Secondary Logo

Journal Logo

Original Research Articles

Method to Quantify Cadence Variability of Individuals with Lower-Limb Amputation

Arch, Elisa S. PhD; Erol, Ozan MSc; Bortz, Connor BS; Madden, Chelsea BS; Galbraith, Matthew BS; Rossi, Anthony BS; Lewis, Jessica BS; Higginson, Jill S. PhD; Buckley, Jenni M. PhD; Horne, John CPO, CPED

Author Information
Journal of Prosthetics and Orthotics: April 2017 - Volume 29 - Issue 2 - p 73-79
doi: 10.1097/JPO.0000000000000124
  • Free


To enhance mobility and thus activity and societal participation,1 prostheses of patients with lower-limb amputations should be prescribed to meet each individuals' needs.2 Improper prosthetic prescriptions may result in suboptimal outcomes for patients, such as a loss of functional mobility.3–5 Currently, prostheses are prescribed by classifying individuals with lower-limb amputation by their Medicare Functional Classification Level, or K-level. The K-level system was developed in the United States to aid in the reimbursement of prescribed prostheses from a third-party payer. According to the Centers for Medicare & Medicaid Clinical Template in Support of Claims for Lower Limb Prostheses, K-level identifies an individual's potential functional ability and is based, in part, on their ability to walk with variable cadence.6 Cadence is defined as the step count per unit time, typically 1 minute,7 and cadence variability is defined as the ability to walk at different step rates.

In individuals with lower-limb amputation, cadence has been shown to significantly influence gait performance8 and be an indicator of walking intensity.9 Furthermore, the ability to walk at variable cadence may influence an individual's ability to navigate different environmental barriers and stay safe. As per the K-level descriptors, the ability to walk at a range of different cadences is an indicator of a higher functional level. Thus, it is important to accurately determine an individual's ability to walk with a variable cadence to ensure each individual is prescribed a prosthesis that is best matched to his or her functional ability. Many other parameters that are considered important in determining functional ability, such as daily step count1,10–13 or performance on clinical measures,1,14,15 have been evaluated in the literature. However, the cadence variability of individuals with lower-limb amputation has never, to our knowledge, been investigated. As a result, there are no reference values or standardized thresholds of cadence variability for each K-level, leaving it up to the clinician to determine what defines cadence variability. Many clinicians determine an individual's ability to vary cadence via visual inspection. This subjective process contributes to the variability of the K-level classification and prosthesis prescription process.14,16 This variability may lead to an individual receiving a suboptimal prosthesis, which can impede an individual's ability to achieve his or her highest possible level of function. The absence of data on cadence variability may be due, in large part, to the lack of a method to measure cadence variability as individuals go about their daily activities in the real world. To garner an individual's ability to walk with a variable cadence, cadence variability should be measured in the real world, as opposed to in a controlled laboratory setting, because cadence measures in these two settings have been shown to be substantially different.7

Activity monitors can quantitatively measure real-world step counts as individuals go about their daily lives.1,17,18 Accelerometer-based monitors, which often provide minute-by-minute step counts, have been shown to accurately count steps among healthy adults19–22 and individuals poststroke and posttraumatic brain injury,23,24 to name a few. One of the first studies to evaluate an activity monitor's step-count accuracy for individuals with lower-limb amputations was by Coleman and colleagues. Results from this study supported validity of the monitor evaluated (the StepWatch), albeit the study was conducted on a small sample size.25 Additional studies since that time have demonstrated the step-count accuracy of a range of activity monitors for individuals with lower-limb amputations.18,26,27 These previous studies used an array of activity monitors, with variability in weight, size, cost, commercial availability, and ease of data access. One newer monitor, the FitBit One, is a wireless, lightweight, low-cost, commercially available option. These advantages have the potential to increase the use of such activity monitors for both research and clinical applications related to the lower-limb amputation population. While the step-count accuracy of FitBit devices has been documented in other populations,1,9,23 they have been shown to have inaccuracies in some situations, such as for slower or faster walking speeds.20,23 Furthermore, to our knowledge, it has not been previously evaluated for individuals with lower-limb amputations.

