impact velocities

of the hip, head and hand, and the time duration of the falls We

compare our results to previous studies with young adults, and to

theoretical predictions from simple mathematical models

2 Methods

2.1 Real-life fall library

This study builds on recent descriptive reports by our team on the circumstances of

falls captured on video cameras in two long-term care (LTC) facilities ( Robinovitch et al.,

2013; Yang et al., 2013 ) Between April 2007 and February 2013, we partnered with two

LTC facilities in the Vancouver area (Delta View, a 312-bed facility in Delta, BC, and New

Vista, a 236-bed facility in Burnaby, BC) to capture video footage of real-life falls in older

adults ( Robinovitch et al., 2013 ) Delta View had a network of 216 digital cameras, while

New Vista had 48 Cameras were located in common areas (dining rooms, hallways, and

lounges) and not bedrooms or bathrooms Each video was recorded at 30 frames

per second and a resolution of 640 by 480 pixels or 720 by 480 pixels ( Fig 1 ) The study

was approved by the Ofﬁce of Research Ethics at SFU Each resident provided written

consent to the facilities for video capture of their images, and these data were shared as

secondary data with the research team Additional written consent was secured from

individuals to share their images.

Each fall video was initially analyzed by a team of three experts (research

assistants and graduate students trained by co-author SNR) using a structured,

validated questionnaire ( Yang et al., 2013 ) The questionnaire probed the cause of fall,

the activity at the time of the fall, the initial direction of the fall (forward, backward,

sideways, or straight down), the landing conﬁguration (forward, backward, or

sideways), the occurrence (if any) of stepping responses, and the occurrence of

impact to the hand(s), knee(s), head and pelvis.

2.2 Fall duration The questionnaire required the team to estimate the exact video frames corresponding to the onset of imbalance leading to the fall, the initiation of the descent stage of the fall (defined as one video frame after the foot contacted the ground in the last recovery step, if any), and the first occurrence of impact to the hand (s), head and pelvis We report two estimates of fall duration before impact to the body parts: the “total fall duration”, defined by the interval between the onset of imbalance and initial impact to the body part, and the “descent duration,” defined as the interval between the onset of fall initiation and impact.

2.3 Impact velocities

To estimate impact velocities, we manually digitized landmark of the pelvis (anterior superior iliac spine), head (ear or forehead) and hand (palm), using a Matlab routine developed by Hedrick (2008) We digitized each frame over the interval starting one frame before fall initiation and ending one frame after impact of the corresponding body part ( Fig 1 (a)) We then applied a two-dimensional direct linear transform (2D DLT) to reconstruct those points as position coordinates in the object space ( Hedrick, 2008; Meershoek, 1977 ) Finally, we used finite difference to estimate time-varying vertical and horizontal velocities The resulting velocity—time traces were fit with a fifth-order polynomial ( Fig 2 ) using Matlab's polyfit function, the approach used in falling experiments by van den Kroonenberg et al (1996) to fit vertical displacement versus time curves for the hip during falls from standing The vertical impact velocity was estimated as the maximum value of the curve fit, based on previous observations (and theoretical considerations) that the peak downward velocity occurs very near to the instant of contact ( Feldman and Robinovitch, 2007; Hsiao and Robinovitch, 1998 ) We also report values of the peak horizontal velocity and the magnitude of horizontal velocity at the instant of peak vertical velocity While our 2D DLT procedure corrected for lens distortion, an important limitation of the technique is the potential for “perspective errors” arising when the digitized points of interest move outside the calibrated image plane.

In an attempt to minimize these errors, we determined DLT calibration coefficients specific to each fall video, by visiting the site of each fall, and recording images of a flat calibration panel from the surveillance camera that captured the fall

Fig 1 Video snapshots showing: (a) forward fall by older adult in long-term care (LTC); (b) forward fall by young adult in the laboratory environment; (c) backward fall by older adult in LTC; and (d) backward fall in young adult in the lab The far-right panel illustrates the 25-marker calibration panel, placed at the exact location of the fall (in the

Trang 3

(see far-right images in Fig 1 ) The panel had dimensions 160 cm 160 cm, and

contained a 5 5 grid of 10 cm diameter circular markers spaced 40 cm apart The

panel was placed with the bottom surface ﬂush to the ground, centered at the

midpoint of the faller's feet (at the moment of fall initiation), and oriented in the plane

of the fall (deﬁned by the line connecting the mid-point of the feet and the location of

the head at the moment of pelvis impact).

