Figures
Abstract
The increased fall risk associated with the use of psychotropic drugs might be caused by underlying problems in postural control that are induced by sedative side-effects of these drugs. The current literature on the effects of psychotropics on postural control only examined acute single-drug effects, and included relatively healthy young elderly. Consequently, it is unclear what the impact of the long-term use of these drugs is on gait in frail older persons with polypharmacy. Therefore, it was aimed in the present study to explore the association between the use of psychotropics, multiple other medications, frailty-related parameters and gait performance in older patients. Eighty older persons (79±5.6 years) were recruited. Comorbid diseases, frailty-related parameters, and medication-use were registered. Trunk accelerations during a 3-minute-walking-task were recorded, whereof walking speed, mean stride times, coefficient of variation (CV) of stride times, and step consistency were determined. Multivariate Partial Least Squares (PLS) regression analysis was used to examine the association between population characteristics and medication-use, versus gait parameters. A PLS-model existing of four latent variables was built, explaining 45% of the variance in four gait parameters. Frailty-related factors, being female, and laxative-use were most strongly associated with lower walking speed, higher mean stride times, higher CV of stride times, and less consistent steps. In conclusion, frailty-related parameters were stronger associated with impaired gait performance than the use of psychotropic drugs. Possibly, at a certain frailty-level, the effect of the deterioration in physical functioning in frailty is so large, that the instability-provoking side-effects of psychotropic drugs have less impact on gait.
Citation: de Groot MH, van Campen JPCM, Kosse NM, de Vries OJ, Beijnen JH, Lamoth CJC (2016) The Association of Medication-Use and Frailty-Related Factors with Gait Performance in Older Patients. PLoS ONE 11(2): e0149888. https://doi.org/10.1371/journal.pone.0149888
Editor: Walter Maetzler, University of Tuebingen, GERMANY
Received: October 25, 2015; Accepted: February 6, 2016; Published: February 22, 2016
Copyright: © 2016 de Groot et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The authors have no support or funding to report.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Older persons visiting a geriatric outpatient clinic are characterized by a combination of physical, psychological and social problems. They often have multiple chronic conditions, and use therefore multiple medications (polypharmacy). These frail old people have a high risk of adverse events, such as falling, hospital admissions, and ultimately death [1,2].
The past decades, many risk factors for falls in elderly people have been identified, which can be classified as either intrinsic (e.g., age-related changes in the sensor-motor system leading to gait and balance deficits), extrinsic (e.g., polypharmacy), or environmental factors [3]. In addition, meta-analyses showed that specific medication classes, including psychotropics and some cardiac drugs, have been associated with an increased risk of falling [4–7]. In a recent literature review [8], we suggested that the increased fall risk with which these psychotropic drug classes are associated, is caused by underlying problems in postural control these drugs can induce. That is because postural instability during daily activities, such as walking, is a primary cause of falling [9]. Walking can be characterized by a wide variety of gait parameters, each characterizing different aspects of the gait pattern. Walking speed is easily assessable and often used in clinical practice to reflect the functional status in older patients [10]. However, it lacks specificity (gait speed is slower in many different pathologies), and does not reveal information about the quality of the walking pattern, i.e., the ability of a person to adapt his/her gait smoothly, and the variability and regularity of the gait pattern. Stride-to-stride variability during walking provides additional information about gait performance, since high variability in stride times in frail elderly can be considered as a marker of abnormal regulation of gait [11,12]. Both walking speed and characteristics indicating the variability in the walking pattern are suggested to be associated with increased fall risk [13,14].
The current literature about the effects of psychotropic drugs on postural control assessed postural control during quiet standing, described only acute single-drug effects, and included relatively healthy young elderly [8]. Thus, the impact of long-term psychotropic drug-use on gait performance in frail older people with multiple comorbid diseases and polypharmacy remains still unclear.
In the present study, we therefore aimed to explore the association between the use of psychotropic drugs and gait performance in a cohort of older patients. Because in this frail older population many factors may be simultaneously present that could influence gait (e.g., the use of multiple medications, the presence of comorbid diseases, cognitive impairment, and/or frailty [11,12,15,16]), we included these factors in our analyses as well. The use of a cross-sectional study design, and a Partial Least Squares (PLS) regression analysis, enabled us to examine the association between the use of psychotropic drugs and gait parameters, in relation to other factors that might influence the walking pattern. We hypothesized that a deterioration in gait performance was not only explained by the use of psychotropics, but that in older patients frailty-related parameters, polypharmacy and comorbidity would also be associated.
