- Email: [email protected]
Experimental identification of potential falls in older adult hospital patients
Journal of Biomechanics 49 (2016) 1016–1020
Contents lists available at ScienceDirect
Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com
Experimental identification of potential falls in older adult hospital patients Aimee Cloutier a, James Yang a,n, Debajyoti Pati b, Shabboo Valipoor b a b
Human-Centric Design Research Laboratory, Department of Mechanical Engineering, Texas Tech University, Lubbock, TX 79409, USA Department of Design, Texas Tech University, Lubbock, TX 79409, USA
art ic l e i nf o
a b s t r a c t
Article history: Accepted 5 February 2016
Patient falls within hospitals have been identified as serious but largely preventable incidents, particularly among older adult patients. Previous literature has explored intrinsic factors associated with patient falls, but literature identifying possible extrinsic or situational factors related to falls is lacking. This study seeks to identify patient motions and activities along with associated environmental design factors in a patient bathroom and clinician zone setting that may lead to falls. A motion capture experiment was conducted in a laboratory setting on 27 subjects over the age of seventy using scripted tasks and mockups of the bathroom and clinician zone of a patient room. Data were post-processed using Cortex and Visual3D software. A potential fall was characterized by a set of criteria based on the jerk of the upper body's center of mass (COM). Results suggest that only motion-related factors, particularly turning, pushing, pulling, and grabbing, contribute most significantly to potential falls in the patient bathroom, whereas only pushing and pulling contribute significantly in the clinician zone. Future work includes identifying and changing precise environmental design factors associated with these motions for an updated patient room and performing motion capture experiments using the new setup. & 2016 Elsevier Ltd. All rights reserved.
Keywords: Older adult patients Falls Patient bathroom and clinician zone Motion capture Motion jerk
1. Introduction Unintentional falls are the leading cause of unintentional injury in the United States (Centers for Disease Control and Prevention, 2015a). Among all age groups, the older adult population suffers the highest risk of falling and the highest risk of injury as a result of falling—nearly 2.5 million unintentional injuries were caused by falls in adults 65 or older in 2013 (Centers for Disease Control and Prevention, 2015a). Approximately 24,000 of these falls resulted in death (Centers for Disease Control and Prevention, 2015b), and 700,000 led to hospitalization, primarily due to a broken hip or head injury (Centers for Disease Control and Prevention, 2015c). Further, older adult falls cause significant financial strain on families and healthcare facilities with the direct medical cost of falls reaching $34 billion (Centers for Disease Control and Prevention, 2015d). A significant percentage of patient falls are identified as preventable as evidenced by the Centers for Medicare and Medicaid's classification of certain falls as “never events,” i.e. errors in medical care for which the risk of occurrence is significantly influenced by procedures of the healthcare organization. Consequently, hospitals n
Corresponding author. E-mail address: [email protected] (J. Yang).
http://dx.doi.org/10.1016/j.jbiomech.2016.02.012 0021-9290/& 2016 Elsevier Ltd. All rights reserved.
no longer receive reimbursement for treating falls that occur within patient rooms (CMS, 2008). This decision, prompted by the serious physical and financial consequences of falls, indicates that it is essential to prevent patient falls whenever possible. The causes of falls are complex and may primarily be categorized as intrinsic, those related to the characteristics of the patient, or extrinsic, those related to environmental conditions. Specific intrinsic and extrinsic factors affecting falls are identified in Hignett and Masud (2006). Other causes of falls include situational activities, such as leaning forward or reaching (Tinetti and Speechley, 1989), and organizational factors related to staffing, policies, and available equipment (Currie, 2008). Roughly 70% of falls are unwitnessed; thus, existing research on patient falls draws data from reports of patient falls (Lee et al., 2011) or creates assessments to determine the patient-specific risk of fall (Kato et al., 2013). Analysis is limited primarily to intrinsic factors (Haines and Waldron, 2011), and literature examining extrinsic factors and situational activities is lacking. Preferable falls prevention solutions include those which minimize the risk of falling while limiting mobility restrictions on the patient (Tinetti and Speechley, 1989). Several previous studies aimed to develop devices for the detection and prevention of falls (Debard et al., 2012; Ferrari et al., 2012; Lockhart et al., 2010; Ni et al., 2012; Wang et al., 2004) or to assess biomechanical factors which may lead to falls (Robinson et al., 1998; Yang and Pai, 2011,
A. Cloutier et al. / Journal of Biomechanics 49 (2016) 1016–1020
2014; Yang et al., 2012). Others have focused on how specific extrinsic factors, such as the bed rail height (Van Leeuwen et al., 2001), floor type (Simpson et al., 2004), or lighting conditions (Heung et al., 2010) influence patient falls. It should be noted that significant differences have been observed between real-world and simulated falls (Klenk et al., 2011), and experimental protocols should be adapted to reflect real-world falls accurately. This study aims to identify motions and activities, with associated environmental design factors, within the clinician zone and patient bathroom which may lead to falls. Motion capture data were collected on subjects performing a natural progression of tasks identified as high risk for falls. No subjects fell throughout the course of the experiment, but a set of criteria based on the jerk, i.e. the rate of change of acceleration, of the upper body center of mass (COM) was used to determine the start and end points for potential falls. Complementary information pertaining to study development and possible design solutions may be found in Pati et al. (under review). 2. Methods 2.1. Participants 27 Subjects over the age of seventy volunteered for this study (sex: 11 males, 16 females; age: 78.375.16 years; height: 165.1711.3 cm; body mass: 80.8717.6 kg, where 7 indicates standard deviation). Nine subjects were assigned to each of three
1017
scenarios: a clinician zone mockup or one of two bathroom mockups. The subjects were required to be physically and mentally sound, able to perform the scripted tasks without any assistance and not be on any medication that might hamper their performing the tasks. The experimental procedures were approved by the Institutional Review Board of Texas Tech University, and all subjects gave written informed consent. 2.2. Experimental protocol At the beginning of each experiment day, the motion capture system (eight Eagle-4 camera system from Motion Analysis, capture volume of 3.048 3.048 2.438 m3, capture rate of 120 frames/s) was calibrated to minimize error from motion of the cameras due to building vibrations and slight tripod adjustments. Subjects were provided with tight-fitting clothing for the motion capture experiments. After putting on the clothing, height and weight measurements were taken along with traditional limb length measurements according to the conventional gait model (Davis et al., 1991). The in-house marker placement protocol from Cloutier et al. (2011) was modified for use in this experiment. A total of 62 reflective markers were placed on each subject's body as shown in Fig. 1. Subjects were strapped into a fall arrest harness system (McMaster-Carr, 181.437 kg capacity, 3.658 7.315 m2 floor area covered) before beginning trials as an added safety measure in case a fall occurred. To prevent fatigue, subjects were instructed to take breaks as frequently as needed. 2.2.1. Patient bathroom trials Due to the limited size of the motion capture volume, tasks were split into clinician zone tasks and bathroom tasks. Two different locations of the bathroom (to the right or left side of the bed) were tested as shown in Fig. 2. Mockups for each bathroom location were created independently, and nine subjects per mockup participated in the bathroom trials (Mockup 1—Subject 1–9, sex: 4 males, 5 females; age: 75.3 7 5.17 years; height: 161.9 710.2 cm; body mass: 79.4 720.1 kg; Mockup 2—Subject 10–18, sex: 5 males, 4 females; age: 78.4 74.45 years; height: 1717 13.4 cm; body mass: 84.7 7 20.4 kg). Subjects began seated on the hospital
Fig. 1. (a) Marker placement protocol with harness (front view). (b) Marker placement protocol with harness (back view).
