Abstract
Background Serial joint range-of-motion (ROM) measurements are an important component of assessments for children with cerebral palsy. Most research has studied ROM stability using group data. Examination of longitudinal intraindividual measures may provide more clinically relevant information about measurement variability.
Objective The aim of this study was to examine the stability of intraindividual longitudinal measurements of hip abduction (ABD), popliteal angle (POP), and ankle dorsiflexion (ADF) ROM measures of children with cerebral palsy.
Design Secondary data analyses were performed.
Methods The stability patterns of individual serial measurements of ABD, POP, and ADF from 85 children (mean age=3.8 years, SD=1.4) collected at baseline (T1), 3 months (T2), 6 months (T3), and 9 months (T4) were examined using T1 as the anchor and bandwidths of ±15 degrees (ABD and POP) and ±10 degrees (ADF) as acceptable variability. Frequencies of stability categories (0°–5°, 5.1°–10°, 10.1°–15°, and >15°) were calculated. Patterns of stability across the 4 time periods also were examined. Group means (T1–T4) were compared using repeated-measures analysis of variance.
Results No significant differences in group means were found except for ABD. Stability patterns revealed that 43.3% to 69.5% of joint measurements were stable with T1 measurements across all 3 subsequent measurements. Stability category frequencies showed that many measurements (ABD=17%, POP=29.9%, and ADF=37.1%) went outside the variability bandwidths even though 39% or more of joint measurements had a change of 5 degrees or less over time.
Limitations Measurement error and true measurement variability cannot be disentangled. The results cannot be extrapolated to other joint ROMs.
Conclusions Individual ROM serial measurement exhibits more variability than group data. Range-of-motion data must be interpreted with caution clinically and efforts made to ensure standardization of data collection methods.
Loss of joint range of motion (ROM) is a major concern in the long-term management of children with cerebral palsy (CP).1,2 It is a prevalent secondary complication of the abnormal muscle tone and prolonged static positioning often associated with CP.3 Thus, monitoring of ROM of both the upper and lower extremities is an important component of assessment and follow-up over the life span. In the lower extremities, hip abduction (ABD), popliteal angle (POP), and ankle dorsiflexion (ADF) measurements are commonly assessed because these movements are often limited in children with CP and decreased ROM in these movements can directly affect their functional abilities. Many children with CP have surgical intervention to maintain joint mobility and optimize daily functioning.4 Physicians and therapists monitor these 3 joint ROMs longitudinally and use significant loss of ROM as an important factor when considering referral for an orthopedic consultation.
Goniometry is the most common clinical procedure for measuring ROM with children with CP. The precision of goniometric measurements with this population has been studied extensively, but different samples, methods, and analyses make it difficult to identify common conclusions.5 Agreement exists that intrarater, intrasession measurements are most reliable6,7 and that precision of measurement decreases when measurements are done by different raters or at different sessions.6,8–12 These research findings raise important questions about how ROM is monitored clinically because the ROM of a child with CP often is monitored over time by different physical therapists and physicians.
The authors of many ROM studies report results as intraclass correlation coefficients. Although this metric provides important information regarding the relationship between measurements and the magnitude of variability, clinicians are more interested in obtaining information about the measurement error in the metric of degrees because it is more clinically relevant. Evaluations of the reliability of ABD, POP, and ADF measurements over time have shown measurement errors ranging from 5,13 7,11,14 and 106 degrees to 15,6 18,12 and 2812 degrees. The measurement error bandwidths for both the studies by Kilgour et al11 and Allington et al13 increase to 18 to 28 degrees when the 95% confidence intervals are considered. For all of these studies, group data were reported. Except for Harris and colleagues'10 single-subject design evaluation of serial shoulder measurements of a child with spastic quadriplegia (reporting intersession measurement variability of ±10°–15°), evaluation of the stability of individual child ROM measurements over time has not been reported. Intraindividual variability is important because reliability of ROM measures may vary due to factors within the child such as severity of involvement and emotional state. Group data, especially when assessed with correlational analyses, may mask the absolute value of intraindividual variability of ROM measurements.
