Independent domains of daily mobility in patients with neurological gait disorders

Disorders of gait and balance and associated mobility impairments are a common complication in neurological diseases and affect around 60% of patients [41]. Highest prevalence for gait and mobility impairments have been reported for patients with Parkinson's disease, followed by vascular encephalopathy, cerebellar ataxia, and sensory deficits. Limitations in mobility function can be debilitating with considerable consequences for patients' functional independence, social participation, and overall quality of life [24, 38]. Moreover, gait impairments in these patients are linked to an increased risk of falling [34, 42] and injuries resulting from falls not only entail substantial medical costs but also determine patients' mortality risk [9].

Disease-specific clinical scales or outcomes from neurological functional diagnostics frequently fail to adequately reflect or predict the degree of mobility restrictions and the risk of falling in different neurological disease cohorts [7, 10, 35]. It appears therefore crucial to directly and comprehensively examine gait and mobility function in these patients. In effect, functional mobility tests (e.g., the Timed Up and Go test [29]) and instrument-based measures of gait stability have been shown to more adequately capture mobility impairments and to more reliably identify a risk of falling in central and peripheral neurological [28, 32, 33, 36, 37, 45] as well as in geriatric patients [1, 13, 25]. However, a drawback of these approaches lies in the fact that they are primarily performed within the laboratory or a specialized clinical setting. They may therefore miss or underestimate the challenges of real-life mobility during which patients actually fall.

In this context, recent advances in daily mobility assessment using body-worn inertial sensors promise a more adequate and specific characterization of patients' mobility impairments in ecologically valid and clinically relevant settings [20, 46]. In contrast to an instrument-based gait assessment in the clinical setting, which focuses on the detailed spatiotemporal features of a patient's gait pattern, body-worn sensors are primarily used to assess patients' mobility from a macroscopic perspective [20]. Accordingly, sensor-based activity monitoring has been focused on measures that allow to quantify how single epochs (also referred to as 'bouts' [3]) of activity (e.g., walking, standing, sitting, lying) alternate and accumulate over time. A plethora of such mobility measures have been proposed in the past (e.g., number of steps, intensity of activity, number and variability of activity bouts, etc. [4, 19, 31, 39]) and there is yet no consensus on what parameters best capture the essence of real-world mobility performance and are clinically most relevant.

Using techniques from factor analysis, Lord and colleagues could previously demonstrate that a large set of available mobility measures can be reduced to a meaningful, small set of independent mobility domains that adequately capture daily-life activity in community-dwelling older adults [19]. Based on this previous work, the aim of the present study was (1) to examine whether an analogous approach might yield a parsimonious and adequate model of daily mobility activity in patients with neurological gait disorders and (2) to explore which explanatory characteristics might determine differences of patients' mobility performance within this model.

Descriptive statistics on personal characteristics as well as outcomes from clinical assessment and daily mobility monitoring can be found in Table 1. Patients and healthy controls did not differ in basic demographics parameters. In accordance to the recruitment criterion on preserved independent walking ability, patients exhibited only moderately impaired functional mobility scores in clinical assessment. They further yielded near-to-normal cognitive function but reported lower confidence in balance and increased concerns about falling. In correspondence, retrospective fall assessment revealed that patients had considerably more often fallen in the past with significantly more severe consequences of falling (according to the Hopkins grading scale). Parameters from daily mobility monitoring in patients were predominantly within the normal range or only moderately impaired, in particular in terms of a reduced daily duration and intensity of ambulatory behavior.

Out of the in total 14 included mobility measures, PCA yielded five orthogonal factors that did not exhibit notable cross-loadings. All item loadings were greater than 0.7 (Table 2). The five obtained mobility domains account for 92.3% of the total variance in activity monitor measures (Fig. 1a). We designate these domains as (1) ambulatory volume (i.e., the total amount of ambulatory activity; accounting for 38.7% of total variance), (2) ambulatory pattern (i.e., the temporal distribution and variability of different bouts of ambulatory activity; accounting for 22.3% of total variance), (3) postural transitions (accounting for 13.3% of total variance), (4) sedentary volume (i.e., the total amount of sedentary activity; accounting for 10.8% of total variance), and (5) sedentary pattern (i.e., the temporal distribution and variability of different bouts of sedentary activity; accounting for 7.2% of total variance). Daily mobility performance in patients and healthy controls was compared based on the five resultant daily mobility domains. This analysis revealed that patients only showed reduced scores in the ambulatory volume domain but had near-to-normal scores in all remaining domains.

In a next step, associations between the obtained mobility domains and various explanatory characteristics were analyzed. For each mobility domain, we found differential moderate associations with demographic items (age, gender), confidence (FES-I), and functional mobility (FGA, TUG) scores (Table 3). These characteristics explained between 19.2% (for ambulatory volume) and 10.2% (for ambulatory pattern) of the total variance within domains. Falls grade (i.e., the most severe consequences of past fall events) and subjective fear of falling (i.e., FES-I) had the overall strongest impact on ambulatory volume. Functional mobility scores (i.e., FGA and TUG) were primarily associated with the two mobility domains related to sedentary behavior.