Fig. 2. (a) Bathroom layout to the right of bed, performed by subjects 1–9. (b) Bathroom layout to the right of bed, performed by subjects 10–18.
1018
A. Cloutier et al. / Journal of Biomechanics 49 (2016) 1016–1020
bed and were instructed by a lab member reading from a script through a natural progression of tasks including entering/leaving the bathroom, opening/closing the bathroom door, lifting/lowering the toilet seat, sit/stand for the toilet, and washing hands. Six distinct combinations of tasks were identified for bathroom trials based on two positions of the intravenous (IV) pole (attached with tape to the right or left hand) and three conditions for the door as shown in Fig. 2, where doors 2 and 3 have the same location but swing in opposite directions. Each combination was repeated by the subject three times; eighteen trials were recorded per subject in a random order.
2.2.2. Clinician zone trials The clinician zone includes the area around the bed used by clinical staff for direct patient care. Nine subjects (subject 19 27, sex: 2 males, 7 females; age: 8174.69 years; height: 162.378.47 cm; body mass: 78.3712.5 kg) participated in tasks pertaining to the clinician zone, whose layout is shown in Fig. 3. Subjects began seated on the hospital bed and were instructed by a lab member reading from a script through a natural progression of tasks including sit/stand for the patient bed and patient chair, walking to/from the bathroom, and drinking from a cup of water on the bedside table. Eight distinct combinations of tasks were identified for clinician zone trials based on two positions of the IV pole (attached with tape to the right or left hand) and four locations for the door. Each combination was repeated three times by the subject; 24 trials were recorded per subject in a random order.
2.3. Data processing A total of 540 trials were recorded. One of every three trials was chosen for data processing. Motion capture trials were first processed using Cortex, the postprocessing software provided by Motion Analysis. Post-processing in Cortex consisted of marker labeling and virtual data filling when marker occlusion occurred. Data were then imported into Visual 3D, and using a body model created specifically for the in-house marker placement protocol, the trajectory of the upper body COM was identified for each trial.
2.4. Data analysis 2.4.1. Identification of potential falls The jerk squared of the upper body's COM was calculated in MATLAB using Eq. (1). Although jerk is not the most traditional method of determining the onset of falls, its success in previous literature as a metric for detecting falls points toward jerk as a useful parameter for falls identification (Bahon and Sole, 2009; Klenk et al., 2011). It was expected that a fall would be preceded by an increase in the jerk of the COM. Fig. 4 shows the jerk squared of the COM for one trial. Two types of high jerk data may be observed. Large, brief increases in the jerk squared (observed around the two-second mark) can be attributed to noise in data collection, which is easily observable from the corresponding frames in the recorded motion capture trial. More consistent increases in jerk are expected to be characteristic of a potential fall. Noisy data points were characterized as outliers based on common statistical methods (Devore et al., 2012). Fig. 4 shows the minor and major outlier thresholds and the median of the data. The start and end points of potential falls were determined using the following criteria: within a frame range, (1) 70% of data points must lie above the median; (2) a maximum of 15% of data points may be minor outliers and (3) a maximum of 10% of data points may be major outliers. Although only one potential fall is shown in Fig. 4, multiple potential fall identifications per trial are possible.
jerk squared ¼
!2 !2 !2 3 3 3 d x d y d z þ þ dt 3 dt 3 dt 3
2.4.2. Classification of potential falls The start and end points of the potential falls identified based on the jerk method were used to extract clips from the accompanying synchronized video recordings. Identified potential falls were generally less than a second long; so, the speed of video clips was reduced so that reviewers could analyze potential falls more closely. Video clips were reviewed by two teams, one composed of healthcare designers and the other composed of registered nurses. Reviewers were asked to identify locations, activities, and motions associated with each fall event in addition to possible design solutions to prevent the fall.
3. Results and discussion In total, 730 potential fall events were identified throughout the 200 trials. Table 1 shows a rank-ordered list of the motions identified by reviewers as associated with potential falls in the bathroom setting. Table 2 provides similar results for the clinician zone trials. Note that the total number of observations in the “frequency of observation” columns is far greater than the total number of falls observed. This is due to the fact that it was possible for multiple motions to be associated with a single potential fall. Tables 1 and 2 reveal pushing and pulling as major factors contributing to potential falls in both the clinician zone and bathroom settings. Additionally, when considering any type of turn (rather than turns of a specific degree), turning items account for 15% of all potential fall events in the bathroom (items 7, 10, 12, 17, 19, 21) and 14% of all potential fall events in the clinician zone (items 6, 10, 13, 19, 20, 21). Thus, turning, pushing, and pulling emerge as the most frequently observed triggers for potential falls in both environments. Two multivariate regression analyses were conducted to determine which motions, environmental factors, and design factors significantly affect potential fall events. Details of the analyses are provided in Pati et al. (under review). In the bathroom setting, some environmental and design factors were found to be significant, but when all significant factors were integrated into one model, only the motion-related factors turning, pushing, pulling,
ð1Þ
Fig. 4. Example of a potential fall for subject 5.
Fig. 3. (a) Clinician zone layout with IV and furniture on right side, performed by subjects 19–27. (b) Clinician zone layout with IV and furniture on left side, performed by subjects 19–27.
