Abstract
Background Balance confidence and fear of falling are factors associated with recurrent falls in people with Parkinson disease (PD). However, the accuracy for predicting falls on the basis of self-report measures has not been widely investigated.
Objective The study objectives were: (1) to compare the accuracy of the Activities-specific Balance Confidence Scale (ABC) and the Falls Efficacy Scale–International (FES-I) with that of the Berg Balance Scale (BBS), Dynamic Gait Index (DGI), Functional Reach Test (FRT), and Timed “Up & Go” Test (TUG) for predicting recurrent falls in people with PD and (2) to explore the ability of combinations of up to 3 tests to predict recurrent falls.
Design This was a prospective cohort study involving 225 people with PD.
Methods Participants were assessed with the ABC, FES-I, BBS, FRT, TUG, and DGI. Participants who reported 2 or more falls in the 12-month follow-up period were classified as recurrent fallers. Areas under the receiver operating characteristic curves were determined, and the Akaike information criterion was used to select the best predictive model.
Results Eighty-four participants (37.3%) were classified as recurrent fallers. Areas under the receiver operating characteristic curves for the ABC, FES-I, TUG, FRT, DGI, and BBS were 0.73, 0.74, 0.72, 0.74, 0.76, and 0.79, respectively. Two-test models provided additional discriminating ability compared with individual measures and had Akaike information criterion values similar to those of 3-test models, particularly the combination of the BBS with the FES-I.
Limitations The lack of an external validation sample was a limitation of this study.
Conclusions The ABC and FES-I demonstrated moderate accuracy in predicting recurrent falls and a predictive ability similar to that of performance-based balance measures, especially the FRT and the TUG. Two-test models showed performance similar to that of 3-test models, suggesting that a combination of 2 measures may improve the ability to predict recurrent falls in people with PD. Specifically, the combination of the BBS with the FES-I may be considered.
Parkinson disease (PD) is one of the most common neurodegenerative diseases affecting elderly people and is a major cause of disability in this population.1 One or more aspects of postural control, such as proprioceptive input,2 muscular strength,3 range of motion, latency and amplitude of postural responses, and limits of stability, are usually affected.4,5 These impairments result in reduced balance and an increased risk of falling.5–7
Falls are common in people with PD, and recurrent falls, especially, can be considered a disabling feature of the disorder. Previous studies showed a high incidence of recurrent falls in people with PD, ranging from 18% to 65% in a 1-year period.8–11 A recent systematic review of falling in people with PD showed a high rate of falls, ranging from 4.7 to 67.6 falls per recurrent faller (person experiencing 2 or more falls) per year (average=20.8).11
Falls can lead to serious injuries and functional limitations, including hip fractures, reduced mobility, and difficulties in performing daily tasks.12 Because some injuries resulting from falling may require hospitalization, falls are associated with high costs for people and the health system.13 In addition, psychosocial consequences, such as fear of falling and social isolation, can have a marked impact on quality of life.14,15 Previous studies demonstrated that measures of fear of falling accounted for 65% of the variance in quality of life in people with PD.16 Fear of falling has been found to be greater in people with PD than in people who are healthy14,17 and has been correlated with PD severity, disability,15 and clinical measures of balance and gait—showing its relationship with postural control in people with PD.14,15
Reduced balance confidence and fear of falling have been shown to be factors associated with recurrent falls in people with PD.18,19 The Activities-specific Balance Confidence Scale (ABC)20 and Falls Efficacy Scale–International (FES-I)21 are self-report measures that have been used to address these issues in people with PD.18,19,22 Both scales have acceptable internal consistency and test-retest reliability in PD.23 Despite measuring similar constructs,24–26 the ABC and the FES-I have different contents25 and address a variety of activities that might lead to falls.27 However, the accuracy for predicting falls on the basis of self-report measures has not been widely investigated. In a 1-year prospective study, Mak and Pang18 found the ABC to have moderate accuracy for predicting recurrent falls in people with PD, but comparisons with performance-based balance measures were not provided.