Assessment of gait and mobility function is gaining increasing importance for clinical diagnostics or monitoring of disease progression, as a measure for the efficacy of interventions, and as a marker for identifying those patients at a high risk of falling. However, for gait as well as for daily mobility assessment, clinicians are often unable to decide, which of the plenty available outcome measures are the most appropriate for each of the above-mentioned applications [20]. In the past, factor analysis techniques have been successfully applied to conceptualize both types of assessment, in terms of a low-dimensional set of independent functional domains (typically 3–5) that adequately cover the essential aspects of the general gait pattern at an microscopic and the everyday mobility performance at an macroscopic perspective [14, 19, 21, 45]. However, most of these conceptualizations and models were developed based on datasets derived from the healthy elderly population, which yet impedes a direct application to the clinical field of neurological gait and balance disorders.

Here we demonstrate that the approach of low-dimensional modelling of real-world mobility behavior can be extended to patients with different forms of neurological gait disorders. Our obtained model includes five independent functional mobility domains, which we designate as ambulatory volume, ambulatory pattern, postural transitions, sedentary volume, and sedentary pattern. This model shares several communalities to the previously presented model from Lord and colleagues that was obtained from a dataset on community-dwelling older adults [19]. In particular, in both models, measures of ambulatory and sedentary activities as well as measures of transitions between these activities group in separate functional domains. Intuitively, quantities of ambulatory and sedentary behavior should yield redundant information as they apparently represent opposing ends of the very same phenomenon. However, both models emphasize that the characterization of sedentary behavior and postural transitions between activities entails complementary information that is required to comprehensively capture the pattern of activities during daily-life routine. In contrast, to Lord's model that is composed of only three domains (i.e., ambulatory behavior, sedentary behavior, and postural transitions), the current model comprises five functional domains inasmuch as ambulatory and sedentary activities are each further differentiated into a domain that reflects the volume and intensity of behavior and a domain that represents the distribution and regularity of behavior. This differentiation is in accordance with previous reports that indicate that both the volume and the diversity of activities reflect different aspects of the general mobility status of individuals [11, 22, 31]. Thus, the discrepancy between the two models presumably results from the apparently different ambulatory status of the two study populations. Accordingly, reduced activity levels in the elderly population in Lord's study might eventuate in generally fewer degrees of freedom within the pattern of daily activities and a collapse of otherwise functionally independent domains of behavior [16].

The current analysis focused on a patient cohort with an early-stage manifestation of different neurological gait impairments with yet retained independent ambulatory status (predominantly from a population at working age). This selection criterion was reflected in the obtained model, according to which patients exhibited only moderately reduced scores in the ambulatory volume domain compared to age-matched healthy participants but near-to-normal scores in all remaining domains. Thus, patients in our cohort tended to maintain their everyday routine (i.e., a near-to-normal pattern of different activities over the day) despite showing initial signs of a disease-related reduction on mobility volume. Regression analysis revealed that reduced performance in ambulatory volume domain was associated to a lower balance confidence and increased concerns about eventual falls. This association indicates that not only fall-related injuries that actually impair mobility function but already a lower balance confidence resulting from an incipient decline of postural function and/or from former falls may considerably affect the amount of everyday mobility [19]. Moreover, higher scores of ambulatory volumes in patients were associated to more severe consequences of falling. In accordance with previous reports, this finding suggests that in particular for patients with early-stage gait disorders a longer exposure to walking situations entails a higher risk for the occurrence of frequent and especially of severe fall events [23].

Consistent with previous work, the overall outcome of regression analysis on mobility domains was only modest [19]. Despite including a comprehensive set of demographic, subjective confidence, and clinical characteristics, we did not identify explanatory characteristics that would convincingly account for individual performance differences along the dimensions of the five different mobility domains. With respect to functional mobility scores (i.e., FGA and TUG), this observation indicates that real-world measures of mobility provide complementary information on patients' mobility status that is not readily available from clinical evaluation. Counterintuitively, patients' quality of life as assessed by the SF-12 score did not at all show any association to the different aspects of patients' daily-life activity. This observation might at first glance contradict the high expectations on long-term mobility assessment as an specific and patient-relevant outcome measure for future clinical trials [46]. However, our focus on a clinical population with predominantly early-stage gait disorders that exhibited near-to-normal daily mobility performance and quality of life scores certainly limits this observation. It is indeed conceivable that more pronounced impairments of daily mobility in advanced stages of neurological gait disorders may be actually more strongly associated to alterations in patients' quality of life. Further studies that include a more comprehensive study population and more specific measures of health-related quality of life are therefore required to corroborate or disprove this observation.

In conclusion, the here established model for daily mobility behavior in neurological gait disorders may provide a convenient framework for future studies on disease-specific mobility impairments and their consequences for patients' well-being, quality of life, and risk of falling. The mainly negative results from regression analysis demand for further research to explore the characteristics that may sufficiently explain performance differences within the proposed functional mobility domains.