Therefore, the purpose of this study was to develop and demonstrate feasibility of a method to quantify real-world cadence variability for individuals with lower-limb amputation using a FitBit One. Of particular interest for this study was developing a method that could differentiate the cadence variability for individuals classified as K2 and K3, as the ability to walk with variable cadence is a key parameter differentiating these two functional levels. Thus, it was hypothesized that the method would be able to differentiate the cadence characteristics of individuals classified as K2 versus those classified as K3.


Method to Quantify Real-World Cadence Variability

The cadence variability method begins by collecting step-count data from an accelerometer-based activity monitor as individuals go about their daily activities. From these data, the cadence at each minute is calculated by taking the number of steps recorded in that minute. A Weibull probability density function (pdf) is then fit to each cadence data set, and the scale parameter (measure of distribution's spread) is calculated to quantify the cadence variability over the duration of the observation period. In addition, to facilitate clarity and visualization, cadence is binned in to categories (bin 1 = 0–10 steps/min; bin 2 = 11–20 steps/min, etc.) to create a histogram depicting the frequency of occurrence of each cadence over the observation period.

The Weibull pdf scale parameter provides a similar measure to the mean in normal distribution.28 Smaller scale parameter values indicate that most of the cadence data are populated at lower cadences resulting in a left skewed distribution. In other words, such a result indicates the subject does not walk at a variety of cadences. On the other hand, higher values of the scale parameter indicate that the subject is able to achieve cadence values that are more widely distributed.

The Weibull pdf was chosen to be used to quantify cadence variability as it can model various data sets while taking on the characteristics of other distribution types.29 Importantly, this behavior allows Weibull to handle skewed data. In contrast, descriptive statistics such as the standard deviation assume a normal distribution and may lead to inaccurate reports of variability with nonnormally distribution, such as skewed data.30 When evaluating cadence over a period, we anticipated most individuals who walk with little to no cadence variability to have a skewed cadence histogram. This belief that the cadence data would not follow a normal distribution was evaluated in the data set used to assess feasibility of the method. Furthermore, the scale parameter captures the shape of a distribution, whereas a descriptive statistic only captures numerical differences among data points without regard to the overall shape of the distribution.

FitBit One Step-Count Accuracy Assessment

While the FitBit One has been found to be accurate in most healthy adults at usual walking speeds,19 step-count accuracy had not previously been reported for individuals with lower-limb amputations. Therefore, the ability of the FitBit One to accurately measure the number of steps taken by individuals with lower-limb amputations walking on level ground while wearing a prosthesis was evaluated using a protocol similar to many other activity monitor step-count accuracy assessments.20,23,31,32 This study was approved by the University of Delaware Institutional Review Board. Nine participants with unilateral lower-limb amputation between the ages of 21 and 85 years who were able to walk for 1 minute were evaluated. Five of the participants were classified as K3, and four participants were classified as K2. Six of the participants had transtibial amputations, while three participants had transfemoral amputations.

When participants arrived at the laboratory, they were equipped with the activity monitor, which was secured near the ankle (connection between the prosthetic foot and pylon) of their prosthesis using the strap included with the FitBit. The number of steps currently on the activity monitor (initial step count) was recorded once the device was secured to the participant's prosthesis, and the participant was standing at the beginning of the pathway. Then, the participant was instructed to walk at his/her comfortable walking speed along a pathway for 1 minute. The pathway included turns of 90° and possibly 180°, depending on how far the participant could walk in 1 minute. During the minute of walking, an investigator counted the number of steps (each heel strike) the participant took. At the end of the minute, the number of steps counted by the investigator as well as the number of steps now on the activity monitor (final step count) was recorded. The number of steps the activity monitor counted during the 1 minute of walking (final step count–initial step count) was determined. Then, the absolute percent difference between the activity monitor and investigator-generated step counts was calculated (absolute value of [(monitor − investigator)/investigator] × 100) to evaluate the step-count accuracy of the FitBit One for this patient population. A paired t-test (α = 0.05) was used to evaluate the statistical significance between the monitor- and investigator-determined step counts across all subjects.