The purpose of this study was to examine the stability of intraindividual longitudinal measurements of ABD, POP, and ADF ROM in children with CP.
Method
Participants
Data from a subsample of children who participated in the Focus on Function (FonF) study, a randomized controlled cluster clinical trial of 2 rehabilitation intervention strategies for children with CP,15 were used for secondary data analyses. In the FonF study, 7 trained assessors measured the children's ABD, POP, and ADF joint ROMs (in degrees) at baseline (T1), 3 months after intervention started (mid-intervention) (T2), 6 months (end of intervention) (T3), and 9 months (3 months postintervention) (T4) using a standardized protocol (Appendix). Each assessor saw the same children in their homes at the 4 assessment times. Each assessor received a half-day of training in the study ROM protocol, and updates in procedure were sent to the assessors, but ongoing interrater or intrarater reliability checks were not done. In the FonF study, treating therapists continued passive stretching programs (as deemed appropriate by the therapist) in one intervention group, whereas passive stretching by therapists was not allowed in the other group.15 Children in both groups received therapy, on average, weekly—a higher frequency than typically provided in rehabilitation centers in Canada.
The results from the FonF study15 revealed no difference in mean ROM changes between the 2 intervention groups over the 4 assessment periods, allowing merging of the data from the 2 groups for these analyses. The sample of the original study was 126 children. These secondary analyses included 85 children (54 male, 31 female) with complete ROM information for the 3 joint measures across all 4 assessments on at least one side; 6 children had complete data for only one side, for a total of 164 measurements for each of the 3 joints measured. The mean age of the sample was 3.8 years (SD=1.4, range=1.0–6.1). The children's functional sitting and mobility status was classified using the Gross Motor Function Classification System (GMFCS)16; in this classification system, level V indicates the most restricted mobility skills. The children in this study represented a range of GMFCS abilities: level I (n=22), level II (n=12), level III (n= 14), level IV (n=17), and level V (n=20). The mean age and GMFCS level and sex distributions of the children excluded from this analysis did not differ significantly from those who were included.
Procedure
We identified “variability bands” representing the amount of measurement variability defined as “stable” or within the expected measurement error for ABD, POP, and ADF joint ROMs. The width of these variability bands was determined by evaluation of variability of serial ROM measurements (in degrees) reported in the literature6,11–14 and consensus among 3 of the authors (J.D., L.W., and J.W.G.), with 20, 8, and 14 years of pediatric clinical experience, respectively, about the magnitude of ROM decrease for each joint measurement that might raise concern clinically. We also considered the criteria developed for the FonF study regarding ROM monitoring. For these secondary analyses, the bandwidth of stability measurements for ABD and POP was set at ±15 degrees, and the stability band for ankle dorsiflexion was identified as ±10 degrees. These bandwidths represent a large margin of degree change before a measurement would be classified as “unstable” and most likely represent a larger range of measurement variability than most clinical criteria used by therapists and physicians to determine true change in ROM.
The baseline measurement (T1) was the “anchor point” to determine longitudinal stability of the 4 measures, and the variability bands for each child were calculated using the value of this first measurement. The absolute difference in measurement of the other 3 measures compared with this anchor measure was used to determine whether the measurements were stable (within the variability band) or unstable (outside the variability band) between a series of 2 measures (T1 to T2, T1 to T3, and T1 to T4). Graphs of each child's measurement history for each joint provided a visual picture of the measurement stability pattern for each child. The Figure provides examples of different patterns of stability for different joints from 3 different children.
Examples of stability patterns. Data from 3 different children. ROM=range of motion, T1=time 1, T3=time 3.
Data Analysis
Repeated-measures analysis of variance (ANOVA) was used to compare group mean ROM values for each joint over the 4 assessment times. The Shapiro-Wilks test for normality was conducted, and the data for some of the joints were not normally distributed. However, the ANOVA is particularly robust to violations of assumption of normality. The Mauchly test of sphericity was conducted on the data, and the Huynh-Feldt correction for degrees of freedom was used when the assumption of sphericity was violated.