Combinations of performance-based measures have been proposed to allow the assessment of different aspects of balance and to improve the identification of a person with PD at risk of falling. These measures include the Berg Balance Scale (BBS); the Functional Reach Test (FRT); the Timed “Up & Go” Test (TUG); the Dynamic Gait Index (DGI)28,29; the Balance Evaluation Systems Test (BESTest), the condensed version of the BESTest (Mini-BESTest); and the Functional Gait Assessment.30–32 Because these tests address different aspects of balance control33,34 and falls are multifactorial,33 self-report measures should be analyzed along with performance-based balance measures for their potential to accurately predict recurrent falls in people with PD. If self-report measures have the same predictive ability as performance-based measures, then self-report measures may be useful for epidemiological or telephone-based surveys when performance evaluations cannot be performed. Additionally, the ABC and FES-I capture information that is related to balance confidence and fear of falling during activities of daily living and that may not be captured by performance-based balance measures. By using both performance-based and self-report measures, treating clinicians will be better able to evaluate findings from examinations and develop treatments to address deficits.
The aims of this study were: (1) to compare the accuracy of the ABC and FES-I with that of the BBS, FRT, TUG, and DGI for predicting recurrent falls in people with PD and, because such assessments are more likely to occur in clinical settings, (2) to explore the ability of combinations of up to 3 tests to predict recurrent falls. We hypothesized that neither self-report measure would show an inferior (poorer) ability to predict recurrent falls than some performance-based balance measures. Furthermore, we hypothesized that the addition of a self-report measure to a model with at least one performance-based balance measure would improve its overall ability to predict recurrent falls.
Method
Participants
Consecutive participants who were able to walk with or without an assistive device and without the assistance of another person were recruited from the Movement Disorders Clinic at the State of Bahia Health Attention Center for the Elderly, Salvador, Bahia, Brazil, between April 2010 and June 2013. The Movement Disorders Clinic is a public referral service through which approximately 700 elderly people with PD and other movement disorders are monitored annually. Figure 1 shows the flowchart for the recruitment and follow-up of participants.
Flowchart showing recruitment and follow-up of participants with Parkinson disease (PD).
All participants were diagnosed with idiopathic PD by a certified neurologist in accordance with UK Parkinson's Disease Society Brain Bank clinical diagnostic criteria.35 Participants were excluded if they had neurological conditions other than PD, cognitive impairment (assessed with the Mini-Mental State Examination and based on cutoff scores determined in previous studies, in accordance with the level of education of each participant)36 or dementia, severe visual disturbance, vestibular dysfunction, or comorbidities that would affect locomotion or balance. Comorbidities were determined on the basis of self-report complemented by chart review. All participants provided written informed consent and were tested during the “on” phase of the medication cycle, approximately 1 to 2 hours after medication intake.
Sample size calculations for the study were conducted with MedCalc 9.5 software (MedCalc Software, Ostend, Belgium). A sample size of 108 participants was estimated on the basis of an alpha level of .05 and a beta level of .10, with an estimated power for the receiver operating characteristic (ROC) curve of .75.
Baseline Assessments
Demographic and clinical data, including age, sex, PD duration, and antiparkinsonian medications taken, were recorded. The levodopa equivalent dose was calculated with a previously reported conversion factor.37 Participants were assessed with the activities of daily living and motor sections of the Unified Parkinson Disease Rating Scale. The scores on the activities of daily living section range from 0 to 52 points (greater disability), and those on the motor section range from 0 to 108 points (greater impairment).38 Additionally, a modified Hoehn and Yahr stage, with scores ranging from 0 to 5 (wheelchair bound or bedridden),39 and a Schwab and England score, ranging from 0% to 100% (completely independent), were assigned.40 The Unified Parkinson Disease Rating Scale was administered by a neurologist (G.T.V.) or a physical therapist (L.R.S.A.) previously trained for this assessment.
