Background

Health professionals commonly use gait speed in the evaluation of functional status in older people. In addition, it is a good predictor for adverse consequences, such as disability, hospitalization or death [1]. In this context, most of the articles about gait speed focus on the differences between people suffering from diseases which cause gait disorders and people who do not [2, 3], on the effectiveness of treatments [4] or on the prediction of adverse outcomes [5]. However, there are a limited number of studies on people without gait disorders, usually focused on establishing reference parameters in healthy people with a clear underrepresentation of older people [6,7,8,9,10,11]. When correlates of usual gait speed in normal or well-functioning older adults are analysed [8,9,10,11], there is a tendency to ignore some factors that are almost exclusive to older people like the prevalence of polypharmacy, recurrent falls, sarcopenia or frailty, although they can affect gait speed.

Our study aims to analyse which factors are associated with changes in gait speed in older people without clinically evident gait disorders.

Methods

Setting and study design

We designed an observational cross-sectional study. Volunteers were companions of patients attending a day hospital or people reached through posters or older people’s associations. Inclusion criteria were being older than 70 years, living in community dwellings and being able to walk without aids with a normal gait, as assessed by visual observation. We described “normal gait” as an independent, symmetric and upright gait, with good hip and lower limbs alternance and no dragging of the feet. A total of 235 volunteered to participate and met the inclusion criteria. However, we had to exclude 115 patients because of the presence of any of the following exclusion criteria: patients with a Mini-Mental State Exam (the Spanish version is called MEC) < 25; Barthel index < 85; terminal status (life expectancy < 6 months); acute medical illness in the past 3 months; major depression or psychiatric follow-up; neurological diseases (including Parkinson’s disease, cerebellar disease or peripheral neuropathy); peripheral artery disease; and coronary heart disease. We excluded one more volunteer, because he did not have all variables measured, resulting in a final study sample of 119 persons.

Measures

Dependent variable: gait speed

We assessed gait speed using the 6-m test. We asked participants to walk 10 m in their usual way to measure the speed at which they normally walked in the street. We started the timer at the 2 m and stopped it at the 8-metre mark. We took two measurements (in metres per second—m/s) and used the best one for the analysis.

Independent variables

We limited the number of variables to have no fewer than five persons per variable in the sample [12].

Sociodemographic variables

We considered gender, age and age2.

Morphological variables

Morphological variables such as height (cm); waist perimeter (cm); waist-to-height ratio (ratio of the waist perimeter to height); body mass index (BMI) in kg/m2; total appendicular mass (TAM) measured through impedanciometry in kilograms (Tanita BC-418MA bioimpedanciometry); ratio TAM/BMI; press strength measured in kilograms [13] and press strength2 were calculated. Apart from the components of sarcopenia individually, we also considered its diagnosis according to the FNIH criteria [14].

Conditions

We measured cognition through the MEC score [15]. We operationalized frailty as the number of Fried’s phenotype criteria standardized to the Spanish population [16, 17]; we also introduced the number of criteria2 in the model.

Low physical activity

We used the item of the Fried’s phenotype to have a measure of the low physical activity performed by people. We considered low physical activity when men walked less than 2.5 h/week and women less than 2 h/ week. Press strength was the other Fried’s item that was incorporated individually into the models.

Falls

We considered the number of falls suffered in the last year and the total number of falls2.

Drugs

The number of drugs and number of drugs2 were considered; we defined the consumption of four or more drugs as polypharmacy [18]. In addition, we incorporated taking psycho-drugs (participants only took antidepressants and sedative anxiolytics—subgroups N05 and N06 of the ATC classification) [19], antihypertensive, antiarrhythmic drugs or diuretics (subgroups C02, C03, C04, C07, C08 and C09 of the ATC classification).

Descriptive variables

The descriptive variables were basic activities of daily living (BADL) (measured with the Barthel index) [20], instrumental activities of daily living (IADL) (measured with the Lawton and Brody Index) [21] and comorbidity (measured with the Charlson Index) [22].

Statistical analyses

We described variables using arithmetical means and standard deviations or proportions where appropriate. We calculated bivariate linear regressions on gait speed for each variable. To obtain the best set of variables associated to gait speed, we used a stepwise regression strategy, where all variables are entered at the first step and eliminated one by one if their statistical significance is higher than 0.1. Excluded variables may re-enter the model if their significance level at inclusion is less than 0.05. We set tolerance at 0.01 [23]. We checked residuals of the final model, as well as outliers (univariate and multivariate) and influential observations. We performed all statistical tests using SPSS for Windows (13.0 version).

Results

The characteristics of the sample are presented in Table 1. The mean age was 78 years, and 71.4% were women. As expected because of the exclusion criteria, the participants were free from significant disability, cognitive impairment and comorbidity. 31.1% were obese (BMI ≥ 30), and 46.2% suffered from overweight (BMI 25–29.9). 12% suffered from sarcopenia and the mean number of Fried’s criteria was 0.42. Weakness and low physical activity were the most common frailty criteria, present in one-tenth of the sample. The mean number of falls in the previous year was 0.64. Polypharmacy and consumption of psycho-drugs and cardiovascular drugs were common (27.7%, 37.8% and 62.2%, respectively). Average gait speed was 1.13 m/s; SD = 0.27 m/s, ranging from 0.32 to 1.96 m/s.