To examine intraindividual longitudinal stability, absolute change in values from T1 to T2, T1 to T3, and T1 to T4 were calculated for each child. Using the variability bandwidths and T1 as the anchor, the stability pattern of each child was examined. Frequencies of absolute change groupings (0°–5°, >5°–10°, >10°–15°, >15°–20°, and >20°) were calculated, and the frequencies of each stability pattern for each joint were calculated.
All analyses were completed first on the total sample (164 joints) and then divided into 2 categories based on GMFCS levels (levels I–III and IV–V) to examine the influence of functional mobility on ROM longitudinal stability patterns. These 2 functional mobility categories also were used in the FonF study.
Role of the Funding Source
The National Institutes of Health (grant R01HD044444) and the Alberta Centre for Child, Family and Community Research funded the FonF study.
Results
The group analyses suggested very little variation over time of the means and standard deviations for each of the 3 joint ROMs examined (Tab. 1). Only the repeated-measures ANOVA for ABD demonstrated a statistically significant difference in mean ROM over the 4 assessment times; this significant difference was present for the total group (n=164 joints) and the GMFCS level I–III subgroup (n=92 joints). The mean ABD values increased at T4 (3 months postintervention) rather than decreased. Overall, the results of the group analyses suggest stability of measurement over time.
Comparison of Group Range-of-Motion Values (in Degrees) for Each Joint Over the 4 Assessment Timesa
Table 2 shows the frequencies of different measurement change categories; for each joint, at least 39% of joint measurements had a change of 5 degrees or less over time, but many paired measurements (eg, ABD=17%, POP=29.9%, and ADF=37.1%) went outside the designated variability bandwidths for each joint. The frequency tables of patterns of stability (Tab. 3) indicate that 43.3% to 69.5% of individual joint measurements for the total sample were stable with the T1 measure across all subsequent assessment times. The remainder of joint measurement patterns had at least one measurement that exceeded the limit of the variability bandwidths. The ADF measurements for children in the GMFCS level IV–V subgroup demonstrated the least amount of stability; 64% had at least one measurement outside of the variability bandwidth. The “no stability with T1” stability category in Table 3 could have been due to the fact that the T1 measurement was the outlier and the other 3 measurements were stable with each other. Further investigation revealed that this was indeed the case for 2/4 (50%) ABD measurements, 9/14 (64%) POP measurements, and 16/19 (84%) ADF measurements in the “no stability with T1” category. For all categories, there was no systematic decrease in ROM for any of the 3 joints; often the instability was caused by an increase rather than a decrease in ROM.
Frequencies of Absolute Difference Scores Between Baseline (T1) and T2, T3, and T4 (N=164 joints)a
Frequencies of Stability Patterns of Joint Measurements for Total Group and by Gross Motor Function Classification System (GMFCS) Categoriesa
Discussion
The results suggest that intraindividual variability of serial measurements of ABD, POP, and ADF is masked when participants' group data are used for analysis. Although the large standard deviations of the group data suggest variability, the group means for each joint ROM over time were very consistent across all assessments. The repeated-measures ANOVAs revealed a statistically significant change (increase) only for ABD measurements. The results of the longitudinal intraindividual analyses are more challenging to interpret. It is reassuring that, in most instances, 40% or more of measurements for all 3 joints were within a 5-degree margin of the T1 measurement at the subsequent 3 assessments (Tab. 2). In the same manner, T2, T3, and T4 measurements for the total sample were within the variability bands for more than 50% of the joint measurements of ABD and POP and 43.3% of the measurements of ADF.