2.4 Laboratory measures of accuracy

We tested the accuracy of our velocity estimates through laboratory falls with

an inverted pendulum and a human participant Each trial was captured with an

8-camera motion capture system recording at 250 Hz (Motion Analysis Corp., Santa

Rosa, CA, USA), and a single surveillance camera identical to the type used in our

partnering LTC sites recording at 30 Hz (model WVC210, Cisco Systems, San Jose, CA).

The surveillance camera was placed at a height of 2.6 m and horizontal distance of 5 m

from the site of the fall, which was typical of the falls we captured in the two long term

care facilities (although there was variability in this distance for the real-life falls).

The pendulum consisted of a 1.57 m long aluminum rod of uniform mass

distribution, connected to a low-friction hinge joint at the ﬂoor Reﬂective markers

were placed on the midpoint (representing the pelvis) and top end (head) of the

rod The pendulum was released from vertical and descended over a 901 arc before

impacting the ground.

Human falls were conducted for three falling directions: forward, backward,

and sideways ( Fig 1 ) In all trials, participants self-initiated the fall, and were

instructed to fall naturally Reﬂective markers were placed on the anterior superior

iliac spines, greater trochanters, sacrum, wrists, elbows, shoulders and forehead.

Trials were conducted at ﬁve falling angles with respect to the axis of the

surveillance camera: 301 (nearly toward the camera), 601, 901 (perpendicular to the

camera axis, providing a sideways view), 1201, and 1501 (nearly away from the camera).

A single trial was acquired for each camera angle, in (for human falls) each of the three

fall directions.

2.5 Inclusion criteria for video-captured falls in older adults

We used the results from our laboratory-based human falls to guide the

selection of video-captured falls in older adults for analysis, based on fall direction.

In particular, for each trial, we calculated the offset error, deﬁned as the difference

between the impact velocity estimate from 3D motion capture and the 2D single

video camera images We then determined, for a given fall direction, the mean and

standard deviation in the offset error across the five camera angles We regarded the technique as acceptable for a given fall direction if the standard deviation in the offset error was 0.35 m/s or less, which would reflect a 95% confidence interval less than or equal to 70.7 m/s in the estimated impact velocity We regarded the mean offset error (for a given direction) as a fixed bias, and subtracted direction-specific mean errors from peak velocities estimated from video analysis.

We also excluded falls with movements that were clearly different than those involved in our calibration trials These included falls with signiﬁcant rotation during descent, falls directly toward or away from the camera, falls directed straight down, and falls involving impact to objects other than the ﬂoor (e.g., walls or furniture) We also excluded falls not involving pelvis impact, or where the pelvis was occluded from camera view during descent or impact, since our primary objective at the onset of the study was characterizing the severity of impact to the pelvis during falls.

2.6 Analysis of video-captured falls in older adults

We provide descriptive results (means and standard deviations) for the vertical and horizontal impact velocity of the pelvis, head and hand We focus our attention more on vertical than horizontal velocity, as the stronger indicator of risk for serious injury ( van den Kroonenberg et al., 1996 ) We also report total fall durations and descent durations for each body part We compare measured vertical velocities to theoretical estimates based on free fall of a falling mass (where impact velocity¼ pffiffiffiffiffiffiffiffiffi2gh

, where g is gravitational acceleration of 9.81 m/s 2

, and h is the vertical descent distance of the body part), and an inverted pendulum with uniformly-distributed mass (where impact velocity¼pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi) We also conducted regression analysis to test whether impact velocity

is associated with fall height.