Table 1 Characteristics of the sample tested by variables
Full size table

As shown in Table 2, bivariate regression showed significant associations in the expected direction between gait speed and the following variables: age, gender, MEC, height, waist-to-height ratio, press strength, TAM, TAM to BMI index, number of drugs, polypharmacy ≥ 4, consumption of psycho-drugs, number of falls, number of frailty criteria and low physical activity. There was no significant association with waist circumference, BMI, sarcopenia, and consumption of antihypertensive drugs, diuretics or beta blockers.

Table 2 Variables could associate on gait
Full size table

The final stepwise model included age (B = − 0.020, p < 0.001); gender (B = − 0.184, p < 0.001); waist-to-height ratio (B = − 0.834, p = 0.002); number of falls (B = − 0.049, p = 0.003) and number of Fried’s frailty criteria (B = − 0.064, p = 0.019). These variables explained over 50% of the inter-subject variability in gait speed (R2 = 0.53).

On average, being a woman or being 10 years older decreased the gait speed by 0.2 m/s (p < 0.001). For every unit increase of the waist-to-height ratio, gait speed decreased by 0.8 m/s (p = 0.002). For each two additional frailty criteria or each two more falls experienced in the previous year, gait speed decreased by 0.1 m/s (p = 0.019 and p = 0.003, respectively).

Discussion

The aim of this study was to assess the relationship between gait speed and the factors that theoretically could affect it in older patients with clinically normal gait, including some that are almost exclusive to older people. Age, gender, waist-to-height ratio, number of frailty criteria and number of falls were the factors we found to be associated with gait speed.

Age, gender and waist circumference and height are classical factors that have been reported to be associated with gait in older and younger people [11]. However, there are just a few studies that analysed the waist-to-height ratio, even though this ratio has proven to be one of the best cardiovascular assessment index for the older people [24]. In the line of these results, the waist-to-height ratio was the only morphological one that remained associated with gait speed in the multivariate model.

Frailty is a pathologic status of older people and its prevalence increases with age [25]. We operationalized it with the Fried’s phenotype. We considered its individual components that could theoretically be related to gait speed, such as press strength and low physical activity, but we also wanted to explore the association of the index as a whole. The more frailty criteria an individual had, the lower his/her gait speed was. Of note, only one person in the sample met the Fried’s slowness criterion, so the association cannot be attributed to the presence of this component that is based on a measure of gait speed.

We found an association between an increasing number of falls and a decrease in gait speed. Other studies support these results [26]. Fear to fall produced by previous falls could reduce gait speed [27]. In any case, with our cross-sectional design, it is impossible to determine if falls are the cause or the consequence of a low gait speed. We found that patients with two or more falls experienced a reduction of gait speed above 0.1 m/s. This is an important change that has shown predictive power of adverse events [1, 5].

Prevalence of falls increases with age and frailty level [28]. Though it is clear that frailty and recurrent falls cannot be considered as factors associated to a healthy population [26, 28], other authors do not include them among their exclusion criteria when choosing their sample [8,9,10,11]. Using the same exclusion criteria, we have found patients who suffer from both conditions.

We have found a very weak association between muscle mass or sarcopenia and gait speed. It must be said that in the stepwise regression, sarcopenia remained in the model until frailty was introduced, suggesting that the presence of frailty explains all the effects of sarcopenia on gait speed. Supporting this explanation, several authors have found a strong association between this phenotype and sarcopenia [29].

Polypharmacy and psycho-drugs variables lost their significance in the multivariate model when we adjusted by age, gender and the number of frailty criteria. Confounding factors could also play a role in the loss of significance. Alternatively, the effect of drugs could be mediated by frailty [28]. In relation to the consumption of antihypertensive drugs, diuretics or beta blockers, they did not show an association even in the bivariate model. This could be due to the fact that what conditions gait speed is changes in these medications, not just their usage [30]. The use of absence of clinical instability in the last 3 months as an inclusion criteria in our study and the review of patients’ clinical records suggest that this circumstance did not happen in our sample.

Regarding cognitive function, even in our sample with no cognitive impairment, we found a bivariate relation of MEC with gait speed. This effect disappeared when we adjusted our model by age, gender and number of frailty criteria. Again, confounding by these variables which are associated with cognitive impairment [28] could be the reason.

According to our results, health professionals should consider age and gender, waist-to-height ratio, number of falls and the number of Fried’s frailty criteria as adjustment factors when studying other variables associated with gait speed and falls. Other studies did not take into account frailty and falls in their designs, and their findings can be biased.

The main strength of this study is that it allowed studying a wide number of variables in people without disability and with a normal gait. This also acts as a limitation because our volunteers were relatively healthy and had a low prevalence of factors that might be associated with gait, such as sarcopenia, physical inactivity or depression; this circumstance may also limit the external validity of the study. However, the need of excluding patients with any disturbance of gait precludes the inclusion of people with other conditions that produce abnormal clinical gait. The main drawback is the cross-sectional nature of the design that prevents us from establishing causal relationships.

Conclusion

There is a lack of information about factors associated with gait speed in older people without gait abnormalities, which forces the scientific and clinical community to use thresholds for “normal” gait speed and consider factors associated with changes in gait speed from studies in younger people. This raises the possibility of bias. This work proposes new factors that should be considered as potential confounders in the study of variations of gait speed in older people with clinically normal gait.