Methods
Ethics Statement
The study was approved by the local Medical Ethics Committee of the MC Slotervaart (registration number: NL33825.048.10). All participants or their legal representatives (when participants had cognitive impairment) gave their written informed consent.
Participants
Eighty patients were recruited consecutively among the elderly that visited the geriatric outpatient clinic of the MC Slotervaart in Amsterdam between October 2010 and April 2013. They were referred to the clinic by their general practitioner for various reasons, including memory complaints, mobility problems, or for evaluating polypharmacy. Patients were eligible to participate in the study if they were at least 65 years or older, and could walk safely for at least 3-minutes without any assistive device. They were excluded when they had any mobility problems due to neurological or orthopedic disorders, or were not able to understand the instructions of the researcher due to severe cognitive or hearing impairment.
During the study period, 619 older adults visited the geriatric outpatient clinic. Thereof, 392 patients (63%) did not meet the in- and exclusion criteria. In total, 227 (37%) patients were eligible, whereof 80 patients (13% of all visitors of the geriatric outpatient clinic) were willing to cooperate in the present study. The other 147 patients refused to cooperate because they felt too tired after the day in the hospital or did not want to cooperate (without giving reason).
Measurements
Population characteristics.
All participants underwent extensive physical and cognitive examination, as part of a Comprehensive Geriatric Assessment (CGA) [17]. General information was obtained from their patient file. All population characteristics were binary coded for statistical analyses (see S1 Dataset). Age was coded as being ≥80 or <80 years, because this was the mean age of the included participants). Cognitive functioning was assessed using the Mini Mental State Examination (MMSE) [18]. Participants scoring ≤23 points were categorized as being cognitive impaired [19]. Fall risk was examined using the LASA fall risk assessment [20]. When participants scored ≥8 points, they were identified as having an increased risk of recurrent falling. Four frailty criteria were registered [1], namely: unintentional weight loss, self-reported exhaustion, low physical activity, and low hand grip strength. Walking speed was a dependent variable in the statistical analyses. Comorbid diseases were registered using the Charlson’s Comorbidity Index (CCI) [21]. Table 1 lists the registered comorbid diseases.
In the right column, the captured variance (%) in every independent variable per latent variable (LV) of the PLS-model, and the total variance is presented. Values with high modeling power (>6.90%)a are relevant to the LVs.
Medication-use.
Medication-use was systematically reviewed by the participant’s physician as part of the CGA. Prior to their visit to the Geriatric Outpatient Clinic, patients were requested to bring the packaging of their medications or a list of their pharmacist with the drugs they use. The physician went through the list with the patients and/or their caregivers to check if the patients used all medications appropriately and/or used other non-prescribed drugs or supplements. The actual medication-use was then registered in the patients’ file, which was used in the present study.
All generic names of the drugs (trade names) were manually looked up and subsequently linked with the corresponding code of Anatomical Therapeutic Chemical (ATC) classification system [23] (see S1 Table for an overview of the prevalence of all drugs in the study population). The medications were clustered into ATC-drug classes (see Table 1 for an overview of the 37 included ATC-drug classes in the present study). For each participant it was binary scored whether the participant used a drug in a specific ATC-drug class or not (see S1 Dataset). Furthermore, the total number of drugs used per participant was registered. Polypharmacy was present when the participant used ≥4 drugs.
Gait parameters.
Participants walked 3-minutes in a 80-meter long hallway at a self-selected speed while trunk accelerations were recorded using an accelerometer (DynaPort Minimod Hybrid; McRoberts BV; sample frequency 100 Hz), which was attached with a band at the level of the lumbar spine. Walking distance was recorded to determine gait speed. Trunk acceleration signals were analyzed using custom-made software in MATLAB (version 2011b; The MathWorks Inc.). Foot contact moments were determined from the peaks of the anterior-posterior-acceleration time-series. From the foot contact moments, stride times were calculated defined as the time interval between two ipsilateral foot contacts. For each participant, mean and CV of stride times were computed. The variability between left and right steps within the strides was quantified by the standard deviation (SD) of the relative timing between ipsilateral foot contact moments, with larger values indicating less consistent steps [24,25]. See De Groot et al. [16] for a detailed description of the calculated gait parameters.