Two self-report measures were used to assess participants. The ABC is a 16-item test used to assess self-perceived balance confidence during the performance of daily activities, such as reaching at eye level and on tiptoe, walking up or down stairs or a ramp, and being bumped into while walking. Scores on the ABC range from 0% (no confidence) to 100% (full confidence).20,41 The FES-I is a 16-item tool used to assess fear of falling during the performance of activities of daily living, including getting dressed or undressed, preparing simple meals, and taking a bath or shower. Scores on the FES-I range from 16 points (not at all concerned) to 64 points (very concerned).21,42 The ABC focuses on mobility issues; the FES-I also addresses aspects related to self-care and community, social, and civic activities.25
Participants were also evaluated with 4 performance-based balance measures addressing different aspects of balance control.33,34 The BBS is a well-established tool for the assessment of static and dynamic standing balance. It consists of 14 items related to functional movements and includes tasks that require anticipatory balance control with and without a change in the base of support, such as getting up from a sitting position and tandem standing. Scores on each item range from 0 point (worst) to 4 points (best), and the total score ranges from 0 to 56 points.43,44
The FRT is used as a test of anticipatory balance control without a change in the base of support. It measures, in centimeters, the maximum distance that a person can reach forward beyond arm's length while standing with a stable base of support.45,46
The TUG reflects functional mobility47 and requires participants to stand up from an armchair, walk forward for 3 m, turn around, walk back to the chair, and sit down; results are recorded in seconds. For the present study, participants were allowed to wear their regular footwear and use their customary walking aids.47,48 Two TUG trials were performed by participants; the second trial was recorded as the test result.47
The DGI is an 8-item test used to evaluate a person's ability to adjust balance while walking in response to changing gait task demands. Items include walking forward, changing gait speed, pivot turning, stepping over obstacles, and stair climbing. Scores on each item range from 0 point (worst) to 3 points (best), with the total score ranging from 0 to 24 points.49,50
Procedure
All assessments were performed in the Movement Disorders Clinic on the day on which each participant was recruited. The self-report and performance-based balance measures were administered in the order described above by one physical therapist (L.R.S.A.); the entire test battery was completed in approximately 60 minutes. Participants were allowed to rest as needed at any time during the evaluation. A fall was defined as “an event which results in a person coming to rest unintentionally on the ground or lower level, not as the result of a major intrinsic event or overwhelming hazard.”51(p118) In relation to fall incidence, participants were monitored for 12 months and asked to record on a calendar every fall experienced during the period of interest and its circumstances. Monthly telephone calls were also made to encourage participants, their family, and caregivers to document all falls and to verify the recorded information. Participants were classified as recurrent fallers if they reported 2 or more falls in the 12-month follow-up period and as non–recurrent fallers if they reported no fall or one fall.
Data Analysis
Statistical analysis was performed with IBM SPSS version 21 (IBM Corp, Armonk, New York). The Kolmogorov-Smirnov normality test indicated that, other than the TUG, all of the measures were normally distributed. Descriptive statistics were calculated for demographic and clinical variables, and comparison of recurrent fallers and non–recurrent fallers was completed with the t test, Mann-Whitney U test, and chi-square or Fisher exact test, as appropriate. Falls that did not meet the definition of a fall provided above and data from participants who were lost to follow-up were removed from the analysis.
To select the best-fitting model for predicting recurrent falls in our population, we followed 3 steps for model building. First, ROC curves were developed for each self-report and performance-based balance measure as a predictor of recurrent falls. This approach was chosen because cutoff scores that were previously developed for elderly people (although not specifically people with PD)43,45,47,49 were reported to have low sensitivity for people with PD.46,52 In the present study, optimal cutoff points were chosen on the basis of the Youden Index, which corresponds to the maximum vertical distance between the ROC curve and the diagonal chance line, the point at which both sensitivity and specificity are maximized.53–55 The area under the curve (AUC), which represents the discriminative ability of a test; likelihood ratios for a positive test result (LR+) and a negative test result; and posttest probabilities were also determined. In general, in the analysis of ROC curves, the larger the AUC, the better the test. An AUC of 0.5 was interpreted as a chance result, AUCs of >0.5 to 0.7 were interpreted as indicating low accuracy, AUCs of >0.7 to 0.9 were interpreted as indicating moderate accuracy, and AUCs of >0.9 were interpreted as indicating high accuracy.54–56 Noninferiority tests were used to compare the AUCs of the self-report measures with each other and with those of each performance-based measure57 and, therefore, to determine whether the accuracy of each self-report measure was not inferior to that of the performance-based measures.
Second, each combination of 1, 2, and 3 dichotomous scales to be used as a predictor was evaluated with a separate logistic regression model, with recurrent falls as the dependent variable. Finally, the Akaike information criterion (AIC) was calculated for each model.58 The AIC is a way of comparing different models for a given outcome by estimating the quality of each model, resulting in the selection of models that are more efficient representations of observed data.59 It is derived from the Kullback-Leibler divergence, which measures the distance between a candidate model and the true model. The closer the distance, the more similar the candidate to the truth. Therefore, the AIC is an estimate of the divergence between a candidate model and the true model. Given that the true model is unknown, the relative differences between models can be used to rank order the models in accordance with their expected Kullback-Leibler divergence. The candidate model with the lowest AIC value also has the lowest expected Kullback-Leibler divergence so that models with lower AIC values can be considered to be closer to the true model.59,60 Given this reasoning, the model with the lowest AIC value was chosen as the best-fitting model for predicting recurrent falls in our sample. A significance level of .05 was set for all statistical tests.