A. Cloutier et al. / Journal of Biomechanics 49 (2016) 1016–1020
and grabbing remained significant. For the clinician zone setting, only the motion-related factors pushing and pulling demonstrated statistical significance. Tables 1 and 2 provide percentages of the total number of observed potential fall motions but do not provide percentages of the total number of times the particular motion was performed, regardless of whether a potential fall was identified or not. This information would be useful because some motions, such as sitting and standing, occur less frequently than others, such as turning. Although turning as an action may result in more potential falls simply because it occurs more often, sitting and standing may yield more potential falls relative to the number of times sitting and standing occur. It is useful to note why these particular motions might increase the risk of fall. One major factor is the increase in required coefficient of friction between the shoe/foot and floor during certain motions. This factor has been identified as problematic in previous literature for pushing and pulling tasks (Lee et al., 1992) and turning tasks (Yamaguchi et al., 2012). Additionally, tasks which cause the center of gravity to move outside of the region of support from the feet are more likely to lead to instability and falls. Slip that occurs during turning causes the base of support and center of gravity to move in opposite directions (Yamaguchi et al., 2012); thus, a slip that occurs during turning is more likely to lead to fall than a slip during straight walking. While it is true that motions are partially influenced by intrinsic factors, they may also be partially considered as functions of the room design and environmental conditions. For this reason, each potential fall related to one of the significant motion-related factors was paired with its associated environmental factors. A large number of potential falls associated with IV pole management and available space suggests that the amount of space provided, while sufficient for moving under normal circumstances, may be too small for maneuvering with the Table 1 Rank-ordered frequencies of motions in the bathrooms associated with potential fall events. Item Motion category
Frequency of observation
Percentage of total observations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
418 395 270 242 219 212 167 131 124 116 96 87 81 77 68 25 24 17 16 13 (10þ 10) 8 7 6 (4 þ 4) (2 þ 2)
14 13 9 8 7 7 5 4 4 4 3 3 3 3 2 1 o1 o1 o1 o1 o1 o1 o1 o1 o1 o1
(1 þ 1þ 1)
o1
27
Pushing Pulling Grabbing Stretching Bending forward Walking Turn 30° Backing Leaning Turn 60° Twisting Turn 90° Sideways shuffle Repositioning Looking down Sitting Turn 180° Squatting Turn 120° Standing Shifting weight; Turn 150° Bending backward Backing Changing hands on IV pole Bending sideward; lifting Looking back; bending knees Cross-arm push/pull awkwardness; looking to side; reaching
1019
Table 2 Rank-ordered frequencies of motions in the clinician zone associated with potential fall events. Item Motion category
Frequency of observation Percentage of total observations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
249 220 174 158 134 99 87 82 73 69 63 62 56 47 (36 þ36)
14 12 9 9 7 5 4 4 4 3 3 3 3 2 2
29 15 12 9 7 5 2 (1 þ1þ 1þ 1þ 1þ 1þ 1þ1)
1 o1 o1 o1 o1 o1 o1 o1
16 17 18 19 20 21 22 23
Pushing Pulling Stretching Sitting Grabbing Turn 30° Leaning Walking Repositioning Turn 90° Twisting Backing Turn 60° Sideways shuffle Looking down; sideways walk Shifting weight Bending forward Squatting Turn 180° Turn 120° Turn 150° Starting to sit Leg lift; leg swing due to chair height; irregular gait to accommodate obstacle; turns her head only; simply raising glass; hand grab upside down; arm stretch across chest; turned twice—opposite direction
addition of the IV pole. Similarly, the layout of the room could be changed so that fewer turns are required in the free space. One major source of potential error is marker occlusion. During setup, steps were taken to reduce marker occlusion as much as possible. Unnecessary pieces of the bathroom frame were eliminated, the shower curtain was removed, and a backless chair was used instead of a traditional patient chair. Despite precautions, marker occlusion was still problematic on the lower body due to the limited number of cameras and the presence of waist-high obstacles in the volume. The loss of lower body data prompted researchers to explore alternate data processing techniques. While the original intent was to use the whole body COM, the researchers ran a linear regression between the trajectories for the whole body COM and the COM of the upper body only for seventeen trials where the lower body data was relatively complete. High R2 values (0.955 70.0564 with all; 0.974 7 0.0234 with removal of one outlier) indicate a close correlation between the two data sets. Data for the upper body COM only were used following this validation process. Other sources of error include atypical behavior from subjects and differences in movement due to restrictions from the harness. Identified potential falls where harness restriction was clearly a relevant factor were eliminated from consideration. Additionally, falls occur due to complex combinations of intrinsic, extrinsic, and situational factors. Differences in intrinsic factors have previously been shown to influence the locations of falls (Hignett et al., 2010). It is possible that intrinsic factors may also influence the types of motions which most often lead to potential falls. However, the list of intrinsic factors is extensive, and patients may possess multiple intrinsic factors in unique combinations. While it would be impractical to consider all of these intrinsic factors and their combinations, future design
1020
A. Cloutier et al. / Journal of Biomechanics 49 (2016) 1016–1020
recommendations may be strengthened by considering intrinsic factors associated with differences in patient falls (such as confusion and frailty) or which affect the largest number of older adult patients. It would also be useful to determine if the presented results remain true for patients below the age of 65. It is also recognized that potential falls often will not lead to real falls. It would be helpful to know if potential falls associated with certain motions lead to real falls more often than others; however, it is difficult to induce fall without a planned slip or trip, which would be unrealistic for a hospital setting. In this work, the jerk squared was chosen as the metric for identifying a potential fall. Future work may include isolating the movements identified as triggers for potential falls and assessing them using other common methods of determining the onset of falls, such as the center of gravity's location with respect to the stability region of the feet. Expanding the data analysis in this way may yield more complete information about which potential falls are likely to lead to real falls.
4. Conclusion In this study, a motion capture experiment was designed to assess activities, motions, and physical design elements which may lead to falls in older adult hospital patients. 27 subjects performed a series of tasks in one of three mock hospital settings (one clinician zone setting and two bathroom settings). A potential fall was identified as a period of time characterized by an increase in the jerk of the COM. In the patient bathroom, turning, pushing, pulling, and grabbing emerged as the only significant factors contributing to potential falls; in the clinician zone, only pushing and pulling emerged as significant factors. Future work involves identifying and changing precise environmental factors in the patient bathroom and clinician zone which may reduce the effects of these motions and performing motion capture experiments to test these new layouts. Another avenue of future work may be to isolate the movements most commonly identified as triggers for potential falls and to perform additional stability-related assessments to determine whether these identified movements remain significant.
Conflict of interest statement The findings and conclusions in this report are those of the authors and do not necessarily represent the view of the National Patient Safety Foundation. The authors also declare that they have no conflict of interest.
Acknowledgments This research work was funded by National Patient Safety Foundation Research Grants Program (2013).
References Bahon, C.A., Sole, G.V., 2009. A jerk threshold-based involuntary lateral movement algorithm. In: Proceedings of the IEEE Conference on Emerging Technologies & Factory Automation, Mallorca, Spain. Centers for Disease Control and Prevention. 2015a. National Estimates of the 10 Leading Causes of Nonfatal Injuries Treated in Hospital Emergency Departments, United States. 〈http://www.cdc.gov/injury/wisqars/leadingcauses.html〉 (accessed 20.10.15). Centers for Disease Control and Prevention. 2015b. 2013, United States Unintentional Fall Deaths and Rates per 1,00,000. cdc.gov. 〈http://webappa.cdc.gov/cgibin/broker.exe〉 (accessed 20.10.15).