Statistical Analyses
To test the association between the various drug classes, the population characteristics, and the comorbid diseases in relation to stride variability, a multivariate PLS regression analysis was computed using the PLS_toolbox for Matlab (version 3.7.1; Eigenvector Research Inc.). PLS-regression is a technique that combines features from principal component analysis and multiple regression, and is not impeded by collinearity among variables [26–28]. PLS has particularly been applied in studies in which a set of dependent variables is predicted from a relatively large set of independent variables with relatively few observations [29–31]. In these studies, similar to our study, continuous, ordinal and binary data were included. It is well known that their exists an interdependency among different gait variables, such as walking speed, mean stride times, and variability in stride times [32–34]. When using PLS, a model can be constructed wherein the dependency among these gait variables is taken into account. That makes the PLS-regression analysis more favorable over the more conventional multiple regression analysis.
In the PLS-regression analysis, 58 independent variables (the population characteristics, comorbid diseases, and medication-use) and four dependent gait parameters were included. In order to give the variables the same scale, i.e., the same importance in the analysis, all independent (binary coded) variables were mean-centered (i.e., means of each column were set to zero), and the dependent gait parameters were scaled to unit variance by dividing each variable by their SD’s and centering them by subtracting their averages [22,35].
PLS-regression analysis searches for a set of latent variables (LVs) explaining as much as possible of the covariance between the independent (the population characteristics, comorbid diseases, and medication-use per ATC-class), and dependent variables (the four gait parameters). With numerous and correlated variables, there is a substantial risk for “over-fitting”, i.e., getting a well-fitting model with little or no predictive power. Cross-validation is a practical and reliable way to test the predictive significance of the model [35]. Cross-validation was performed by dividing the data into eight groups (method: Venetian Blinds), and then developing eight parallel models from reduced data with one of the groups deleted. Then, differences between the actual and the predicted values for the gait parameters were calculated for the deleted data. The sums of squares of these differences were computed and collected from all the parallel models to form the predictive residual sum of squares (PRESS), which estimates the predictive ability of the model. The optimal number of LVs was determined by stop adding LVs as soon as the Predicted REsidual Sum of Squares (PRESS) decreased [36]. From the PRESS-values of the final model, the predictive ability of the model was indicted by the parameter Q2, the predicted variation, or: the goodness of prediction. This value can be compared to R2, the explained variation, which is the capacity of the population characteristics, comorbid diseases and medication-use to explain the variance among the gait variables, or: the goodness of fit.
From the PLS model, a set of model parameters was computed (see Table 2). A downside of using PLS is that it is relatively difficult to interpret model parameters, since these are biased and p-values and confidence intervals are lacking. By interpreting the model parameters simultaneously, the associations between the independent variables (the population characteristics, comorbid diseases, and medication-use) were examined in relation to the gait parameters. The relevance of each independent variable to the LVs was indicated by the percentage variance that was captured by the LV [22]. Variables with low modeling power, i.e., around A/K (in our study, A = number of LVs = 4; and K = number of independent variables = 58), were of little relevance. Thus variables with variance below 4/58 = 6.90% are not important to the LV [22,29]. Therefore, variables with high captured variance for the same LV were clustered in the same LV.
The association between each independent variable and the gait parameters was indicated by regression coefficients. A higher positive or negative regression coefficient means that the independent variable is stronger related to the gait parameter. The importance of each variable in explaining the variation among gait parameters was given by the variable importance in projection (VIP) value. A VIP-value higher than 1 means that the independent variable is influential to the gait parameters. At last, score and weight plots illustrate and summarize the relationship between observations (the participants) and independent variables (population characteristics, comorbid diseases and medication-use), respectively, with respect to the latent variables.
Since the model parameters are biased, the important variables in relation to the gait parameters have overall higher modeling power (high captured variance), have higher positive or lower negative regression coefficients, and have a VIP-value higher than one. Moreover, these important variables have higher positive or lower negative weight-values, and are therefore clustered in the upper right or lower left quadrant of the weight plot. Based on the scores and weights of the model, predicted values for the gait parameters were calculated to illustrate the fit of the model [22,28,36]. See for a more detailed mathematical description of the PLS regression analysis the Supporting Information (S1 Appendix).