Results
Two hundred twenty-nine people with PD were assessed at baseline, and 4 were lost to follow-up (Fig. 1). Of the 225 final participants, 122 (54.2%) were men, and the mean age was 70.7 years (SD=6.6 years). In the 12-month follow-up period, 84 participants (37.3%) were classified as recurrent fallers. This percentage corresponded to the pretest probability of being a recurrent faller. Comparisons of nonrecurrent fallers and recurrent fallers are shown in Table 1.
Demographic and Disease-Specific Characteristics and Balance-Related Measures for Recurrent Fallers and Nonrecurrent Fallersa
Analysis of Individual Self-Report and Performance-Based Balance Measures
The ROC plots for self-report and performance-based balance measures are shown in Figure 2. The AUCs ranged from 0.72 (TUG) to 0.79 (BBS) and were equal for the FES-I and FRT (0.74) and similar for the ABC (0.73) and TUG (0.72) (Tab. 2). The AUC 95% confidence intervals for all measures overlapped. No significant differences were found between the AUC of FES-I and the AUC of ABC or between the AUCs of these self-report measures and those of the performance-based balance measures (P>.05).
Receiver operating characteristic (ROC) curves for the Berg Balance Scale (BBS), Dynamic Gait Index (DGI), Functional Reach Test (FRT), Falls Efficacy Scale–International (FES-I), Activities-specific Balance Confidence Scale (ABC), and Timed “Up & Go” Test (TUG). AUC=area under the curve, CI=confidence interval.
Analysis of Individual Self-Report and Performance-Based Balance Measures for the 12-Month Follow-up Perioda
The AIC values, suggested cutoff scores, and validity indexes are shown in Table 2. The FES-I had slightly higher values for sensitivity, LR+, likelihood ratio for a negative test result, and posttest probabilities with positive and negative test results than the ABC. Moreover, the FES-I had a lower (better) AIC value (122) than the ABC (AIC=131), and their values were close to those for the FRT (AIC=121) and the TUG (AIC=126), respectively. Both self-report measures had higher values for sensitivity and lower values for specificity than the FRT and TUG.
The posttest probabilities of a participant being a recurrent faller ranged from 53% for the ABC to 65% for the FRT, given a positive test result, and from 24% for the FRT to 17% for the BBS, given a negative test result (Tab. 2).
Validity indexes reported in previous prospective studies for performance-based balance measures over a 12-month follow-up period61,62 are also shown in Table 2 to allow comparisons of our data with those of previous studies.
Analysis of Collective Self-Report and Performance-Based Balance Measures
Higher values for specificity and lower values for sensitivity were found for combinations of up to 3 tests with positive results than for most individual measures. Also, the LR+ and posttest probability of a participant being a recurrent faller, given a positive test result, were higher than those for individual measures.
Table 3 shows the AIC values and the validity indexes for each combination of 2 tests with positive results. Six of 15 combinations of 2 tests resulted in lower (better) AICs (range=98–104) than the BBS, the individual measure with the lowest AIC (105). The best models were found when the BBS was combined with the FES-I (AIC=98) and when the BBS was combined with the DGI or TUG (AIC=101). Table 4 shows the indexes for each combination of 3 tests with positive results. The data show a predominance of high values for specificity and similar AICs for the 2-test models and the 3-test models. In this condition, the best models (AIC=97 and 98) were found when the BBS and FES-I were combined with the DGI, TUG, or FRT.
Analysis of Two-Test Modelsa
Analysis of Three-Test Modelsa
Discussion
To our knowledge, this is the first study to prospectively compare 2 self-report measures and several commonly used performance-based balance measures for accuracy in predicting recurrent falls in people with PD. The ability of the FES-I or the ABC to predict recurrent falls was not inferior to that of performance-based measures, with the FES-I having slightly higher validity indexes than the ABC and lower (better) AIC values. Furthermore, combinations of up to 3 tests improved the quality of the predictive model, and 2-test models provided results similar to those obtained with 3-test models. These data favored the choice of a combination of 2 scales to improve the identification of a person with PD at risk of recurrent falls.