Feasibility Assessment of Cadence Variability Method

Once the cadence variability method was developed and step-count accuracy of the activity monitor established, a feasibility assessment was conducted to determine if the method could differentiate the real-world cadence characteristics of individuals classified as K2 versus K3. Thirty-one individuals with a unilateral transtibial or transfemoral amputation, aged 21 to 85 years, were recruited for this portion of the study. Ultimately, data from 27 participants were analyzed in this study as data from the remaining participants were unable to be used (see Results section for details). Of the participants included in the analysis, 16 individuals (sex: 11 male, 5 female; amputation level: 10 transtibial, 6 transfemoral) were classified as K3 and 11 individuals (sex: 10 male, 1 female; amputation level: 10 transtibial, 1 transfemoral) were classified as K2. Average age, height, and mass demographics for each group are provided in Table 1. K-level classifications were assigned by the participants' health care providers who prescribed the prosthesis. For this study, the participants' most recent K-level that was used to prescribe the prosthesis they were currently wearing was used.

Table 1
Table 1:
Demographic and Cadence Data

At the laboratory, participants were fit with a fully charged Fitbit One activity monitor, secured near the ankle of their prosthesis using the strap included with the FitBit. Then the participants were sent home with the FitBit for a 7-day observation period that began the day after they received the activity monitor. A 7-day observation period was used as it has been shown to be sufficient to establish natural activity patterns for adults25,33 and is a standard length of time for real-world activity monitoring in many published studies.1,11,12,17 Participants were instructed to perform their regular daily activities, wearing the device at all times that they were wearing their prosthesis. Participants were provided with a prepaid and addressed postage box and instructed to mail back the activity monitor at the conclusion of the observation period.

Once the activity monitor was returned, minute-by-minute step-count data were extracted from the device via FitaBase, a cloud-based fee-for-service data aggregation platform.34 FitaBase provides minute-by-minute step-count data for each day, and because the date of each participant's clinical assessment was recorded, step-count data could be extracted for just the 7 days of interest for the observation period. Data were inspected for inconsistencies, such as single days with no steps. If the inconsistency could not be explained and deemed typical activity via discussions with the participants, observation period data were deemed invalid and that participant was excluded from analysis in this study. An example of valid explanations for an inconsistency was a participant stating that he/she regularly did not wear the prosthesis on a particular day of the week to explain a day with no steps recorded.

For each participant with a valid set of data, cadence (number of steps per minute) at each minute was calculated. Minute-by-minute cadence data were evaluated for normality using a Kolmogorov-Smirnov test with a 95% confidence interval to assess the belief that the cadence data sets would not follow a normal distribution, then each subject's cadence variability (Weibull pdf scale parameter) over the 7-day observation period. In addition, mean and maximum cadence values over the 7-day observation period were determined. Finally, mean and standard deviations of these parameters for the K2 and K3 groups were calculated and a Welch's t-test with a confidence interval of 0.95 was used to evaluate differences between the K2 and K3 groups.


Walking speeds for the nine participants who underwent the step-count accuracy assessment ranged from 0.54 to 1.44 m/s. The average number of steps counted by the activity monitor during the 1 minute of walking was 96.33 ± 13 steps, while the average number of steps counted by the investigator was 94 ± 10.2 steps. The percent differences between the activity monitor and investigator-generated step counts ranged from −5.8% to 9.5%, with the activity monitor undercounting the number of steps for one participant, agreeing perfectly with the investigator step count for one participant and overcounting the number of steps for the remaining seven participants. There were no apparent differences in step count based on walking speed. The average absolute percent difference between the two step counts was 3.4% ± 3.0%. There was no significant difference between the activity monitor and investigator-generated step counts (P = 0.637). There were no apparent differences in the step-count accuracy data between the participants classified as K2 and those classified as K3.

