Psychometric properties of instrumented postural sway measures recorded in community settings in independent living older adults

Normal aging is related to declines in different body systems such as cardiovascular, sensory, musculoskeletal, and cognitive function, all of which have been associated with increased risk of falling [1]. It is well documented that aging itself also is associated with a decline in muscle strength, balance, and functional mobility [2]. Maintaining postural stability is imperative for older adults to perform activities of daily living safely and independently within their society and thereby avoid falls [3]. Balance impairments are risk factors that contribute to mobility limitations and falls in older adults [1].

Because maintaining balance and mobility is important to successful aging, the assessment of balance is important for identifying older adults who are at high risk of falling, and also for developing appropriate exercise interventions to address any impairments. In order to achieve postural stability during standing, a person must be able to control the vertical projection of the center of mass within the base of support in the antero-posterior (AP, forward-backward) and medio-lateral (ML, side-to-side) directions. The measurement of body sway using an accelerometer around the waist can be used to record these movements of the center of mass, which is an advantage over wrist-mounted accelerometers commonly in use. Reliable and valid assessment instruments are necessary to obtain consistent and repeatable measurements for static standing balance. Currently, the most common methods to examine balance in clinical settings include observation-based measures; yet these measures have been shown to have examiner’s bias [4], suffer from floor and ceiling effects [5], cover limited aspects of balance, and often lack sensitivity to detect small changes in balance [6]. These drawbacks are major concerns for both clinicians and researchers who treat balance impairments and investigate the effectiveness of different balance interventions.

Over the last two decades, quantitative assessments of postural sway during standing using tools such as force plates have been used to assess postural stability and identify balance dysfunction in elderly population. Force plates have demonstrated good to excellent reliability for recording postural sway. However, because of the expense, space requirements, and lack of portability, their clinical utility in the community has been limited. Recent advances have provided an alternative quantitative method to assess balance that is inexpensive and portable by using body-worn accelerometers. Accelerometers are used to quantify postural sway during standing, and have been shown to have the ability to discriminate between test conditions that require different levels of postural control, between fallers and non-fallers, and young versus older adults [7‐9]. Assessing balance by using accelerometers has been applied to different populations including people with Parkinson disease [10], stroke, children, and with community-dwelling older adults [11, 12]. Previous studies that have used accelerometers have demonstrated good to excellent test-retest reliability of postural sway measurements during the static standing balance [8, 12]. However, these accelerometer reliability studies were limited to clinical and lab settings, and had not been investigated outside in the community. Recently, a study by Saunders et al., found good to excellent test-retest reliability in using a tri-axial accelerometer to assess postural stability in people who live in independent living facilities [9].

To bridge the gap between expensive and immobile instruments and task-based measures, and by taking advantage of technological advancements in accelerometers, postural stability can be quantified portably and inexpensively outside of a lab setting. These tools can serve understudied populations, such as people living in community settings, who may have difficulty getting transportation to research labs. Therefore, the aim of this study was to establish the psychometric properties of balance measurements in older adults using an accelerometer.

Demographic and clinical characteristics of the study sample are summarized in Table 1. The subsample used for the reliability tests was 4 years younger than the total sample used for the validity examination. Compared with the total sample, the reliability subsample had a higher prevalence of diabetes, more comorbidities, and higher BMI.

Across the six balance conditions, the sway measure that produced the greatest reliability was the normalized path length in the mediolateral direction, with ICC scores ranging from 0.61 to 0.81. In addition, some of the other sway measures had excellent reliability for specific test conditions. Only two measures had poor reliability: the AP RMS and AP NPL during semi-tandem stance. The current study had greater reliability coefficients compared with other published studies during the conditions on level surface with eyes open and closed [8, 28‐30], possibly because the age range of our participants was larger, which may have produced greater intersubject variability.

Conversely, the current study had lower reliability than the study of Saunders et al. (2015) [9], who reported ICCs ranging from 0.77–0.93 for standing on a firm surface with eyes open and closed and ICCs from 0.76–0.95 for standing on foam surface. There are several possible reasons for the higher reliability in the Saunders study. In the Saunders study, they used the average of three trials for each balance condition, which would increase the ICC value compared to one trial in our study. It has been shown previously that test–retest reliability increased as the number of trials increase [31]. In the present study, to avoid fatigue of the elderly participants, only one trial was done. In addition, the retest session for the Saunders study was conducted within the same day. Evaluating test–retest reliability within-day has been shown to improve the ICC estimate as compared to between-day estimation [29]. Finally, they used a different foam surface than we used, and foam density and thickness can affect postural stability [32].

Our results for the NPL parameters were consistent with previous findings that used similar accelerometers for standing on a foam surface with eyes open and eyes closed in the AP direction [15, 33]. However, our results in these two conditions were slightly lower than results from Rine et al., (2013) [12], who reported an ICC of 0.88 for standing on foam with eyes open and 0.87 with eyes closed. In their study the retesting was done within the same day which could have yielded these higher ICC values.