3 Results 3.1 Laboratory falls

In our laboratory experiments with the inverted pendulum (Fig 3), the average difference (across theﬁve falling angles) between the 2D DLT and 3D motion capture techniques was 0.008 m/s (SD¼0.04) for the vertical impact velocity, and 0.11 m/s (SD¼0.07) for the

Fig 2 Traces from typical trials of the velocity of the pelvis versus time for: (a) forward fall by an older adult in long-term care (LTC); (b) backward fall by an older adult in LTC; (c) forward fall by a young adult in the lab; and (d) backward fall by a young adult in the lab Vertical velocities are shown in dashed lines/ circles, and horizontal velocities in solid lines/squares.

Trang 4

Fig 3 Agreement from laboratory experiments between the 2D DLT technique and 3D motion capture for vertical and horizontal impact velocities Results are shown for the mid-point of a 1.57 m length pendulum, and for the pelvis for a human participant falling in the forward, backward, and sideways directions (at ﬁve different camera angles; see text for explanation) Sideways falls in human exhibited larger variability between camera angles (vertical velocity SD¼0.59 m/s, horizontal velocity SD¼0.54 m/s) than forward and backward falls The individual plots clearly show that 2 SD is less than 0.7 m/s in all but the sideways direction.

Trang 5

horizontal impact velocity (where a negative mean reﬂects

over-estimation from the 2D DLT technique)

In laboratory-based human falls (Fig 3), the DLT technique met

our criteria for acceptable accuracy for forward and backward falls,

but not sideways falls The mean difference in vertical impact

velocity between the 2D DLT and 3D motion capture techniques

was 0.26 m/s (SD¼0.21) for forward falls, 0.06 m/s (SD¼0.21)

for backward falls, and 1.01 m/s (SD was 0.59) for sideways falls

The mean difference in horizontal impact velocity was 0.15 m/s

(SD¼0.32) for forward falls, 0.16 m/s (SD¼0.18) for backward falls,

and 0.13 m/s (SD¼0.54) m/s for sideways falls For both human

and inverted pendulum falls, the accuracy was not inﬂuenced by the

falling angle with respect to the camera axis (p40.5 by linear

regression) For human falls, the mean error was not different

(p40.1) between forward and backward falls, and was equal to

0.16 m/s for both vertical and horizontal velocity The pooled SD

for forward and backward falls was 0.23 m/s for vertical velocity and

0.24 m/s for horizontal velocity

3.2 Falls by older adults

Between April 2007 and February 2013, we captured 813 falls

experienced by 306 individuals (Fig 4) Based on our acceptance

criteria and the results from our laboratory-based falling

experi-ments, we excluded cases where the initial fall direction was

sideways (n¼152) We excluded an additional 636 cases based on

other exclusion criteria, as described inFig 4

Ourﬁnal analysis included 25 falls (23 from Delta View, 2 from

New Vista) experienced by 23 older adults (Tables 1 and 2) There

were 21 backward falls and 4 forward falls, all of which involved

pelvis impact The average age of the faller was 80.3 yrs (SD¼9.8),

and 61% (n¼14) were female The most common cause of imbalance

was incorrect weight shifting (12 of 25 cases), followed by hit/bump (7 cases) Trips, collapses, and loss-of-support each accounted for

2 falls The most common activity at the time of falling was standing (14 of 25 cases), followed by walking (8 cases), and transferring from standing to sitting (3 cases) Two falls occurred while using a walker (video IDs 23 and 24) Stepping after the onset of imbalance occurred

in 16 of 25 falls

Head impact occurred in 48% of cases (n¼12; all 4 forward falls, and 8 of 21 backward falls), and hand impact occurred in 84% (n¼21; all 4 forward falls, and 17 of 21 backward falls) In 76% of cases involving hand impact (n¼16), the hand impacted before the pelvis or head All four forward falls involved impact to the knee (s) before the pelvis