Feasibility of the cadence variability method to differentiate between individuals classified as K2 versus K3 was conducted on a total of 27 participants. Two participants were eliminated from analysis due to invalid days during the 7-day observation period. It is unknown if these invalid days were due to errors in the FitBit or if the subjects did not wear the monitor those days. In addition, two participants were excluded from this study because the activity monitor was lost during the observation period. Finally, results from the Kolmogorov-Smirnov test showed that all subject's cadence data sets rejected the null hypothesis, indicating the data sets were not in the form of a standard normal distribution. Thus, analysis continued with the Weibull pdf as opposed to a standard deviation to quantify cadence variability.

Averages of the mean cadence, maximum cadence, and cadence variability data are shown in Table 1. Cadences recorded were as high as 121 steps/min during the 7-day observation period, although most individuals classified as K3 had no higher than 110 steps/min, with individuals classified as K2 typically never reaching cadences that high. On average, the mean and maximum cadences for individuals classified as K2 were significantly lower than those classified as K3 (mean, P = 0.047; maximum, P = 0.036). Individuals classified as K2 also had less variability in their cadence over the 7-day observation period when compared with those classified as K3 (Figure 1) as quantified by the significantly lower average scale parameter for the individuals classified as K2 than K3 (P = 0.037) (Table 1). While most participants' data fell within one standard deviation of the group mean for the participant's assigned classification level, a few participant's cadence data were above/below this one standard deviation threshold (Table 2). Participants C and D were classified as K3, but all of their cadence data (mean, maximum, variability) were near or below the K2 group average. In contrast, participants A and B were classified as K2 but had cadence data that reflected K3 level function.

Figure 1
Figure 1:
Cadence histograms. Representative cadence histogram of individuals classified as a K2 (left) and a K3 (right). Weibull distribution shown (line) and the scale parameter indicated for each histogram. As seen in these representative histograms, individuals classified as K2 had a more narrow distribution in their cadence data, as indicated by the lower scale parameter, while individuals classified as K3 has a wider cadence distribution, as seen by the higher scale parameter.
Table 2
Table 2:
Outlying Cadence Data


This study developed and demonstrated feasibility of a method to evaluate real-world cadence variability. Results showed that the cadence variability method developed could differentiate the cadence characteristics between K2 and K3 functional levels for most participants. To our knowledge, this is the first study to present a method to quantify and document real-world cadence variability in this population.

A key characteristic differentiating the K2 and K3 classification levels is that individuals classified as K3 have “the ability or potential for ambulation with variable cadence.”35 However, before this study, there was no method to quantify the real-world cadence variability of individuals with lower-limb amputation. Therefore, clinicians are currently left to determine whether an individual can walk with a variable cadence via qualitative measures or visual inspection, all within a clinic setting. This subjectivity can result in great interclinician variability in K-level classification and prosthesis prescription as well as K-levels that inaccurately reflect the individual's functional ability in the real world. These limitations of the current classification system often make it difficult for clinicians to justify to third-party payers the prostheses they have prescribed. The historic lack of data and method to quantify cadence variability may have been due, in large part, to the challenges of measuring real-world cadence. With the rise in availability of activity monitors, tracking real-world cadence is now feasible and the cadence variability method developed in this study provides an objective way to investigate real-world cadence characteristics of individuals with lower-limb amputations.

The FitBit One appeared to provide an acceptable tool to quantitatively count the steps taken for nine subjects with lower-limb amputations classified as K2 or K3 during level-ground, forward walking. From the step-count accuracy assessment, the 3.4% difference between the activity monitor and investigator-generated step counts is on par with steps counts of the most accurate activity monitors and well below the ±10% difference deemed acceptable for device-based step counts.20 All participants performed at least one 90° turn and four of the nine participants performed at least one 180° turn during the step-count accuracy assessment. In our accuracy assessment, the activity monitor tended to overestimate the number of steps. Reports vary in the literature, with some accelerometer-based activity monitors typically overestimating the number of steps while others underestimating the step count.20,23,26,27

The demographics of the study participants were likely representative of the K2 and K3 population. There were no significant differences in height and mass between the two K-level groups; however, the K2 participants in this study were, on average, significantly older than the K3 participants, which is likely a good representation of the K2 and K3 populations. Therefore, the K3 individuals in this study may have had vocational activities that required or promoted higher or more variable cadence activity, while the individuals classified as K2 may not have. It is important to note that this study did not assess an individual's ability to walk at variable cadences, because the participants were never instructed to try to walk at different cadences. Instead, this study assessed the typical level of cadence variability of individuals classified as K2 and K3. However, an individual's vocational needs are supposed to be taken in to account when prescribing a prosthesis.35