Is a rate of stability of approximately 50% good or bad news? From a clinical perspective, the results suggest that clinicians should be cautious when measuring and interpreting serial ROM measurements. Two characteristics of the study warrant this caution. First, the wide variability bands used to define stability were probably larger in magnitude than those that therapists use clinically as indicators of change in ROM measurements. Therapists often become concerned when ADF measurements decrease by less than 10 degrees and ABD and POP measurements decrease by less than 15 degrees. The significance of the magnitude of an absolute change in a joint measurement cannot be determined without also considering the initial ROM value. For example, if a child's hip abduction ROM is initially 45 degrees and decreases to 30 degrees, it is less concerning clinically than if a child has initial hip abduction ROM of 20 degrees and it decreases to 10 degrees. The medical (eg, hip subluxation), functional (eg, mobility restrictions), and caregiving (eg, dressing and perinatal care) implications associated with the second example result in more clinical concern than the ROM changes in the first example even though the absolute magnitude of ROM change is less in the second situation (10°) versus the first example (15°). Both the initial ROM value and the amount of decrease in ROM over time need to be considered when determining when to be concerned about a decrease in ROM values. Cutoff or critical ROM values of concern need to be established. Currently, these types of ROM cutoffs in children and adolescents are not standardized. Decrease in joint ROM is used to identify the need for referral to an orthopedic surgeon in internationally adopted hip surveillance guidelines for children with CP.17 Discussion regarding cutoff values of concern would be useful clinically. In the FonF study,15 critical cutoff values of ROM for each of the 3 joints were identified in consultation with a pediatric physiatrist, and children received individual follow-up if their measurements fell below the identified critical level regardless of the absolute change in their measurements. Clinical decisions cannot be made by considering only absolute change in magnitude of joint ROMs.
The second factor to consider when interpreting the results is that clinically the variation of longitudinal intraindividual measurements may be larger than those obtained from these analyses. Assessors in the clinical trial attended a 1-day training session, used a standardized measurement protocol, and recorded the average of 2 measurements, and therapists were paired with the same children throughout the duration of the study—all criteria assumed to improve measurement reliability.6,10,13 In physical therapist practice, clinicians often do not follow the same standardized protocol for measuring joint ROMs, even within an institution, and different therapists often measure a child over 3-, 6-, and 9-month assessment intervals.
The term “measurement variability” has been intentionally used when presenting the results rather than “measurement error.” It is not possible to determine whether the observed variability in measurement is measurement error or true change in joint ROM. The fact that measurements did not systematically increase or decrease but rather often fluctuated between a decrease and increase from the T1 value over the next 3 assessments suggests that the nature of joint ROM is variable over time, and a decrease in ROM may not always represent a fixed decrease in joint ROM.
For children with CP, spasticity is often considered to be a major factor contributing to measurement variability.6,8,10,12,13 However, Kilgour et al11 reported similar error of measurement between children with CP and a control group and discounted the influence of spasticity. Evaluation of joint measurement variability in 8 healthy adults also revealed intersession variability (±10° using the same rater),18 similar to the measurement variability observed with children with CP, suggesting that spasticity may not be the sole culprit contributing to measurement variability. The influence of spasticity in measurement error warrants further evaluation, but the lack of valid and reliable clinical measures of muscle tone in children and adults with CP19,20 makes this research challenging. Clinically, the results suggest that if a decrease in ROM is suspected, it should be measured more than once by the same physical therapist using a standardized protocol.
Environmental factors also need to be considered when considering measurement variability. In the original study,15 therapists assessed the children in their homes. This location could be considered ideal because it is assumed that children would be more comfortable and relaxed in their own homes and that there would be consistency over time in the home environment. Interestingly, some assessors commented in their study notes that the home environment had a negative effect on assessment procedures; they mentioned the activity of siblings and television background noise as distractions for the children. They also noted that the lack of a plinth or high mat made it more difficult to ensure optimal and consistent starting and end positions for measurement. On the other hand, children are often more anxious in a clinical setting, and this factor also may have influenced the stability of measurements. Correct positioning7 and correct end-range joint positioning11 have been identified as important for measuring POP joint ROM and may apply to other joint ROMs.
The relationship established between the therapist and the child is another important environmental consideration. One of the main sources of measurement variability identified by McDowell et al12 was child-assessor variability; children had greater ROM values with different therapists, and the differences were not systematic. They attributed the difference to lack of specific training of the therapists, but the relationship established between the child and therapist also may have influenced this finding.