Centers for Disease Control and Prevention. 2015c. Costs of Falls Among Older Adults. cdc.gov. 〈http://www.cdc.gov/homeandrecreationalsafety/falls/fallcost. html〉 (accessed 20.10.15). Centers for Disease Control and Prevention. 2015d. Important Facts about Falls. cdc.gov. 〈http://www.cdc.gov/homeandrecreationalsafety/falls/adultfalls. html〉 (accessed 25.01.15). Cloutier, A., Boothby, R., Yang, J., 2011. Motion capture experiments for validating optimization-based human models. In: Proceedings of the HCI International, 3rd International Conference on Digital Human Modeling. Orlando, Florida. CMS. 2008. Never Events. From 〈http://downloads.cms.gov/cmsgov/archiveddownloads/SMDL/downloads/SMD073108/pdf〉 (accessed 25.01.15). Currie, L., 2008. Fall and injury prevention. In: Hughes, R.G. (Ed.), Patient Safety and Quality: An Evidence-Based Handbook for Nurses. Agency for Healthcare Research and Quality, Rockville, MD. Davis, R., Ounpuu, S., Tyburski, D., Gage, J., 1991. A gait analysis data collection and reduction technique. Hum. Mov. Sci. 10, 575–587. http://dx.doi.org/10.1016/ 0167 9457(91)90046. Debard, G., Karsmakers, P., Deschodt, M., Vlaeyen, E., Dejaeger, E., Milisen, K., Tuytelaars, T., 2012. Camera-based fall detection on real world data. In: Dellaert, F., Frahm, J.-M., Pollefeys, M., leal-Taixe, L., Rosenhahn, B. (Eds.), Outdoor and Large-Scale Real-World Scene Analysis, 10. Springer, Berlin, Heidelberg, pp. 356–375. Devore, J.L., 2012. Probability and Statistics for Engineering and the Sciences. Brooks/Cole CENGAGE Learning, Boston, MA. Ferrari, M., Harrison, B., Rawashdeh, O., Hammond, R., Avery, Y., Rawashdeh, M., Maddens, M., 2012. Clinical feasibility trial of a motion detection system for fall prevention in hospitalized older adults. Geriatr. Nurs. 33 (3), 177–183. http://dx. doi.org/10.1016/j.gerinurse.2011.11.011. Haines, T.P., Waldron, N.G., 2011. Translation of falls prevention knowledge into action in hospitals: what should be translated and how should it be done? J. Saf. Res. 42 (6), 431–442 10.1016.j.jsr.2011.10.003. Heung, M., Adamowski, T., Segal, J.H., Malani, P.N., 2010. A successful approach to fall prevention in an outpatient hemodialysis center. Clin. J. Am. Soc. Nephrol. 5 (10), 1775–1779. http://dx.doi.org/10.2215/CJN.01610210. Hignett, S., Masud, T., 2006. A review of environmental hazards associated with inpatient falls. Ergonomics 49 (5 6), 605–616. http://dx.doi.org/10.1080/ 00140130600568949. Hignett, S., Sands, G., Youde, J., Griffiths, P., 2010. Targeting environmental factors to reduce elderly in-patient falls. In: Proceedings of the 1st International Conference on Human Factors and Ergonomics in Healthcare. Miami, Florida. Kato, S., Tsuru, S., Iizuka, Y., 2013. A structural model for patient fall risk and method for determining countermeasures. J. Qual. 20 (5), 503–520. http://dx. doi.org/10.6220/joq. Klenk, J., Becker, C., Lieken, F., Nicolai, S., Maetzler, W., Alt, W., Zijlstra, W., Hausdorff, J.M., van Lummel, R.C., Chiari, L., Lindemann, U., 2011. Comparison of acceleration signals of simulated and real-world backward falls. Med. Eng. Phys. 33 (3), 368–373. http://dx.doi.org/10.1016/j.medengphy.2010.11.003. Lee, K.S., Chaffin, D.B., Parks, C., 1992. A study of slip potential during cart pushing and pulling. IIE Trans. 24 (5), 139–146. http://dx.doi.org/10.1080/07408179208964254. Lee, T.-T., Liu, C.-Y., Kuo, Y.-H., Mills, M.E., Fong, J.-G., Hung, C., 2011. Application of data mining to the identification of critical factors in patient falls using a webbased reporting system. Int. J. Med. Inform. 80 (2), 141–150. http://dx.doi.org/ 10.1016/j.ijmedinf.2010.10.009. Lockhart, T.E., Barth, A.T., Zhang, X., Songra, R., Abdel-Rahman, E., Lach, J., 2010. Portable, Non-invasive Fall Risk Assessment in End Stage Renal Disease Patients on Hemodialysis. Wireless Health 2010, San Diego, California. Ni, B., Dat, N.C., Moulin, P., 2012. RGDB-camera based get-up event detection for hospital fall prevention. In: Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Kyoto, Japan. Pati, D., Valipoor, S., Cloutier, A., Yang, J. (Under review). Physical design factors contributing to patient falls. Journal of Patient Safety. Robinson, C.J., Purucker, M.C., Faulkner, L.W., 1998. Design, control, and characterization of a sliding linear investigative platform for analyzing lower limb stability (SLIPS-FALLS). IEEE Trans. Rehabil. Eng. 6 (3), 334–350. http://dx.doi.org/ 10.1109/IEMBS.1996.646538. Simpson, A.H., Lamb, S., Roberts, P.J., Gardner, T.N., Evans, J.G., 2004. Does the type of flooring affect the risk of hip fracture? Age Aging 33 (3), 242–246. http://dx. doi.org/10.1093/ageing/afh071. Tinetti, M., Speechley, M., 1989. Prevention of falls among the elderly. New Engl. J. Med. 320 (16), 1055–1058. http://dx.doi.org/10.1056/NEJM198904203201606. Van Leeuwen, M., Bennett, L., West, S., Wiles, V., Grasso, J., 2001. Patient falls from bed and the role of bedrails in the acute care setting. Aust. J. Adv. Nurs. 19 (2), 8–13. Wang, C.H., Hsueh, I.P., Sheu, C.F., Yao, G., Hsieh, C.L., 2004. Psychometric properties of 2 simplified 3-level balance scales used for patients with stroke. J. Am. Phys. Ther. Assoc. 84 (5), 430–438. Yamaguchi, T., Yano, M., Onodera, H., Hokkirigawa, K., 2012. Effect of turning angle on falls caused by induced slips during turning. J. Biomech. 45 (2012), 2624–2629. http://dx.doi.org/10.1016/j.jbiomech.2012.08.006. Yang, F., Pai, Y.C., 2011. Automatic recognition of falls in gait-slip training: harness load cell based criteria. J. Biomech. 44 (2011), 2243–2249. http://dx.doi.org/ 10.1016/j.jbiomech.2011.05.039. Yang, F., Espy, D., Bhatt, T., Pai, Y.C., 2012. Two types of slip-induced falls among community dwelling older adults. J. Biomech. 45 (2012), 1259–1264. http://dx. doi.org/10.1016/j.jbiomech.2012.01.036.