Results from this study showed that the cadence variability method could distinguish between individuals classified as K2 versus K3. As expected based on the K-level descriptors, individuals classified as K2 walked, on average, with significantly lower mean cadence, lower maximum cadence, and less cadence variability than those classified as K3. However, cadence characteristic quantified by the method presented in this study did not align with individual's assigned K-levels for four participants, as these participants' cadence variability was greater than ±1 standard deviation of the classification level's mean. The discrepancies were in both directions, with two individuals classified as K2 having cadence data in line with K3 level averages and two individuals classified as K3 having cadence data in line with K2 level averages. However, these discrepancies do not invalidate the cadence variability method. Because the current K-level classification system is heavily subjective, these individuals may have been misclassified by their health care provider. In addition, because cadence variability is not the only factor that is used to define K-level, these participants may have had other factors that justified the K-level they were assigned to. Furthermore, the time since the K-level was assigned and current prosthesis prescribed were not considered for this study, so it is possible that these subjects' functional levels had changed but they had not been reclassified and prescribed new prostheses.

While the human subjects data used in this study served to assess feasibility of the method to differentiate cadence variability between individuals classified as K2 and K3, future work is needed before there is wide-scale implementation of this method into clinical practice. These future studies should aim to determine a threshold for the scale parameter, above which defines a variable cadence, as well as normative values for the K2 and K3 classification levels.

It is recognized that the K-levels are assigned before a prosthesis is prescribed; however, cadence and cadence variability cannot truly be assessed unless an individual is walking while wearing a prosthesis. Thus, the method presented in this study could be used for individuals with their initial prostheses, and the cadence variability data could be used to drive and justify prescription of their definitive prosthesis. Furthermore, this study can be built upon in the future by developing clinical measures that are predictive of an individual's real-world cadence variability to facilitate a standardized, a priori K-level classification. Minimizing the subjectivity in the current process will improve prosthesis prescription, which has the potential to enhance an individual's mobility and function. Such advancements will decrease health care costs by reducing comorbidities associated with limited activity and participation as well as the rate at which individuals need new prostheses due to improper fit.

Some limitations of this study should be noted. In this study, investigator-generated step counts during the FitBit One step-count accuracy assessment were done in real time. Videotaping the experiment would have enabled the investigators to recount steps to ensure accuracy of the investigator-generated count. Furthermore, while the accuracy of the FitBit was evaluated using methods similar to those in the literature, step-count accuracy was only evaluated in a controlled setting during level-ground, forward walking in nine subjects. However, this procedure assessed accuracy of the FitBit in counting steps for different gait patterns for individuals with lower-limb amputation, which is arguably the most important component to assess. Further, level-ground, forward walking is the primary walking task when measuring cadence and the ability to vary cadence. However, accuracy should be tested during complex walking tasks in a free-living, or simulated free-living, environment27,36 across a range of walking speeds for each subject. In addition, while the walking speeds of subjects included in the step-count accuracy assessment were quite varied, step-count accuracy was only assessed for each subject's self-selected walking speed. People likely walk at a range of speeds in the free-living environment. In summary, accuracy testing was considered sufficient for the development and demonstration of a new method for quantifying real-world cadence variability for individuals with lower-limb amputation. It was not the purpose of this study to comprehensibly validate the FitBit One activity monitor. Before this method is implemented into clinical practice, the medical device status and step-count accuracy of the activity monitor being used should be comprehensively evaluated. Devices that are listed by the US Food and Drug Administration (FDA) offer additional security against false data because the FDA audits its quality control systems. The StepWatch (Modus Health LLC, Washington, DC, USA) and ActiGraph (ActiGraph, Pensacola, FL, USA) are examples of FDA Class II medical devices.