The AUC is a summary statistic of diagnostic accuracy across all possible cutoff scores54 and can be used to compare the discriminatory powers of diagnostic tests.54–56 All of the self-report and performance-based balance measures used in the present study showed moderate accuracy for predicting recurrent falls in people with PD. The BBS and DGI had the highest predictive accuracy, although it was not statistically different from that of the FES-I and ABC. Moreover, the BBS and DGI were the individual measures with the lowest AIC values. The FES-I and FRT had the same AUCs and similar AIC values; similar findings were obtained for the ABC and TUG. Given that the higher the AUC, the better the test,54–56 and given that the lower the AIC, the better the model,59,60 it may be argued that the BBS and DGI have better predictive ability than the FES-I and ABC. Taken together, these results suggest that the ability of both self-report measures to predict recurrent falls is comparable to that of the performance-based balance measures, especially the FRT and TUG.
These findings of the present study are in line with those of previous studies, in which the ABC was found to have moderate accuracy for discriminating fallers with PD.18,63 However, those studies did not provide 95% confidence intervals for validity indexes, a noninferiority analysis, or a comparison of self-report and performance-based measures.18,63 In the present study, the FES-I had a slightly higher validity index and a lower AIC value than the ABC, indicating better performance. However, because the validity indexes were so similar, the differences may not have been clinically meaningful; therefore, we suggest that the choice of a specific measure should be based on the aspects addressed and the target population.25 For example, the FES-I covers self-care activities, whereas the ABC does not.25 However, for a person with PD and a high level of function, the ABC would be more appropriate than the FES-I because the ABC includes tasks that are more challenging for postural control.24
In the present study, the similarities in predictive ability among the FES-I, ABC, FRT, and TUG may have been related to the content addressed by these measures. The FRT measures limits of stability, and poor performance is represented by a loss of balance during reaching. Reaching is a common daily activity that can be performed by a person who is cleaning the house, preparing meals, getting dressed, taking a shower, or reaching for something above the head or on the ground—all tasks that are evaluated by the FES-I.21,25 The ABC is a scale that clearly focuses on mobility, rating many ambulatory activities20,25; because the TUG also addresses functional mobility,47 their relationship may be linked in that way.
Some important points about self-report and performance-based balance measures must be considered. Performance-based measures, although relatively free from response bias, represent just a single point in time. For a disease such as PD, which involves high variability in balance even within 1 day, these measures may not provide an accurate representation of day-to-day balance performance. Likewise, self-report measures do not provide clinicians with information about physical performance during all activities, perhaps leading to a misunderstanding of the mechanisms of falls. On the other hand, these scales rate fear of falling (FES-I) and balance confidence (ABC) during activities of daily living, indoors and outdoors; therefore, they reflect real-life situations in which a fall can occur and presumably would take into account day-to-day variability in balance. Given these properties, self-report measures, although more likely to be biased (because people can underestimate or overestimate their deficits), may provide a good indication of balance capabilities because they likely incorporate respondents' sense of their best and worst balance capabilities for each task. Furthermore, they can provide insight into reasons for a fall and guide fall prevention programs.
Various combinations of self-report and performance-based balance measures provided a diversity of validity indexes, depending on which measures were chosen. The AIC was chosen as our criterion for selecting which combination of up to 3 measures had the best predictive capability, minimizing rates of both false-positive and false-negative results. The AIC estimated the quality of each model; the basic principle of an information-theoretic criterion is to select statistical models that simplify the description of data.59 Therefore, the model with the lowest AIC was selected as the best-fitting model for predicting recurrent falls in our population. However, depending on the clinical scenario, scale combinations that are focused on minimizing rates of false-negative results may be chosen.
In the present study, combinations of 2 or 3 scales improved the ability to predict recurrent falls given that lower (better) AIC values were obtained with the combinations than with individual measures; these results are in line with those of previous studies because of the assessment of multiple balance components.22,28,29,32 Moreover, most of the 3-test models had predictive ability similar to that of the 2-test models but had high values for specificity, which may have increased rates of false-negative results. Therefore, we suggest that a combination of 2 tests may be the best choice. Most models allowed for higher LR+ and posttest probabilities of a participant being a recurrent faller, given a positive test result, than individual measures. Despite the fact that the posttest probabilities of a participant being a recurrent faller, given a negative test result, were similar for models with the lowest AIC values (BBS combined with FES-I, DGI, or TUG) and for individual measures, the former models produced the lowest probabilities. The best-fitting 2-test model was the combination of the BBS and FES-I. This model produced lower a AIC value than the individual measure with the lowest AIC value. Moreover, this model had reasonable values for sensitivity and specificity and higher posttest probability, given a positive test result, and LR+; these results indicated that recurrent fallers were 3.85 times more likely to have positive test results (ie, to have a BBS score of ≤49 points and an FES-I score of >29 points) than nonrecurrent fallers.64
Despite the possibility of a ceiling effect,32,65 some tasks of the BBS have been described as common situations for falls in people with PD; these include transferring, reaching, turning, and gait.27 Furthermore, the fear of falling is evaluated by similar tasks in the FES-I. These properties may explain why these tests had the best ability to predict recurrent falls. A mixed model (self-report and performance-based balance measures) incorporates the strengths of both types of measures and perhaps mitigates the weaknesses as well. Furthermore, it may help clinicians understand the level of balance impairment in a clinical setting and may provide insight about other mechanisms that can be related to falls, especially daily activities in which people have a greater fear of falling. Moreover, the FES-I can be completed by the patient or caregiver; this feature is important in clinical settings with time constraints.