Limitations
The use of T1 as the anchor measurement increased the frequency of instability patterns if the other 3 measurements were outside of the designated variability band with T1 but within the variability band with each other. We chose not to consider the data as poor but rather viewed it as a stability pattern that was clinically relevant. When designing our analyses approach, we considered moving the error band at each assessment time and comparing adjacent measurements rather than using T1 as the anchor measurement. However, doing so moved the variability values each time and created a very large window of variability over the total time period. We chose to use T1 as the main comparator because, clinically, decreased values of ROM over a period of 6 to 12 months are usually considered worrisome. Another limitation of the study was that because of small and unequal sample sizes between the 2 GMFCS classification categories, only trends can be suggested regarding the effect of mobility restrictions on the stability of ROM measures over time. Ongoing reliability evaluations, both interrater and intrarater, were not conducted. It was challenging to have interrater reliability evaluations with therapists situated across the country. Intrarater measurement variability also was challenging because it would have meant another visit to the child's home. The training and the standardization of measurement protocol used in the FonF study were rigorous to minimize measurement error. Individual child ROM measurements that were clinically concerning were followed up with each child's therapist. The results presented cannot be interpolated to measures of other joint ROMs.
Clinical Implications/Future Directions
In addition to consideration of measurement error, identification of “critical values” of ROM for each joint is needed to assist in the interpretation of changing ROM values and the intervention needed. Discussion and consensus among clinicians regarding the identification of critical values could lead to the development of clinical guidelines regarding the interpretation of serial ROM measurements with children with CP. Therapists would have a significant role to play in the identification of critical values, especially in identifying the ROM requirements of essential functional activities such as a child's ability to move to a sitting or standing position independently or to toilet independently. Range-of-motion requirements also could be identified for activities that might affect a child's participation in the community, such as riding a bicycle. Instead of considering “typical” or “full” ROM values, therapists could start to consider “functional” ROM values for specific activities.
Based on our study, serial ROM data in children with CP need to be interpreted with caution. The evaluation of longitudinal intraindividual data revealed a larger magnitude of variability than group data results even when measurement was standardized and the same rater assessed the children over time. These results make it challenging for clinicians to interpret the significance of changes in ROM with children with CP. Other researchers8,10 have discussed this dilemma and cautioned against making clinical decisions based primarily on ROM changes. Previous literature has stressed the importance of standardized measurement protocols, using the same assessor over time, blinding the assessor to previous measurements, and paying attention to environmental influences, but these practical suggestions do not appear to be routinely integrated into clinical measurement protocols. Clinically, it often is not possible for the same therapist to measure ROM over time, but the literature is clear that this limitation decreases measurement variability considerably. In large rehabilitation centers, the introduction of a “measurement therapist” role could be a viable solution. This therapist would be responsible for the measurement of most children's ROM over time, using a standardized protocol. Even then, our results suggest that variability in goniometric measurements will persist. Given the variability of intraindividual measurements, ROM data cannot be the sole indicator of success of procedures such as botulinum toxin injections, serial casting, and surgical interventions, nor can a decrease in ROM be the only reason to justify such procedures. The functional abilities of a child and families' goals and concerns need to be considered when making such clinical decisions.
Appendix.
Focus on Function Study Range-of-Motion Guidelines for Assessorsa
a ASIS=anterior superior iliac spine, PSIS=posterior superior iliac spine. The appendix may not be used or reproduced without written permission from the authors.
b The manual is available on the CanChild website: http://www.canchild.ca/en/measures/saromm.asp.
Footnotes
Dr Darrah, Dr Gorter, and Dr Law provided concept/idea/research design. Dr Darrah, Dr Wiart, and Dr Gorter provided writing. Dr Darrah, Dr Wiart, and Dr Law provided data collection. Dr Darrah and Dr Wiart provided data analysis. Dr Law provided project management and consultation (including review of manuscript before submission). Dr Darrah and Dr Law provided fund procurement. Dr Darrah provided institutional liaisons and clerical support.
The National Institutes of Health (grant R01HD044444) and the Alberta Centre for Child, Family and Community Research funded the Focus on Function study. The authors acknowledge the coinvestigators in that study: Nancy Pollock, Brenda Wilson, Dianne J. Russell, Stephen D. Walter, Peter Rosenbaum, and Barb Galuppi.
- Received August 16, 2013.
- Accepted February 16, 2014.
- © 2014 American Physical Therapy Association
References
Issue highlights