Results
Descriptive Characteristics of the Study Population
Eighty older persons were included in the present study (aged 79±5.6 years). Cognitive impairment was present in 35 participants (44%); the median MMSE score was 24 points (range: 11–30). Fourteen participants (18%) scored ≥8 points on the LASA fall risk assessment, indicating an increased risk for recurrent falls. Considering the frailty-related indicators, 13 participants (16%) reported ≥5kg unintentional weight loss, 21 participants (26%) had self-reported exhaustion, 12 participants (15%) had low physical activity, and 20 participants (25%) had low grip strength. Twenty-nine participants (36%) had ≥2 comorbid diseases on the CCI. Dementia, myocardial infarct, and diabetes were most present in the included population (respectively 27%, 26%, and 14%). See Table 1 for the prevalence of other comorbid diseases.
The participants used 5.5±3.9 drugs; 52 participants (65%) were categorized as having polypharmacy, because they used ≥4 drugs. Twenty-nine participants (36%) used one or more psychotropic medications that are known to be associated with increased fall risk, e.g. antiepileptics (N03A), antipsychotics (N05A), anxiolytics (N05BA), hypnotics (N05C), and antidepressants (N06A). Medication-use per ATC-class is presented in Table 1.
Mean walking speed was 0.90±0.24 m/s, mean stride time was 1.16±0.13 seconds, the variation (CV) of stride times was 3.81±1.61%, and step consistency was 4.67±1.68°.
Association Between Population Characteristics, Comorbid Diseases, and Medication-Use in Relation to Gait Parameters
The PLS-regression testing the association between population characteristics, comorbid diseases and medication-use in relation to gait parameters yielded a model of four LVs (see Table 3), explaining 27.6% of the variance between the population characteristics, comorbid diseases, and medication-use, and 45.4% of the variance in gait parameters (R2). Adding more LVs to the model did not improve R2, as indicated by the PRESS.
The variable most relevant to the model, i.e., with the most variance captured by the model, was the presence of polypharmacy (63%). The first LV captured most variance of the following variables: using drugs for ulcer and reflux disease (A02B; 39%), polypharmacy (36%), and two frailty criteria: low physical activity (25%), and self-reported exhaustion (25%). The second LV captured most variance in polypharmacy (20%), and in using drugs acting on the cardiovascular system (ATC-drug classes C09, C10, C03, C07, and C01D). Most variance by the third LV was captured in being female (25%), having low hand grip strength (24%), and increased fall risk (21%). By the fourth LV, most variance was captured in having myocardial infarction in medical history (19%), and using agents acting on the renin-angiotensin system (C09; 14%), and beta-blocking agents (C07; 10%). The variance captured by the model for all variables per LV are presented in Table 1.
In Fig 1, the VIP-values and regression coefficients of all population characteristics, comorbid diseases and medication-use for the gait parameters are presented. The most important variables to the model (highest VIP-values) were unintentional weight loss, self-reported exhaustion, being female, and the use of drugs for constipation (A06). These variables were associated with lower walking speed (negative regression coefficients), and higher mean and CV of stride times and less consistent steps (positive regression coefficients).
VIP-values (gray bars; left Y-axis) and regression coefficients (black dots (●); right Y-axis) of all population characteristics, comorbid diseases and medications used are presented for (A) walking speed, (B) mean stride times, (C) CV of stride times, and (D) step consistency. The variables are placed on the horizontal axis and sorted according to the height of the regression coefficient. Variables placed at the outer left and right side of the graph have a stronger association with the gait parameter than the variables in the middle of the graph. Variables with a VIP-value of >1 are important to the model. Note that right of the vertical dotted line the variables are presented that are associated with impaired gait ability, that is negative regression coefficients for walking speed, and positive regression coefficients for mean stride times, CV of stride times and step consistency. See Table 1 for a description of the ATC-drug classes.
The association between the participants and the independent variables (population characteristics, comorbid diseases, and medication-use) in relation to the gait parameters is revealed in Fig 2. The position of the participants in a given direction in the score plot (Fig 2A) is influenced by the independent variables lying in the same direction in the weight plot (Fig 2B). Participants in the upper right quadrant of the score plot had overall higher stride variability, whereas persons clustered in the lower left quadrant had lower stride variability. Furthermore, in the weight plot, the relations between the independent variables are visualized. Variables with high regression coefficients and high VIP-values, e.g., unintentional weight loss, self-reported exhaustion, and being female (see Fig 1), are situated in the upper right quadrant of the weight plot, illustrating that they are associated with each other (as can be also seen by the captured variance of LV1 as presented in Table 1), and are associated with higher values for the gait parameters (higher regression coefficients; see Fig 1).