Table 2 reports estimated impact velocities for each fall, after subtracting direction-speciﬁc mean offset errors (for vertical velocity: 0.26 m/s for forward falls and 0.06 m/s backward falls; for horizontal velocity: 0.15 m/s for forward falls and 0.16 m/s backward falls) Over the 25 falls by older adults, the vertical impact velocity averaged 2.14 m/s (SD¼0.63) for the pelvis, 2.91 m/s (SD¼0.86) for the head, and 2.87 m/s (SD¼1.60) for the hand (Table 2 and Figs 5 and 6) For eight backward falls involving impact to both the pelvis and head, the vertical impact velocity was 2.67 (SD¼0.82) for the head and 1.98 (SD¼0.45) m/s for the pelvis The horizontal impact velocity averaged 1.16 m/s (SD¼1.42) for the pelvis, 2.64 m/s (SD¼1.12) for the head, and 1.52 m/s (SD¼1.14) for the hand The total fall duration averaged 1271 ms (SD¼648) for the pelvis, 1730 ms (SD¼805) for the head, and 1188 ms (SD¼702) for the hand The descent duration averaged 593 ms (SD¼255) for the pelvis, 757 ms (SD¼217) for the head, and 479 ms (SD¼230) for the hand The vertical impact velocity of the pelvis averaged 2.19 m/s (SD¼0.61) in trials where the hand(s) impacted before the pelvis, compared to 2.41 m/s (SD¼0.85) in falls not involving hand

Trang 6

Table 1

Participant characteristics and descriptive data for 25 falls from 23 older adults.

Video

ID

Faller

ID

Age Sex Body mass

(kg)

Parkinson's disease

Alzheimer's disease

Stroke Hypertension COPD Diabetes Fall

direction

Pelvis impact

Head impact

Hand impact

Knee impact

Cause of fall

Activity

at the time of fall

Stepping response

Injury noted

Fall direction: F¼forward, B¼backward; Cause of fall: T¼trip/stumble, B¼hit/bump, IT¼incorrect transfer, LOS¼loss of support, C¼leg collapse; Activity at the time of fall: W¼walking, S¼standing, T¼transferring;¼missing data; COPD¼chronic obstructive pulmonary disease.

Trang 7

Impact velocities, fall durations, and fall heights for 25 falls from 23 older adults.

Video ID Faller

ID

Peak vertical velocity (m/s)

Peak horizontal velocity (m/s)

Horizontal velocity at peak vertical velocity (m/s)

Total fall duration (ms)

Descent duration (ms)

Fall height (cm)

Forward falls

Backward falls

Total fall duration¼time interval between the moment of imbalance and impact; Descent duration¼time interval between fall initiation and impact; Fall height¼vertical descent distance from fall initiation to impact;¼data not available.

a Not able to digitize.

Trang 8

impact, and 2.19 m/s (SD¼0.58) in trials where steps occurred

after imbalance versus 2.31 m/s (SD¼0.81) in falls not involving

steps (Table 2)

When compared to theoretical predictions based on free-fall

from each measured fall height (Table 2andFig 7), our vertical

pelvis impact velocities averaged 46.0% (SD¼14.95) lower than

predictions from a falling mass model, and 38.0% (SD¼17.3) lower

than predictions from an inverted pendulum model Similarly, our

vertical head and hand impact velocities averaged 37.4% (SD¼15.3)

and16.9% (SD¼55.4) lower than free-fall predictions, and 27.7%

(SD¼17.7) and 4.0% (SD¼63.9) lower than pendulum fall

predic-tions, respectively Furthermore, from regression analysis (SPSS,

Version 18.0), we found that fall height associated with the vertical

impact velocity of the head (v¼0.022nhþ0.5, R2¼0.403, p¼0.036),

but not of the pelvis (p¼0.19) or hand (p¼0.41)

4 Discussion

This is the ﬁrst study to our knowledge to report impact

velocities and fall duration from real-life falls in older adults Our

results provide important baseline measures of fall severity for the

design and assessment (through mechanical testing systems or

mathematical models) of interventions for fall injury prevention,

including wearable hip protectors, helmets and compliantﬂooring

Impact velocity is a measure of fall severity that is important for

the design and testing of injury prevention strategies Our measured

vertical impact velocities for older adults averaged 16% lower than

the mean value for the pelvis (of 2.55 m/s (SD¼0.85)) and 9% greater

than the mean value for the wrist (2.64 m/s (SD¼0.66)) reported by

Hsiao and Robinovitch (1998)from a laboratory study with young

adults falling on gym mats after receiving a sudden (slip)

perturba-tion This previous study included 20 backward and 11 sideways falls

in the analysis, and only reported average values for all trials, without

separating the results by fall direction (Hsiao and Robinovitch, 1998) Furthermore, our vertical impact velocities averaged 38% lower for the pelvis, and 28% lower for the head, than theoretical predictions from an inverted pendulum model, based on fall height Moreover, the fall height associated with vertical impact velocity of the head, but not of the pelvis or hand