Moreover, the inability of some activity monitors including the FitBit to distinguish subconscious leg movements while sitting from a step has been noted.20 However, it is not believed that individuals with lower-limb amputations frequently subconsciously wiggle their leg while sitting due to the difficulty of such a movement with a prosthesis. Because the activity monitor was placed on the prosthetic leg in this study, it is not believed that the data was substantially influenced by this device limitation. In addition, while the Weibull pdf is ideal for modeling skewed data, as often arises in a cadence data set, the Weibull pdf cannot accurately model bimodal distributions. However, none of the cadence data sets in this study exhibited bimodal distributions. It is hard to imagine a situation where a cadence data set would exhibit a strong bimodal distribution as this would mean that the participant only walked at slow and fast cadences, not middle-level rates, over the 7-day observation period. Finally, only individuals with unilateral lower-limb amputation were included in this study. Outcomes of this study may differ for individuals with bilateral lower-limb amputations.


This study presented a novel method to quantify the real-world cadence variability of individuals with lower-limb amputations, a capability that has never been presented before. Results from this study demonstrated feasibility of using this method to differentiate the cadence variability between individuals classified as K2 and those classified as K3.


1. Parker K, Kirby RL, Adderson J, Thompson K. Ambulation of people with lower-limb amputations: relationship between capacity and performance measures. Arch Phys Med Rehabil 2010;91(4):543–549.
2. Van der Linde H, Hofstad CJ, Geurts AC, et al. A systematic literature review of the effect of different prosthetic components on human functioning with a lower-limb prosthesis. J Rehabil Res Dev 2004;41(4):555–570.
3. Hsu MJ, Nielsen DH, Lin-Chan SJ, Shurr D. The effects of prosthetic foot design on physiologic measurements, self-selected walking velocity, and physical activity in people with transtibial amputation. Arch Phys Med Rehabil 2006;87(1):123–129.
4. Klute GK, Kallfelz CF, Czerniecki JM. Mechanical properties of prosthetic limbs: adapting to the patient. J Rehabil Res Dev 2001;38(3):299–307.
5. Kahle JT, Highsmith MJ, Hubbard SL. Comparison of nonmicroprocessor knee mechanism versus C-Leg on Prosthesis Evaluation Questionnaire, stumbles, falls, walking tests, stair descent, and knee preference. J Rehabil Res Dev 2008;45(1):1–14.
6. HCFA Common Procedure Coding System HCPCS 2001. Washington, DC: Centers for Medicare and Medicaid Services; 2001.
7. Tudor-locke C, Barreira TV, Brouillette RM, et al. Preliminary comparison of clinical and free-living measures of stepping cadence in older adults. J Phys Act Health 2013;10:1175–1180.
8. Cortés A, Viosca E, Hoyos JV, et al. Optimisation of the prescription for trans-tibial (TT) amputees. Prosthet Orthot Int 1997;21(3):168–174.
9. Rowe DA, McMinn D, Peacock L, et al. Cadence, energy expenditure, and gait symmetry during music-prompted and self-regulated walking in adults with unilateral transtibial amputation. J Phys Act Health 2014;11(2):320–329.
10. Lin SJ, Winston KD, Mitchell J, et al. Physical activity, functional capacity, and step variability during walking in people with lower-limb amputation. Gait Posture 2014;40(1):140–144.
11. Stepien JM, Cavenett S, Taylor L, Crotty M. Activity levels among lower-limb amputees: self-report versus step activity monitor. Arch Phys Med Rehabil 2007;88(7):896–900.
12. Hordacre B, Barr C, Crotty M. Use of an activity monitor and GPS device to assess community activity and participation in transtibial amputees. Sensors (Basel) 2014;14(4):5845–5859.