The score-plot (A) shows the relationship between the participants. The participants are categorized according to their CV of stride times: quartile 1 (Q1; lowest variability; < 2.6%), quartile 2 (Q2; 2.7–3.4%), quartile 3 (Q3; 3.4–4.4%), and quartile 4 (highest variability in stride times; >4.4%). Coupled to the weight plot (B), the inter-relatedness among the patient characteristics and stride variability is revealed. The weight plot can be considered as a coordinate system, where the underlying structure between the variables in relation to stride variability is revealed. The most influential variables are situated far from origo, with the variables at upper right quadrant associated with higher stride variability, and those situated in the lower left quadrant are associated with lower variability.
The fit and predictive ability of the model are illustrated in Fig 3. The observed versus predicted values of the PLS model for walking speed (R2 = 60%; Q2 = 92%), mean stride times (R2 = 56%; Q2 = 98%), CV of stride times (R2 = 39%; Q2 = 88%), and for step consistency (R2 = 26%; Q2 = 87%) are presented in this figure.
For (A) walking speed, (B) mean stride times, (C) CV of stride times, and (D) step consistency the observed and predicted values are presented. The striped line represents the fit-line, and the dotted line is the 1:1-line.
In summary, the results show that impairments in gait performance are strongest associated with unintentional weight loss, self-reported exhaustion, being female and the use of laxatives (A06). These variables had overall most variance captured by the model, and the highest VIP-values and regression coefficients.
Discussion
In the present cross-sectional study, we aimed to determine the association between population characteristics, comorbid diseases, medication-use and gait parameters in a population of older patients visiting a geriatric outpatient clinic. Therefore, a PLS-regression model was created testing the association between the use of medications in various ATC-drug classes, comorbid diseases, and common factors in a geriatric population, versus gait parameters. In total, 58 variables were included, explaining 45% of the total variance in four gait parameters, namely walking speed, mean stride times, variation in stride times, and step consistency. R2- and Q2-values showed that the model had an appropriate fit and a good predictive ability (see Fig 3).
We created a statistical model including characteristics common in a population of older patients visiting a geriatric outpatient clinic. 45% of the variance in four gait parameters was explained by this model. Gait is a complex motor behavior requiring adequate integration of sensor-motor information. With aging, there is an age-related deterioration of postural control due to progressive loss of functioning of the neurophysiological system, resulting in, amongst others, visual problems, vestibular impairment, affected proprioception, changes in central processing mechanisms, reduced muscle strength, and slower reaction times [37,38]. These factors are also present in our geriatric study population, but were not as such included in our model. Therefore, these not included aging-related factors presumably account for the unexplained variance in gait parameters in our model.
Nevertheless, most variance in the gait variables was explained by the first LV, and the clustered variables in this LV were therefore strongest related with the gait parameters. In particular, frailty-related criteria were clustered in this first LV, such as unintended weight loss, self-reported exhaustion, low physical activity, and low hand grip strength (see Table 1 and Fig 2), and these were associated with lower walking speed, and higher mean stride times, increased variability of stride times, and less consistent steps. The identified association between these frailty-related indicators and gait parameters is in line with previous research [15,39]. Montero-Odasso et al. [15] suggested that frailty reflects a general deterioration in the neurophysiological system. At the same time, gait can be considered as a complex task that is highly regulated by neurophysiological systems. Consequently, the decreased walking speed, increased stride variability, and worse step consistency that was seen in frail elderly and was found to be related with frailty-related parameters as found in the present study, may reflect the multisystem reduction in neurophysiological capacity.
Based on a previous literature review [8], we anticipated a strong association between the use of psychotropic drugs and gait performance. However, the association of the use of antiepileptics (N03A) and antidepressants (N06A) with impaired gait ability was moderate, while hardly any association with other psychotropic drug classes was observed. We suggest that the effect of the deterioration in physical functioning in the frailest elderly is so large, that the added impact of the side-effects of psychotropics on gait regulation is relatively small. What also may have contributed to the weak associations between gait performance and psychotropic drugs is that the dosages of drugs that were taken by our study population (see S1 Table) was much lower than reported in the aforementioned literature review [8]. This underpins the recommendation of Lader et al. [40] that psychotropic drugs—when administered—should only be prescribed at the lowest effective dose and for a restricted period of time to limit the side-effects of these drugs.