Contents lists available at ScienceDirect
Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com
Experimental identification of potential falls in older adult hospital patients Aimee Cloutier a, James Yang a,n, Debajyoti Pati b, Shabboo Valipoor b a b
Human-Centric Design Research Laboratory, Department of Mechanical Engineering, Texas Tech University, Lubbock, TX 79409, USA Department of Design, Texas Tech University, Lubbock, TX 79409, USA
art ic l e i nf o
a b s t r a c t
Article history: Accepted 5 February 2016
Patient falls within hospitals have been identified as serious but largely preventable incidents, particularly among older adult patients. Previous literature has explored intrinsic factors associated with patient falls, but literature identifying possible extrinsic or situational factors related to falls is lacking. This study seeks to identify patient motions and activities along with associated environmental design factors in a patient bathroom and clinician zone setting that may lead to falls. A motion capture experiment was conducted in a laboratory setting on 27 subjects over the age of seventy using scripted tasks and mockups of the bathroom and clinician zone of a patient room. Data were post-processed using Cortex and Visual3D software. A potential fall was characterized by a set of criteria based on the jerk of the upper body's center of mass (COM). Results suggest that only motion-related factors, particularly turning, pushing, pulling, and grabbing, contribute most significantly to potential falls in the patient bathroom, whereas only pushing and pulling contribute significantly in the clinician zone. Future work includes identifying and changing precise environmental design factors associated with these motions for an updated patient room and performing motion capture experiments using the new setup. & 2016 Elsevier Ltd. All rights reserved.
Keywords: Older adult patients Falls Patient bathroom and clinician zone Motion capture Motion jerk
1. Introduction Unintentional falls are the leading cause of unintentional injury in the United States (Centers for Disease Control and Prevention, 2015a). Among all age groups, the older adult population suffers the highest risk of falling and the highest risk of injury as a result of falling—nearly 2.5 million unintentional injuries were caused by falls in adults 65 or older in 2013 (Centers for Disease Control and Prevention, 2015a). Approximately 24,000 of these falls resulted in death (Centers for Disease Control and Prevention, 2015b), and 700,000 led to hospitalization, primarily due to a broken hip or head injury (Centers for Disease Control and Prevention, 2015c). Further, older adult falls cause significant financial strain on families and healthcare facilities with the direct medical cost of falls reaching $34 billion (Centers for Disease Control and Prevention, 2015d). A significant percentage of patient falls are identified as preventable as evidenced by the Centers for Medicare and Medicaid's classification of certain falls as “never events,” i.e. errors in medical care for which the risk of occurrence is significantly influenced by procedures of the healthcare organization. Consequently, hospitals n
Corresponding author. E-mail address: [email protected] (J. Yang).
http://dx.doi.org/10.1016/j.jbiomech.2016.02.012 0021-9290/& 2016 Elsevier Ltd. All rights reserved.
no longer receive reimbursement for treating falls that occur within patient rooms (CMS, 2008). This decision, prompted by the serious physical and financial consequences of falls, indicates that it is essential to prevent patient falls whenever possible. The causes of falls are complex and may primarily be categorized as intrinsic, those related to the characteristics of the patient, or extrinsic, those related to environmental conditions. Specific intrinsic and extrinsic factors affecting falls are identified in Hignett and Masud (2006). Other causes of falls include situational activities, such as leaning forward or reaching (Tinetti and Speechley, 1989), and organizational factors related to staffing, policies, and available equipment (Currie, 2008). Roughly 70% of falls are unwitnessed; thus, existing research on patient falls draws data from reports of patient falls (Lee et al., 2011) or creates assessments to determine the patient-specific risk of fall (Kato et al., 2013). Analysis is limited primarily to intrinsic factors (Haines and Waldron, 2011), and literature examining extrinsic factors and situational activities is lacking. Preferable falls prevention solutions include those which minimize the risk of falling while limiting mobility restrictions on the patient (Tinetti and Speechley, 1989). Several previous studies aimed to develop devices for the detection and prevention of falls (Debard et al., 2012; Ferrari et al., 2012; Lockhart et al., 2010; Ni et al., 2012; Wang et al., 2004) or to assess biomechanical factors which may lead to falls (Robinson et al., 1998; Yang and Pai, 2011,
A. Cloutier et al. / Journal of Biomechanics 49 (2016) 1016–1020
2014; Yang et al., 2012). Others have focused on how specific extrinsic factors, such as the bed rail height (Van Leeuwen et al., 2001), floor type (Simpson et al., 2004), or lighting conditions (Heung et al., 2010) influence patient falls. It should be noted that significant differences have been observed between real-world and simulated falls (Klenk et al., 2011), and experimental protocols should be adapted to reflect real-world falls accurately. This study aims to identify motions and activities, with associated environmental design factors, within the clinician zone and patient bathroom which may lead to falls. Motion capture data were collected on subjects performing a natural progression of tasks identified as high risk for falls. No subjects fell throughout the course of the experiment, but a set of criteria based on the jerk, i.e. the rate of change of acceleration, of the upper body center of mass (COM) was used to determine the start and end points for potential falls. Complementary information pertaining to study development and possible design solutions may be found in Pati et al. (under review). 2. Methods 2.1. Participants 27 Subjects over the age of seventy volunteered for this study (sex: 11 males, 16 females; age: 78.375.16 years; height: 165.1711.3 cm; body mass: 80.8717.6 kg, where 7 indicates standard deviation). Nine subjects were assigned to each of three
1017
scenarios: a clinician zone mockup or one of two bathroom mockups. The subjects were required to be physically and mentally sound, able to perform the scripted tasks without any assistance and not be on any medication that might hamper their performing the tasks. The experimental procedures were approved by the Institutional Review Board of Texas Tech University, and all subjects gave written informed consent. 2.2. Experimental protocol At the beginning of each experiment day, the motion capture system (eight Eagle-4 camera system from Motion Analysis, capture volume of 3.048 3.048 2.438 m3, capture rate of 120 frames/s) was calibrated to minimize error from motion of the cameras due to building vibrations and slight tripod adjustments. Subjects were provided with tight-fitting clothing for the motion capture experiments. After putting on the clothing, height and weight measurements were taken along with traditional limb length measurements according to the conventional gait model (Davis et al., 1991). The in-house marker placement protocol from Cloutier et al. (2011) was modified for use in this experiment. A total of 62 reflective markers were placed on each subject's body as shown in Fig. 1. Subjects were strapped into a fall arrest harness system (McMaster-Carr, 181.437 kg capacity, 3.658 7.315 m2 floor area covered) before beginning trials as an added safety measure in case a fall occurred. To prevent fatigue, subjects were instructed to take breaks as frequently as needed. 2.2.1. Patient bathroom trials Due to the limited size of the motion capture volume, tasks were split into clinician zone tasks and bathroom tasks. Two different locations of the bathroom (to the right or left side of the bed) were tested as shown in Fig. 2. Mockups for each bathroom location were created independently, and nine subjects per mockup participated in the bathroom trials (Mockup 1—Subject 1–9, sex: 4 males, 5 females; age: 75.3 7 5.17 years; height: 161.9 710.2 cm; body mass: 79.4 720.1 kg; Mockup 2—Subject 10–18, sex: 5 males, 4 females; age: 78.4 74.45 years; height: 1717 13.4 cm; body mass: 84.7 7 20.4 kg). Subjects began seated on the hospital
Fig. 1. (a) Marker placement protocol with harness (front view). (b) Marker placement protocol with harness (back view).