These results suggest that, in the falls we analyzed, older adults utilized mechanisms to absorb energy during descent, and reduce their impact velocity (and risk for injury) These included attempts to recover balance by stepping (which occurred in 64% of falls), and impacting the ground with the hands before the pelvis or head (which occurred in 84% of falls) In previous falling experiments with young adults, pelvis impact velocities were decreased 22% by taking a step after imbalance, and 18% by impacting the hands before the pelvis (Feldman and Robinovitch, 2007) While our small sample precluded meaningful statistical analysis, our trends agree with these findings Pelvis impact velocities averaged 9% lower in falls involving hand impact (compared to no hand impact), and 5% lower in falls involving stepping (compared to no stepping) Additional mechan-isms beyond the scope of this study may have contributed to velocity reduction, including squatting during descent and contacting the pelvis with the trunk relatively upright (Robinovitch et al., 2004) The fall duration is of interest since it reflects the time available for the faller to initiate and execute protective responses to avoid injury during landing Our study shows that a duration of nearly 1200 ms is available to initiate protective responses between the moment of imbalance, and the instant of impact to the pelvis (which occurred on average at 1271 ms) or hand (which occurred on average at 1188 ms) This is considerably longer than the fall durations reported byHsiao and Robinovitch (1998)in their laboratory falls due to sudden slip perturbations, where the average interval between the onset of the perturbation and pelvic contact was 715 ms (SD¼160), while wrist contact was 680 ms (SD¼116) This reflects that real-life falls in older adults occur over a considerably slower time interval than falls

Fig 5 Mean values from the 25 falls in older adults (with standard deviations shown as error bars) of: (a) vertical impact velocity; (b) horizontal impact velocity; (c) total fall duration; and (d) descent duration In each case, values of shown for each of the pelvis, head and hand.

Trang 9

recorded in laboratory experiments with young adults, where a rather

severe, sudden perturbation is necessary to overcome balance

recov-ery responses Another study (Robinovitch et al., 2005) found that the

time required for older adults to move their hands into a protective

position to arrest a fall averages 615 ms (SD¼88) This is well below

our average value of total fall duration but similar to our mean

descent duration In our study, upper limb protective responses were

typically initiated soon after the onset of imbalance (Fig 6), which

likely contributed to the observation of hand impact in 84% of falls

There are important limitations to our study Our results are based

on falls experienced by residents in LTC, and may not apply to

healthier community dwelling older adults, or young adults Our small

sample size prevented us from examining how falls associate with

intrinsic factors such as physical and cognitive function or medications

Larger studies are required to relate the kinematics of falls to the

clinical context We only included falls that involved pelvis impact,

leading us to exclude many forward falls Our video footage was

collected at 30 fps, and therefore our resolution in detecting fall

initiation and impact times was limited to the duration of one frame

of the video (33 ms), or about one-ﬁfteenth (7%) of the shortest

descent duration we report (479 ms for the hand) We calculated

velocities afterﬁtting displacement versus time traces with a

ﬁfth-order polynomial, which may have resulted inﬁltering or loss of valid

kinematic information However, our approach was similar to that

used in previous video-based laboratory measures of fall impact

velocities in humans (van den Kroonenberg et al., 1996) Furthermore,

we found that aﬁfth-order polynomial provided the best agreement to

velocity estimates from 3D motion capture (recording at 250 Hz) in our inverted pendulum calibration tests Furthermore, analyzing the complex movements of falls from planar video is challenging, due to the out-of-plane motions of the body segments that often accompany during descent In our laboratory falls, we found acceptable accuracy

in our velocity estimates for forward and backward falls, where the trajectory of body parts tended to remain parallel to the calibration plane However, measurement accuracy was unacceptable for side-ways falls (where knee and trunkﬂexion often caused out-of-plane movement of the pelvis), which were excluded from analysis The exclusion of sideways falls was unfortunate given that hip fractures are most likely to occur from sideways falls (Greenspan et al., 1994; Nevitt and Cummings, 1993) Vertical impact velocities for the pelvis aver-aged 2.23 for backward falls and 1.66 m/s for forward falls, and one might expect similar values for sideways falls Future studies might test this hypothesis by capturing sideways falls with 3D cameras, or with multiple cameras and 3D analysis techniques