Beside the strong association between frailty-related characteristics and gait parameters, and the moderate association of psychotropic drugs with gait, in the present study also other medication classes were found to be associated with impaired gait ability. The use of laxatives (A06), agents acting on sensory organs (ATC-class S), and drugs for peptic ulcer and reflux disease (A02B) were strongly associated with lower walking speed and higher stride variability as was indicated by high VIP-values and regression coefficients (Fig 1), and illustrated in the score and weight plots (Fig 2). These medication-classes are among the most prescribed medications, and are much more commonly used in individuals with indicators of frailty. For instance, laxatives are used for constipation that, in turn, may be caused by inactivity (one of the indicators of frailty) or the use of constipation-associated medications (e.g., opioid analgesics) [41]. Therefore, these kind of drugs may act as frailty markers [42,43].
The results of the present study correspond with a study of Askari et al. [43] who reported that besides known “fall-risk-increasing drugs”, such as analgesics (N02), anti-Parkinson drugs (N04) and psychoanaleptics (N06), also relatively new classes showed significant association with recurrent falls in elderly, namely nasal preparations (R01), ophthalmologicals (S01), and drugs for acid-related disorders (A02). Especially the finding that drugs for acid-related disorders were associated with recurrent falling and increased stride variability is interesting, because these drugs are commonly used in our study population (see S1 Table) and in the older population in general [44]. Proton-pump inhibitors (PPIs; A02BC), a subclass of drugs for acid-related disorders (A02), have been associated with an increased risk of fractures [45]. It is assumed that PPIs reduce calcium absorption, possibly leading to reduced bone mineral density that would lead to increased fracture risk [46]. Another explanation could be that PPI-use may lead to muscle weakness and gait disorders [46]. This might explain our finding that these drugs were associated with impaired gait ability, and with recurrent falling as Askari et al. [43] found. However, further longitudinal research is necessary to confirm this suggestion.
Conclusion
In the present study, we found an association between frailty-related parameters and impaired walking ability, as was expressed by lower walking speed and increased stride variability. The absence of a strong association between the use of psychotropic drugs and gait parameters is in accordance to our hypothesis that results from studies in healthy and relatively young elderly cannot be extrapolated to a frail population, because comorbid diseases and the use of many medications are complicating factors in this population leading to multisystem reduction in the neurophysiological system. Therefore, in future studies among frail elderly we recommend to develop models which include frailty-related parameters.
Supporting Information
S1 Appendix. Detailed mathematical description of the PLS regression analysis.
https://doi.org/10.1371/journal.pone.0149888.s001
(PDF)
S1 Dataset. Dataset containing the prevalence of the population characteristics, comorbid diseases, and medication-use per ATC-drug class (as listed in Table 1), and the raw values of the gait parameters.
https://doi.org/10.1371/journal.pone.0149888.s002
(XLS)
S1 Table. Prevalence of the medications used by the study’s participants and the dosage range.
https://doi.org/10.1371/journal.pone.0149888.s003
(PDF)
Acknowledgments
The authors received no specific funding for this work. The authors have declared that no competing interests exist.
Author Contributions
Conceived and designed the experiments: MHdG JPCMvC CJCL. Performed the experiments: MHdG. Analyzed the data: MHdG NMK CJCL. Contributed reagents/materials/analysis tools: MHdG JPCMvC NMK CJCL. Wrote the paper: MHdG JPCMvC NMK OJdV JHB CJCL.