Previous prospective studies provided comparisons of performance-based balance measures in people with PD but did not include self-report measures.61,62 Duncan et al62 found the BESTest, Mini-BESTest, BBS, and Functional Gait Assessment to have moderate accuracy for predicting recurrent falls in people with PD over a 12-month period. The values that they reported are similar to our AUCs for both self-report measures, but their confidence intervals were close to or crossed the cutoff for a chance result (AUC=0.5).62 Moreover, most of the balance measures were less sensitive and had specificity and LR+ comparable to or higher than those of the FES-I and ABC.62 Duncan et al61 noted that the shortened version of the BESTest (Brief-BESTest) had accuracy similar to that of the previously examined balance measures for predicting recurrent falls over a 12-month period. In both studies, the predictive values at 6 months were also investigated, and the predictive accuracy at 6 months was found to be better than that at 12 months.61,62 However, comparisons with our results are limited because such an analysis was not an aim of the present study. Our 2-test models had higher validity indexes for predicting who would be a recurrent faller in the next 12 months than the individual measures evaluated by Duncan et al.62
The main strength of the present study was the simultaneous testing of various fall prediction scales, including 2 self-report measures, in a single sample population. Moreover, we provided cutoff scores based on a prospective classification of recurrent fallers and using a method that reduced rates of false-positive and false-negative results, in line with the approaches used in previous studies.30,32,61 We suggest that the cutoff scores should be used to help clinicians understand balance capabilities and identify who has a higher risk of falls, not as definitive dichotomous scales, as also suggested in other studies.30,32,61,62
A limitation of the present study is the lack of an external validation sample. Cohort studies comparing the abilities of various balance measures to predict recurrent falls can be found in the literature,61,62 but we are not aware of prospective studies designed to compare self-report and performance-based balance measures in people with PD. Future research should evaluate whether self-report measures have a better ability to predict recurrent falls at different follow-up periods and whether they add predictive power in models with the more recently developed BESTest, Mini-BESTest, Brief-BESTest, and Functional Gait Assessment.30,32,61
In conclusion, self-report measures were shown to have accuracy similar to that of performance-based balance measures (especially the FRT and TUG) for predicting recurrent falls over a 12-month period in people with PD. The FES-I had slightly higher validity indexes than the ABC, although this difference may not have been clinically meaningful. Two-test models had performance similar to that of 3-test models, suggesting that a combination of 2 measures may improve the ability to predict recurrent falls. Specifically, the use of the BBS in combination with the FES-I may be considered as a way to improve the identification of people with PD at risk of recurrent falls in the next year.
Footnotes
All authors provided concept/idea/research design. Ms Almeida provided writing. Ms Almeida, Dr Valenca, and Ms Negreiros provided data collection. Ms Almeida and Dr Oliveira-Filho provided data analysis. Ms Almeida, Dr Valenca, Dr Pinto, and Dr Oliveira-Filho provided project management. Dr Valenca and Ms Negreiros provided participants and facilities/equipment. Dr Valenca, Ms Negreiros, Dr Pinto, and Dr Oliveira-Filho provided consultation (including review of manuscript before submission). The authors thank Antônio Cunha Porto Maia for his statistical consultations.
This study was approved by the Comitê de Ética em Pesquisa da Secretaria da Saúde do Estado da Bahia, Escola Estadual de Saúde Pública, Salvador, Bahia, Brazil.
- Received March 20, 2015.
- Accepted January 17, 2016.
- © 2016 American Physical Therapy Association
References
Issue highlights