In summary, based on analysis of 25 video-captured falls experi-enced by 23 older adults in long-term care, we found that the vertical impact velocity averaged 2.14 m/s for the pelvis, and 2.91 m/s for the head These values are 38% and 28% lower, respectively, than theore-tical predictions from an inverted pendulum model based on fall height Furthermore, the average vertical impact velocity of the pelvis was 16% lower than values reported for young individuals in labora-tory falling experiments The duration of the fall averaged 1271 ms from the moment of imbalance, and 583 ms from the start of descent,

to the instant of pelvis impact Theseﬁrst measures of the kinematic

Fig 6 Sample traces of the vertical and horizontal velocity of the pelvis, head, and hand versus time for (a) forward fall and (b) backward fall by older adults in long-term care Both falls resulted in impact to the hand, pelvis, and head In the forward fall, impact occurred ﬁrst to the hand This seemed to reduce the subsequent vertical velocity

of the pelvis and head, which impacted the ground near-simultaneously During the backward fall, impact ﬁrst occurred near-simultaneously to the pelvis and hand This seemed to reduce the vertical velocity of the head, before a ﬁnal rapid increase as the trunk rotated downward.

Trang 10

proﬁles of real-life falls in older adults should inform the development

and testing of fall prevention technology, including wearable hip

protectors, helmets, and compliant ﬂooring, and contribute to the

design of exercise programs to train fall protective responses

Conﬂict of interest statement

None of the authors above have anyﬁnancial or personal

relation-ships with other people or organizations that could inappropriately

inﬂuence this work, including employment, consultancies, stock ownership, honoraria, paid expert testimony, patent applications/ registrations, and grants or other funding

Acknowledgments This research was supported by the operating grants from the Canadian Institutes for Health Research (funding reference nos AMG-100487 and TIR-103945)

References

ASTM A.S.f.T.a.M., 2007 Standard Performance Speciﬁcation for Ice Hockey Helmets.

DeGoede, K.M., Ashton-Miller, J.A., Liao, J.M., Alexander, N.B., 2001 How quickly can healthy adults move their hands to intercept an approaching object? Age and gender effects J Gerontol Ser A Biol Sci Med Sci 56, 584–588

Feldman, F., Robinovitch, S.N., 2007 Reducing hip fracture risk during sideways falls: evidence in young adults of the protective effects of impact to the hands and stepping J Biomech 40, 2612–2618

Greenspan, S.L., Myers, E.R., Maitland, L.A., Resnick, N.M., Hayes, W.C., 1994 Fall severity and bone mineral density as risk factors for hip fracture in ambulatory elderly J Am Med Assoc 271, 128–133

Grisso, J.A., Schwarz, D.F., Wishner, A.R., Weene, B., Holmes, J.H., Sutton, R.L., 1990 Injuries in an elderly inner-city population J Am Geriatr Soc 38, 1326–1331

Harvey, L.A., Close, J.C., 2012 Traumatic brain injury in older adults: characteristics, causes and consequences Injury 43, 1821–1826

Hedrick, T.L., 2008 Software techniques for two- and three-dimensional kinematic measurements of biological and biomimetic systems Bioinspir Biomim 3,

034001

Hsiao, E.T., Robinovitch, S.N., 1998 Common protective movements govern unex-pected falls from standing height J Biomech 31, 1–9

Knoefel, F., Patrick, L., Taylor, J., Goubran, R., 2013 Dual-stiffness ﬂooring: can it reduce fracture rates associated with falls? J Am Med Dir Assoc 14, 303–305

Laing, A.C., Robinovitch, S.N., 2009 Low stiffness ﬂoors can attenuate fall-related femoral impact forces by up to 50% without substantially impairing balance in older women Accid Anal Prev 41, 642–650