References
- 1. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci. 2001;56: M146–M156. pmid:11253156
- 2. de Vries OJ, Peeters GMEE, Lips P, Deeg DJH. Does frailty predict increased risk of falls and fractures? A prospective population-based study. Osteoporos Int. 2013;24: 2397–2403. pmid:23430104
- 3. American Geriatrics Society, British Geriatrics Society, American Academy of Orthopaedic Surgeons Panel on Falls Prevention. Guideline for the prevention of falls in older persons. J Am Geriatr Soc. 2001;49: 664–672. pmid:11380764
- 4. Ham AC, Swart KMA, Enneman AW, van Dijk SC, Oliai Araghi S, van Wijngaarden JP, et al. Medication-related fall incidents in an older, ambulant population: The B-PROOF study. Drugs Aging. 2014;31: 917–927. pmid:25398701
- 5. Bloch F, Thibaud M, Dugué B, Brèque C, Rigaud A-S, Kemoun G. Psychotropic drugs and falls in the elderly people: updated literature review and meta-analysis. J Aging Health. 2011;23: 329–346. pmid:20947876
- 6. Woolcott JC, Richardson KJ, Wiens MO, Patel B, Marin J, Khan KM, et al. Meta-analysis of the impact of 9 medication classes on falls in elderly persons. Arch Intern Med. 2009;169: 1952–1960. pmid:19933955
- 7. Hartikainen S, Lönnroos E, Louhivuori K. Medication as a risk factor for falls: critical systematic review. J Gerontol A Biol Sci Med Sci. 2007;62: 1172–1181. pmid:17921433
- 8. de Groot MH, van Campen JPCM, Moek MA, Tulner LR, Beijnen JH, Lamoth CJC. The effects of fall-risk-increasing drugs on postural control: a literature review. Drugs Aging. 2013;30: 901–920. pmid:24005984
- 9. Ganz DA, Bao Y, Shekelle PG, Rubenstein LZ. Will my patient fall? JAMA. 2007;297: 77–86. pmid:17200478
- 10. Abellan van Kan G, Rolland Y, Andrieu S, Bauer JM, Beauchet O, Bonnefoy M, et al. Gait speed at usual pace as a predictor of adverse outcomes in community-dwelling older people. An International Academy on Nutrition and Aging (IANA) task force. J Nutr Heal Aging. 2009;13: 881–889.
- 11. Lamoth CJC, van Deudekom FJ, van Campen JPCM, Appels BA, de Vries OJ, Pijnappels M. Gait stability and variability measures show effects of impaired cognition and dual tasking in frail people. J Neuroeng Rehabil. 2011;8: 2. pmid:21241487
- 12. IJmker T, Lamoth CJC. Gait and cognition: the relationship between gait stability and variability with executive function in persons with and without dementia. Gait Posture. 2012;35: 126–130. pmid:21964053
- 13. Hamacher D, Singh NB, Van Dieen JH, Heller MO, Taylor WR. Kinematic measures for assessing gait stability in elderly individuals: a systematic review. J R Soc Interface. 2011;8: 1682–1698. pmid:21880615
- 14. Menant JC, Schoene D, Lord SR. Single and dual task tests of gait speed are equivalent in the prediction of falls in older people: A systematic review and meta-analysis. Ageing Res Rev. 2014;
- 15. Montero-Odasso M, Muir SW, Hall M, Doherty TJ, Kloseck M, Beauchet O, et al. Gait variability is associated with frailty in community-dwelling older adults. J Gerontol A Biol Sci Med Sci. 2011;66A: 568–576.
- 16. de Groot MH, van der Jagt-Willems HC, van Campen JPCM, Lems WF, Beijnen JH, Lamoth CJC. A flexed posture in elderly patients is associated with impairments in postural control during walking. Gait Posture. 2014;39: 767–772. pmid:24268470
- 17. Ellis G, Langhorne P. Comprehensive geriatric assessment for older hospital patients. Br Med Bull. 2004;71: 45–59. pmid:15684245
- 18. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12: 189–198. pmid:1202204
- 19. Tombaugh TN, McIntyre NJ. The Mini-Mental State Examination: A comprehensive review. J Am Geriatr Soc. 1992;40: 922–935. pmid:1512391
- 20. Pluijm SMF, Smit JH, Tromp EAM, Stel VS, Deeg DJH, Bouter LM, et al. A risk profile for identifying community-dwelling elderly with a high risk of recurrent falling: results of a 3-year prospective study. Osteoporos Int. 2006;17: 417–425. pmid:16416256
- 21. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40: 373–383. pmid:3558716
- 22.
Eriksson L, Johansson E, Kettaneh-Wold N, Trygg J, Wikström C, Wold S. Multi- and megavariate data analysis. Part I—Basic principles and applications. Umea, Sweden: Umetrics; 2006.
- 23.
WHO Collaborating Centre for Drug Statistics Methodology. Guidelines for ATC classification and DDD assignment 2013. Oslo, Norway: Norwegian Institute of Public Health; 2012.
- 24. Roerdink M, Lamoth CJC, Kwakkel G, van Wieringen PCW, Beek PJ. Gait coordination after stroke: benefits of acoustically paced treadmill walking. Phys Ther. 2007;87: 1009–1022. pmid:17553922
- 25. Plotnik M, Giladi N, Hausdorff JM. A new measure for quantifying the bilateral coordination of human gait: effects of aging and Parkinson’s disease. Exp Brain Res. 2007;181: 561–570. pmid:17503027
- 26.