13. Halsne EG, Waddingham MG, Hafner BJ. Long-term activity in and among persons with transfemoral amputation. J Rehabil Res Dev 2013;50(4):515–530. Available at: Accessed June 18, 2014.
14. Gailey RS, Roach KE, Applegate EB, et al. The amputee mobility predictor: an instrument to assess determinants of the lower-limb amputee's ability to ambulate. Arch Phys Med Rehabil 2002;83(5):613–627.
15. Lin SJ, Bose NH. Six-minute walk test in persons with transtibial amputation. Arch Phys Med Rehabil 2008;89(12):2354–2359.
16. Kaluf B. Evaluation of mobility in persons with limb loss using the amputee mobility predictor and the prosthesis evaluation questionnaire - mobility subscale: a six-month retrospective chart review. J Prosthetics Orthot 2014;26(2):70–76.
17. Albert MV, McCarthy C, Valentin J, et al. Monitoring functional capability of individuals with lower limb amputations using mobile phones. PLoS One 2013;8(6):e65340.
18. Ramstrand N, Nilsson KA. Validation of a patient activity monitor to quantify ambulatory activity in an amputee population. Prosthet Orthot Int 2007;31(2):157–166.
19. Takacs J, Pollock CL, Guenther JR, et al. Validation of the Fitbit One activity monitor device during treadmill walking. J Sci Med Sport 2014;17(5):496–500.
20. Fortune E, Lugade V, Morrow M, Kaufman K. Validity of using tri-axial accelerometers to measure human movement - Part II: Step counts at a wide range of gait velocities. Med Eng Phys 2014;36(6):659–669.
21. Karabulut M, Crouter SE, Bassett DR Jr. Comparison of two waist-mounted and two ankle-mounted electronic pedometers. Eur J Appl Physiol 2005;95(4):335–343.
22. Feito Y, Bassett DR, Thompson DL. Evaluation of activity monitors in controlled and free-living environments. Med Sci Sports Exerc 2012;44(4):733–741.
23. Fulk GD, Combs SA, Danks KA, et al. Accuracy of 2 activity monitors in detecting steps in people with stroke and traumatic brain injury. Phys Ther 2014;94(2):222–229.
24. Macko RF, Haeuber E, Shaughnessy M, et al. Microprocessor-based ambulatory activity monitoring in stroke patients. Med Sci Sports Exerc 2002;34(3):394–399.
25. Coleman KL, Smith DG, Boone DA, et al. Step activity monitor: long-term, continuous recording of ambulatory function. J Rehabil Res Dev 1999;36(1):8–18.
26. Bussmann JB, Culhane KM, Horemans HL, et al. Validity of the prosthetic activity monitor to assess the duration and spatio-temporal characteristics of prosthetic walking. IEEE Trans Neural Syst Rehabil Eng 2004;12(4):379–386.
27. Dudek NL, Khan OD, Lemaire ED, et al. Ambulation monitoring of transtibial amputation subjects with patient activity monitor versus pedometer. J Rehabil Res Dev 2008;45(4):577–585.
28. ReliaSoft. Life Data Analysis Reference. Available at: Accessed May 12, 2015.
29. Murthy DNP, Xie M, Jiang R. Weibull Models. New York: Wiley; 2004.
30. Chau T, Young S, Redekop S. Managing variability in the summary and comparison of gait data. J Neuroeng Rehabil 2005;2(22):2–22.
31. Schneider PL, Crouter SE, Lukajic O, Bassett DR Jr. Accuracy and reliability of 10 pedometers for measuring steps over a 400-m walk. Med Sci Sports Exerc 2003;35(10):1779–1784.
32. Chou TGR, Webster JB, Shahrebani M, et al. Characterizaion of step count accuracy of actigraph activity monitor in persons with lower limb amputation. J Prosthet Orthot 2009;21(4):208–214.
33. Clemes SA, Griffiths PL. How many days of pedometer monitoring predict monthly ambulatory activity in adults? Med Sci Sports Exerc 2008;40(9):1589–1595.
34. Small Steps Lab LLC. Fitabase. Available at: Accessed April 9, 2015.
35. Dear Physician. Washington, DC: Centers for Medicare & Medicaid Services; 2011.
36. Hickey A, John D, Sasaki JE, et al. Validity of activity monitor step detection is related to movement patterns. J Phys Act Health 2016;13(20):145–153.

activity monitor; cadence variability; functional level; lower-limb amputation; real-world walking

Copyright © 2017 American Academy of Orthotists and Prosthetists