Majumder, S., Roychowdhury, A., Pal, S., 2008 Effects of trochanteric soft tissue thickness and hip impact velocity on hip fracture in sideways fall through 3D ﬁnite element simulations J Biomech 41, 2834–2842

Meershoek, L., 1977 Matlab Routines for 2D Analysis with Non-Perpendicular Camera Angle International Society of Biomechanics

Mills, N.J., 1996 The biomechanics of hip protectors Proc Inst Mech Eng Part H J Eng Med 210, 259–266

Minns, J., Dodd, C., Gardner, R., Bamford, J., Nabhani, F., 2004a Assessing the safety and effectiveness of hip protectors Nurs Stand 18, 33–38

Minns, J., Nabhani, F., Bamford, J.S., 2004b Can ﬂooring and underlay materials reduce hip fractures in older people? Nurs Older People 16, 16–20

Nevitt, M.C., Cummings, S.R., 1993 Type of fall and risk of hip and wrist fractures: the study of osteoporotic fractures The Study of Osteoporotic Fractures Research Group J Am Geriatr Soc 41, 1226–1234

Palvanen, M., Kannus, P., Parkkari, J., Pitkajarvi, T., Pasanen, M., Vuori, I., Jarvinen, M., 2000 The injury mechanisms of osteoporotic upper extremity fractures among older adults: a controlled study of 287 consecutive patients and their 108 controls Osteoporos Int 11, 822–831

Robinovitch, S.N., Brumer, R., Maurer, J., 2004 Effect of the “squat protective response” on impact velocity during backward falls J Biomech 37, 1329–1337

Robinovitch, S.N., Evans, S.L., Minns, J., Laing, A.C., Kannus, P., Cripton, P.A., Derler, S., Birge, S.J., Plant, D., Cameron, I.D., Kiel, D.P., Howland, J., Khan, K., Lauritzen, J.B.,

2009 Hip protectors: recommendations for biomechanical testing-an interna-tional consensus statement (Part I) Osteoporos Int 20, 1977–1988

Robinovitch, S.N., Feldman, F., Yang, Y., Schonnop, R., Leung, P.M., Sarraf, T., Sims-Gould, J., Loughin, M., 2013 Video capture of the circumstances of falls

in elderly people residing in long-term care: an observational study Lancet

381, 47–54

Robinovitch, S.N., Hayes, W.C., McMahon, T.A., 1991 Prediction of femoral impact forces in falls on the hip J Biomech Eng 113, 366–374

Robinovitch, S.N., Inkster, L., Maurer, J., Warnick, B., 2003 Strategies for avoiding hip impact during sideways falls J Bone Miner Res 18, 1267–1273

Robinovitch, S.N., Normandin, S.C., Stotz, P., Maurer, J.D., 2005 Time requirement for young and elderly women to move into a position for breaking a fall with outstretched hands J Gerontol Ser A Biol Sci Med Sci 60, 1553–1557

van den Kroonenberg, A.J., Hayes, W.C., McMahon, T.A., 1996 Hip impact velocities and body conﬁgurations for voluntary falls from standing height J Biomech.

29, 807–811

Yang, Y., Schonnop, R., Feldman, F., Robinovitch, S.N., 2013 Development and validation of a questionnaire for analyzing real-life falls in long-term care captured on video BMC Geriatr 13, 40

Fig 7 Comparison for the 25 falls by older adults between measured and model

predictions of vertical impact velocities (based on fall height) for (a) pelvis,

(b) head, and (c) hand On average, the measured vertical impact velocities for

the pelvis were 48.0% (SD¼14.2) lower than free-fall model predictions and 40.0%

(SD¼16.4) lower than inverted pendulum predictions The vertical impact

velo-cities for the head were 38.4% (SD¼16.5) and 28.8% (SD¼19.0) lower than free-fall

and inverted pendulum model predictions, respectively, and the vertical hand

impact velocities of the hand were 18.8% (SD¼55.3) and 6.3% (SD¼63.9) lower

than free-fall and inverted pendulum model predictions, respectively.

Định dạng
Số trang	10
Dung lượng	5,6 MB