Wold H. Estimation of principal components and related models by iterative least squares. In: Krishnaiah PR, editor. Multivariate Analysis. New York, NY, USA: Academic Press; 1966. pp. 391–420.
- 27. Haenlein M, Kaplan AM. A beginner’s guide to Partial Least Squares Analysis. Underst Stat. 2004;3: 283–297.
- 28. Boulesteix A-L, Strimmer K. Partial least squares: a versatile tool for the analysis of high-dimensional genomic data. Brief Bioinform. 2007;8: 32–44. pmid:16772269
- 29. Kosse NM, de Groot MH, Vuillerme N, Hortobágyi T, Lamoth CJC. Factors related to the high fall rate in long-term care residents with dementia. Int Psychogeriatrics. 2015;27: 803–814.
- 30. Marzetti E, Lees HA, Manini TM, Buford TW, Aranda JM, Calvani R, et al. Skeletal muscle apoptotic signaling predicts thigh muscle volume and gait speed in community-dwelling older persons: an exploratory study. PLoS ONE. 2012;7: e32829. pmid:22389725
- 31. Eriksson S, Gustafson Y, Lundin-Olsson L. Risk factors for falls in people with and without a diagnose of dementia living in residential care facilities: a prospective study. Arch Gerontol Geriatr. 2008;46: 293–306. pmid:17602762
- 32. Beauchet O, Annweiler C, Lecordroch Y, Allali G, Dubost V, Herrmann FR, et al. Walking speed-related changes in stride time variability: effects of decreased speed. J Neuroeng Rehabil. 2009;6: 32. pmid:19656364
- 33. Kang HG, Dingwell JB. Separating the effects of age and walking speed on gait variability. Gait Posture. 2008;27: 572–577. pmid:17768055
- 34. Allali G, Ayers EI, Verghese J. Multiple modes of assessment of gait are better than one to predict incident falls. Arch Gerontol Geriatr. 2015;60: 389–393. pmid:25772422
- 35. Wold S, Sjöström M, Eriksson L. PLS-regression: a basic tool of chemometrics. Chemom Intell Lab Syst. 2001;58: 109–30.
- 36. Abdi H. Partial least squares regression and projection on latent structure regression (PLS regression). WIREs Comp Stat. 2010;2: 97–106.
- 37. Sturnieks DL, St George R, Lord SR. Balance disorders in the elderly. Clin Neurophysiol. 2008;38: 467–478.
- 38. Papegaaij S, Taube W, Baudry S, Otten E, Hortobágyi T. Aging causes a reorganization of cortical and spinal control of posture. Front Aging Neurosci. 2014;6: 28. pmid:24624082
- 39. Schwenk M, Howe C, Saleh A, Mohler J, Grewal G, Armstrong D, et al. Frailty and technology: a systematic review of gait analysis in those with frailty. Gerontology. 2014;60: 79–89. pmid:23949441
- 40. Lader M, Tylee A, Donoghue J. Withdrawing benzodiazepines in primary care. CNS Drugs. 2009;23: 19–34. pmid:19062773
- 41. Wald A. Constipation in elderly patients. Pathogenesis and management. Drugs Aging. 1993;3: 220–231. pmid:8324297
- 42. Teramura-Grönblad M, Hosia-Randell H, Muurinen S, Pitkala K. Use of proton-pump inhibitors and their associated risks among frail elderly nursing home residents. Scand J Prim Health Care. 2010;28: 154–159. pmid:20586539
- 43. Askari M, Eslami S, Scheffer AC, Medlock S, de Rooij SE, van der Velde N, et al. Different risk-increasing drugs in recurrent versus single fallers: are recurrent fallers a distinct population? Drugs Aging. 2013;30: 845–851. pmid:23959914
- 44. Mouli VP, Ahuja V. Proton pump inhibitors: concerns over prolonged use. Trop Gastroenterol. 2011;32: 175–84. pmid:22332332
- 45. Kwok CS, Yeong JK-Y, Loke YK. Meta-analysis: risk of fractures with acid-suppressing medication. Bone. 2011;48: 768–776. pmid:21185417
- 46. Lewis JR, Barre D, Zhu K, Ivey KL, Lim EEM, Hughes J, et al. Long-term proton-pump inhibitor therapy and falls and fractures in elderly women: a prospective cohort study. J Bone Miner Res. 2014;29: 2489–2497. pmid:24825180