Fig. 2. (a) Bathroom layout to the right of bed, performed by subjects 1–9. (b) Bathroom layout to the right of bed, performed by subjects 10–18.
1018
A. Cloutier et al. / Journal of Biomechanics 49 (2016) 1016–1020
bed and were instructed by a lab member reading from a script through a natural progression of tasks including entering/leaving the bathroom, opening/closing the bathroom door, lifting/lowering the toilet seat, sit/stand for the toilet, and washing hands. Six distinct combinations of tasks were identified for bathroom trials based on two positions of the intravenous (IV) pole (attached with tape to the right or left hand) and three conditions for the door as shown in Fig. 2, where doors 2 and 3 have the same location but swing in opposite directions. Each combination was repeated by the subject three times; eighteen trials were recorded per subject in a random order.
2.2.2. Clinician zone trials The clinician zone includes the area around the bed used by clinical staff for direct patient care. Nine subjects (subject 19 27, sex: 2 males, 7 females; age: 8174.69 years; height: 162.378.47 cm; body mass: 78.3712.5 kg) participated in tasks pertaining to the clinician zone, whose layout is shown in Fig. 3. Subjects began seated on the hospital bed and were instructed by a lab member reading from a script through a natural progression of tasks including sit/stand for the patient bed and patient chair, walking to/from the bathroom, and drinking from a cup of water on the bedside table. Eight distinct combinations of tasks were identified for clinician zone trials based on two positions of the IV pole (attached with tape to the right or left hand) and four locations for the door. Each combination was repeated three times by the subject; 24 trials were recorded per subject in a random order.
2.3. Data processing A total of 540 trials were recorded. One of every three trials was chosen for data processing. Motion capture trials were first processed using Cortex, the postprocessing software provided by Motion Analysis. Post-processing in Cortex consisted of marker labeling and virtual data filling when marker occlusion occurred. Data were then imported into Visual 3D, and using a body model created specifically for the in-house marker placement protocol, the trajectory of the upper body COM was identified for each trial.
2.4. Data analysis 2.4.1. Identification of potential falls The jerk squared of the upper body's COM was calculated in MATLAB using Eq. (1). Although jerk is not the most traditional method of determining the onset of falls, its success in previous literature as a metric for detecting falls points toward jerk as a useful parameter for falls identification (Bahon and Sole, 2009; Klenk et al., 2011). It was expected that a fall would be preceded by an increase in the jerk of the COM. Fig. 4 shows the jerk squared of the COM for one trial. Two types of high jerk data may be observed. Large, brief increases in the jerk squared (observed around the two-second mark) can be attributed to noise in data collection, which is easily observable from the corresponding frames in the recorded motion capture trial. More consistent increases in jerk are expected to be characteristic of a potential fall. Noisy data points were characterized as outliers based on common statistical methods (Devore et al., 2012). Fig. 4 shows the minor and major outlier thresholds and the median of the data. The start and end points of potential falls were determined using the following criteria: within a frame range, (1) 70% of data points must lie above the median; (2) a maximum of 15% of data points may be minor outliers and (3) a maximum of 10% of data points may be major outliers. Although only one potential fall is shown in Fig. 4, multiple potential fall identifications per trial are possible.
jerk squared ¼
!2 !2 !2 3 3 3 d x d y d z þ þ dt 3 dt 3 dt 3
2.4.2. Classification of potential falls The start and end points of the potential falls identified based on the jerk method were used to extract clips from the accompanying synchronized video recordings. Identified potential falls were generally less than a second long; so, the speed of video clips was reduced so that reviewers could analyze potential falls more closely. Video clips were reviewed by two teams, one composed of healthcare designers and the other composed of registered nurses. Reviewers were asked to identify locations, activities, and motions associated with each fall event in addition to possible design solutions to prevent the fall.
3. Results and discussion In total, 730 potential fall events were identified throughout the 200 trials. Table 1 shows a rank-ordered list of the motions identified by reviewers as associated with potential falls in the bathroom setting. Table 2 provides similar results for the clinician zone trials. Note that the total number of observations in the “frequency of observation” columns is far greater than the total number of falls observed. This is due to the fact that it was possible for multiple motions to be associated with a single potential fall. Tables 1 and 2 reveal pushing and pulling as major factors contributing to potential falls in both the clinician zone and bathroom settings. Additionally, when considering any type of turn (rather than turns of a specific degree), turning items account for 15% of all potential fall events in the bathroom (items 7, 10, 12, 17, 19, 21) and 14% of all potential fall events in the clinician zone (items 6, 10, 13, 19, 20, 21). Thus, turning, pushing, and pulling emerge as the most frequently observed triggers for potential falls in both environments. Two multivariate regression analyses were conducted to determine which motions, environmental factors, and design factors significantly affect potential fall events. Details of the analyses are provided in Pati et al. (under review). In the bathroom setting, some environmental and design factors were found to be significant, but when all significant factors were integrated into one model, only the motion-related factors turning, pushing, pulling,
ð1Þ
Fig. 4. Example of a potential fall for subject 5.
Fig. 3. (a) Clinician zone layout with IV and furniture on right side, performed by subjects 19–27. (b) Clinician zone layout with IV and furniture on left side, performed by subjects 19–27.