The test–retest reliability during standing in semi-tandem and tandem stance was higher for the ML direction as opposed to the AP directions for both NPL and RMS sway. The semi-tandem and tandem stance conditions place more emphasis on the control of stance in the ML axis than AP, which seems to be more clinically relevant as ML sway has been associated with fall history [34]. Similarly, Moe-Nilssen et al. found higher ICCs for RMS acceleration in the ML (ICC = 0.84) than AP (ICC = 0.69) during standing on 1 foot where the base of support is more limited in the ML direction, thus providing support to our current findings [28].

The estimate of absolute reliability as indicated by the SEM and MDC provide researchers and clinicians with the ability to quantify the error during measurement and accurately estimate the true change on balance performance. Williams et al. 2016, reported similar MDC values for standing on a firm surface with eyes open and eyes closed using a triaxial accelerometer [29]. A smaller SEM and MDC indicates a more reliable measure. Larger SEM and MDC measures in this study may be attributed to: greater within-subject variability that is expected in older adults compared to other age groups; lack of a familiarization trial before test measurement, and not including more than one trial per session. In addition, the length of trial recording influences the reliability estimates with longer recordings associated with higher reliability. A duration of up to 120 s are suggested to reduce measurement error [35]. We used a 30-s sampling duration to match the abilities of older adults, who might not tolerate standing for an optimum duration.

Postural sway increased as the balance conditions became more challenging, thus demonstrating face validity of the accelerometer measurements. When somatosensory input was reduced by using a foam pad, the older adults generated greater body sway compared with standing on firm surface. Moreover, during conditions where visual inputs were absent, body sway increased as compared to eyes open conditions. Therefore, this has direct impact on older adults’ everyday lives, especially those with peripheral neuropathy or visual impairments who tend to have difficulty maintaining postural stability when walking on a carpeted floor or in a dark room. Our results are consistent with previous studies using a similar accelerometer [8, 25, 28]. In addition, the current results showed that the NPL sway in the AP axis when standing on foam with eyes closed was larger than the sway of healthy older adults with a mean age 47 years from a previous study that used a similar accelerometer, which further validates the measurements [33].

The Spearman correlation results showed a significant correlation in 17/24 of the balance parameters with the total SPPB score, and in 22 of the 24 of the correlations with the balance component of the SPPB, indicating convergent validity. To the best of our knowledge this is the first study that examined the correlation between balance accelerometry and the SPPB. Among all the included balance parameters, the highest correlation coefficients between sway measures and the balance component of the SPPB were the ML RMS sway during standing in semi-tandem and tandem stances (Spearman rho = 0.43 and 0.44, respectively). A simple explanation for this finding is that the semi-tandem and tandem balance conditions used for the accelerometer test mirrors the SPPB balance subtest. Previous studies showed similar results when comparing center of pressure measures using a force platform with clinical-based measures such as the SPPB [36, 37]. However, the moderate correlation indicates that different aspects of balance are being measured by the accelerometer-based measurements. The GES was significantly correlated with 15/24 of the sway measures. The highest value of correlation coefficients among the sway measures occurred in the foam, eyes open condition, and semi-tandem and tandem stances. These results indicate that individuals with greater sway had less confidence in their walking during everyday activities. Although, the correlation coefficients were significant, the strength of the relationship between the GES and sway measures was weak. This weak relationship could be explained by that the GES represents a person’s rating of their own confidence performing different walking-related tasks, whereas the balance accelerometry captures balance performance in standing only. A study that used another self-efficacy scale, such as the Activities-specific Balance Confidence (ABC) scale, which was highly correlated with the GES, showed a similar correlation between postural sway and the ABC scale [38].

The strengths of the current study are several. First, balance performance was quantified using a reliable method established in this specific population: i.e. older adults who live in independent living facilities. Second, we included various balance conditions that were designed to challenge and examine different balance sensory systems. Interpretation of the current findings should be considered in light of the following limitations. The sample in the current study was not randomly chosen from the parent study’s sample because this was an ancillary study to a multi-site cluster randomized trial, in which a subsample of the sites were chosen. However, baseline characteristics in our study were similar as compared to parent study. Another limitation is that we only included static standing balance conditions that examined one aspect of the balance system. Future research that includes dynamic balance tasks such as those in the Berg Balance scale could be done to explore the psychometric properties further. The reason for not including the dynamic conditions in this study is that older adults may not have tolerated a longer testing time, given that most of testing sessions were done after they finished testing from the parent study within the same day.

The dual-axis accelerometer provides a feasible, reliable, and inexpensive method for testing standing balance in older adults. Among the included sway measures, the ML NPL measures demonstrated the highest test-retest reliability. Therefore, we recommend using these parameters to obtain a highly reliable measurement of sway in this population. Implementing the accelerometer technology may help investigators access understudied older populations living in independent living facilities, and will allow clinicians to examine objective measurements in real-life environments. Hopefully through the use of technology clinicians and therapists can prescribe interventions based on the individual’s objectively identified balance deficits.