A. Cloutier et al. / Journal of Biomechanics 49 (2016) 1016–1020
and grabbing remained significant. For the clinician zone setting, only the motion-related factors pushing and pulling demonstrated statistical significance. Tables 1 and 2 provide percentages of the total number of observed potential fall motions but do not provide percentages of the total number of times the particular motion was performed, regardless of whether a potential fall was identified or not. This information would be useful because some motions, such as sitting and standing, occur less frequently than others, such as turning. Although turning as an action may result in more potential falls simply because it occurs more often, sitting and standing may yield more potential falls relative to the number of times sitting and standing occur. It is useful to note why these particular motions might increase the risk of fall. One major factor is the increase in required coefficient of friction between the shoe/foot and floor during certain motions. This factor has been identified as problematic in previous literature for pushing and pulling tasks (Lee et al., 1992) and turning tasks (Yamaguchi et al., 2012). Additionally, tasks which cause the center of gravity to move outside of the region of support from the feet are more likely to lead to instability and falls. Slip that occurs during turning causes the base of support and center of gravity to move in opposite directions (Yamaguchi et al., 2012); thus, a slip that occurs during turning is more likely to lead to fall than a slip during straight walking. While it is true that motions are partially influenced by intrinsic factors, they may also be partially considered as functions of the room design and environmental conditions. For this reason, each potential fall related to one of the significant motion-related factors was paired with its associated environmental factors. A large number of potential falls associated with IV pole management and available space suggests that the amount of space provided, while sufficient for moving under normal circumstances, may be too small for maneuvering with the Table 1 Rank-ordered frequencies of motions in the bathrooms associated with potential fall events. Item Motion category
Frequency of observation
Percentage of total observations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
418 395 270 242 219 212 167 131 124 116 96 87 81 77 68 25 24 17 16 13 (10þ 10) 8 7 6 (4 þ 4) (2 þ 2)
14 13 9 8 7 7 5 4 4 4 3 3 3 3 2 1 o1 o1 o1 o1 o1 o1 o1 o1 o1 o1
(1 þ 1þ 1)
o1
27
Pushing Pulling Grabbing Stretching Bending forward Walking Turn 30° Backing Leaning Turn 60° Twisting Turn 90° Sideways shuffle Repositioning Looking down Sitting Turn 180° Squatting Turn 120° Standing Shifting weight; Turn 150° Bending backward Backing Changing hands on IV pole Bending sideward; lifting Looking back; bending knees Cross-arm push/pull awkwardness; looking to side; reaching
1019
Table 2 Rank-ordered frequencies of motions in the clinician zone associated with potential fall events. Item Motion category
Frequency of observation Percentage of total observations
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
249 220 174 158 134 99 87 82 73 69 63 62 56 47 (36 þ36)
14 12 9 9 7 5 4 4 4 3 3 3 3 2 2
29 15 12 9 7 5 2 (1 þ1þ 1þ 1þ 1þ 1þ 1þ1)
1 o1 o1 o1 o1 o1 o1 o1
16 17 18 19 20 21 22 23
Pushing Pulling Stretching Sitting Grabbing Turn 30° Leaning Walking Repositioning Turn 90° Twisting Backing Turn 60° Sideways shuffle Looking down; sideways walk Shifting weight Bending forward Squatting Turn 180° Turn 120° Turn 150° Starting to sit Leg lift; leg swing due to chair height; irregular gait to accommodate obstacle; turns her head only; simply raising glass; hand grab upside down; arm stretch across chest; turned twice—opposite direction
addition of the IV pole. Similarly, the layout of the room could be changed so that fewer turns are required in the free space. One major source of potential error is marker occlusion. During setup, steps were taken to reduce marker occlusion as much as possible. Unnecessary pieces of the bathroom frame were eliminated, the shower curtain was removed, and a backless chair was used instead of a traditional patient chair. Despite precautions, marker occlusion was still problematic on the lower body due to the limited number of cameras and the presence of waist-high obstacles in the volume. The loss of lower body data prompted researchers to explore alternate data processing techniques. While the original intent was to use the whole body COM, the researchers ran a linear regression between the trajectories for the whole body COM and the COM of the upper body only for seventeen trials where the lower body data was relatively complete. High R2 values (0.955 70.0564 with all; 0.974 7 0.0234 with removal of one outlier) indicate a close correlation between the two data sets. Data for the upper body COM only were used following this validation process. Other sources of error include atypical behavior from subjects and differences in movement due to restrictions from the harness. Identified potential falls where harness restriction was clearly a relevant factor were eliminated from consideration. Additionally, falls occur due to complex combinations of intrinsic, extrinsic, and situational factors. Differences in intrinsic factors have previously been shown to influence the locations of falls (Hignett et al., 2010). It is possible that intrinsic factors may also influence the types of motions which most often lead to potential falls. However, the list of intrinsic factors is extensive, and patients may possess multiple intrinsic factors in unique combinations. While it would be impractical to consider all of these intrinsic factors and their combinations, future design
1020
A. Cloutier et al. / Journal of Biomechanics 49 (2016) 1016–1020
recommendations may be strengthened by considering intrinsic factors associated with differences in patient falls (such as confusion and frailty) or which affect the largest number of older adult patients. It would also be useful to determine if the presented results remain true for patients below the age of 65. It is also recognized that potential falls often will not lead to real falls. It would be helpful to know if potential falls associated with certain motions lead to real falls more often than others; however, it is difficult to induce fall without a planned slip or trip, which would be unrealistic for a hospital setting. In this work, the jerk squared was chosen as the metric for identifying a potential fall. Future work may include isolating the movements identified as triggers for potential falls and assessing them using other common methods of determining the onset of falls, such as the center of gravity's location with respect to the stability region of the feet. Expanding the data analysis in this way may yield more complete information about which potential falls are likely to lead to real falls.
4. Conclusion In this study, a motion capture experiment was designed to assess activities, motions, and physical design elements which may lead to falls in older adult hospital patients. 27 subjects performed a series of tasks in one of three mock hospital settings (one clinician zone setting and two bathroom settings). A potential fall was identified as a period of time characterized by an increase in the jerk of the COM. In the patient bathroom, turning, pushing, pulling, and grabbing emerged as the only significant factors contributing to potential falls; in the clinician zone, only pushing and pulling emerged as significant factors. Future work involves identifying and changing precise environmental factors in the patient bathroom and clinician zone which may reduce the effects of these motions and performing motion capture experiments to test these new layouts. Another avenue of future work may be to isolate the movements most commonly identified as triggers for potential falls and to perform additional stability-related assessments to determine whether these identified movements remain significant.
Conflict of interest statement The findings and conclusions in this report are those of the authors and do not necessarily represent the view of the National Patient Safety Foundation. The authors also declare that they have no conflict of interest.
Acknowledgments This research work was funded by National Patient Safety Foundation Research Grants Program (2013).
References Bahon, C.A., Sole, G.V., 2009. A jerk threshold-based involuntary lateral movement algorithm. In: Proceedings of the IEEE Conference on Emerging Technologies & Factory Automation, Mallorca, Spain. Centers for Disease Control and Prevention. 2015a. National Estimates of the 10 Leading Causes of Nonfatal Injuries Treated in Hospital Emergency Departments, United States. 〈http://www.cdc.gov/injury/wisqars/leadingcauses.html〉 (accessed 20.10.15). Centers for Disease Control and Prevention. 2015b. 2013, United States Unintentional Fall Deaths and Rates per 1,00,000. cdc.gov. 〈http://webappa.cdc.gov/cgibin/broker.exe〉 (accessed 20.10.15).
Centers for Disease Control and Prevention. 2015c. Costs of Falls Among Older Adults. cdc.gov. 〈http://www.cdc.gov/homeandrecreationalsafety/falls/fallcost. html〉 (accessed 20.10.15). Centers for Disease Control and Prevention. 2015d. Important Facts about Falls. cdc.gov. 〈http://www.cdc.gov/homeandrecreationalsafety/falls/adultfalls. html〉 (accessed 25.01.15). Cloutier, A., Boothby, R., Yang, J., 2011. Motion capture experiments for validating optimization-based human models. In: Proceedings of the HCI International, 3rd International Conference on Digital Human Modeling. Orlando, Florida. CMS. 2008. Never Events. From 〈http://downloads.cms.gov/cmsgov/archiveddownloads/SMDL/downloads/SMD073108/pdf〉 (accessed 25.01.15). Currie, L., 2008. Fall and injury prevention. In: Hughes, R.G. (Ed.), Patient Safety and Quality: An Evidence-Based Handbook for Nurses. Agency for Healthcare Research and Quality, Rockville, MD. Davis, R., Ounpuu, S., Tyburski, D., Gage, J., 1991. A gait analysis data collection and reduction technique. Hum. Mov. Sci. 10, 575–587. http://dx.doi.org/10.1016/ 0167 9457(91)90046. Debard, G., Karsmakers, P., Deschodt, M., Vlaeyen, E., Dejaeger, E., Milisen, K., Tuytelaars, T., 2012. Camera-based fall detection on real world data. In: Dellaert, F., Frahm, J.-M., Pollefeys, M., leal-Taixe, L., Rosenhahn, B. (Eds.), Outdoor and Large-Scale Real-World Scene Analysis, 10. Springer, Berlin, Heidelberg, pp. 356–375. Devore, J.L., 2012. Probability and Statistics for Engineering and the Sciences. Brooks/Cole CENGAGE Learning, Boston, MA. Ferrari, M., Harrison, B., Rawashdeh, O., Hammond, R., Avery, Y., Rawashdeh, M., Maddens, M., 2012. Clinical feasibility trial of a motion detection system for fall prevention in hospitalized older adults. Geriatr. Nurs. 33 (3), 177–183. http://dx. doi.org/10.1016/j.gerinurse.2011.11.011. Haines, T.P., Waldron, N.G., 2011. Translation of falls prevention knowledge into action in hospitals: what should be translated and how should it be done? J. Saf. Res. 42 (6), 431–442 10.1016.j.jsr.2011.10.003. Heung, M., Adamowski, T., Segal, J.H., Malani, P.N., 2010. A successful approach to fall prevention in an outpatient hemodialysis center. Clin. J. Am. Soc. Nephrol. 5 (10), 1775–1779. http://dx.doi.org/10.2215/CJN.01610210. Hignett, S., Masud, T., 2006. A review of environmental hazards associated with inpatient falls. Ergonomics 49 (5 6), 605–616. http://dx.doi.org/10.1080/ 00140130600568949. Hignett, S., Sands, G., Youde, J., Griffiths, P., 2010. Targeting environmental factors to reduce elderly in-patient falls. In: Proceedings of the 1st International Conference on Human Factors and Ergonomics in Healthcare. Miami, Florida. Kato, S., Tsuru, S., Iizuka, Y., 2013. A structural model for patient fall risk and method for determining countermeasures. J. Qual. 20 (5), 503–520. http://dx. doi.org/10.6220/joq. Klenk, J., Becker, C., Lieken, F., Nicolai, S., Maetzler, W., Alt, W., Zijlstra, W., Hausdorff, J.M., van Lummel, R.C., Chiari, L., Lindemann, U., 2011. Comparison of acceleration signals of simulated and real-world backward falls. Med. Eng. Phys. 33 (3), 368–373. http://dx.doi.org/10.1016/j.medengphy.2010.11.003. Lee, K.S., Chaffin, D.B., Parks, C., 1992. A study of slip potential during cart pushing and pulling. IIE Trans. 24 (5), 139–146. http://dx.doi.org/10.1080/07408179208964254. Lee, T.-T., Liu, C.-Y., Kuo, Y.-H., Mills, M.E., Fong, J.-G., Hung, C., 2011. Application of data mining to the identification of critical factors in patient falls using a webbased reporting system. Int. J. Med. Inform. 80 (2), 141–150. http://dx.doi.org/ 10.1016/j.ijmedinf.2010.10.009. Lockhart, T.E., Barth, A.T., Zhang, X., Songra, R., Abdel-Rahman, E., Lach, J., 2010. Portable, Non-invasive Fall Risk Assessment in End Stage Renal Disease Patients on Hemodialysis. Wireless Health 2010, San Diego, California. Ni, B., Dat, N.C., Moulin, P., 2012. RGDB-camera based get-up event detection for hospital fall prevention. In: Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Kyoto, Japan. Pati, D., Valipoor, S., Cloutier, A., Yang, J. (Under review). Physical design factors contributing to patient falls. Journal of Patient Safety. Robinson, C.J., Purucker, M.C., Faulkner, L.W., 1998. Design, control, and characterization of a sliding linear investigative platform for analyzing lower limb stability (SLIPS-FALLS). IEEE Trans. Rehabil. Eng. 6 (3), 334–350. http://dx.doi.org/ 10.1109/IEMBS.1996.646538. Simpson, A.H., Lamb, S., Roberts, P.J., Gardner, T.N., Evans, J.G., 2004. Does the type of flooring affect the risk of hip fracture? Age Aging 33 (3), 242–246. http://dx. doi.org/10.1093/ageing/afh071. Tinetti, M., Speechley, M., 1989. Prevention of falls among the elderly. New Engl. J. Med. 320 (16), 1055–1058. http://dx.doi.org/10.1056/NEJM198904203201606. Van Leeuwen, M., Bennett, L., West, S., Wiles, V., Grasso, J., 2001. Patient falls from bed and the role of bedrails in the acute care setting. Aust. J. Adv. Nurs. 19 (2), 8–13. Wang, C.H., Hsueh, I.P., Sheu, C.F., Yao, G., Hsieh, C.L., 2004. Psychometric properties of 2 simplified 3-level balance scales used for patients with stroke. J. Am. Phys. Ther. Assoc. 84 (5), 430–438. Yamaguchi, T., Yano, M., Onodera, H., Hokkirigawa, K., 2012. Effect of turning angle on falls caused by induced slips during turning. J. Biomech. 45 (2012), 2624–2629. http://dx.doi.org/10.1016/j.jbiomech.2012.08.006. Yang, F., Pai, Y.C., 2011. Automatic recognition of falls in gait-slip training: harness load cell based criteria. J. Biomech. 44 (2011), 2243–2249. http://dx.doi.org/ 10.1016/j.jbiomech.2011.05.039. Yang, F., Espy, D., Bhatt, T., Pai, Y.C., 2012. Two types of slip-induced falls among community dwelling older adults. J. Biomech. 45 (2012), 1259–1264. http://dx. doi.org/10.1016/j.jbiomech.2012.01.036.