Abstract
Background:
Predictors of falls in people with Parkinson’s disease (PD) who have not previously fallen are yet to be identified.
Objectives:
We aimed to identify predictors of all falls and recurrent falls in people with PD who had not fallen in the previous year and to explore the timing of falls in a 12-month follow-up period.
Methods:
Participants with PD (n = 130) were assessed by disease-specific, self-report and balance measures. Falls were recorded prospectively for 12 months. Univariate and multivariate analyses were performed. Kaplan-Meier survival analysis was used to investigate time to falling.
Results:
Forty participants (31%) had ≥1 fall during follow-up and 21 (16%) had ≥2 falls. Disability, reduced balance confidence and greater concern about falling were associated with ≥1 fall in univariate analyses. Additionally, PD duration and severity, freezing of gait and impaired balance were associated with ≥2 falls (p < 0.05). Disability (Schwab and England scale, Odds Ratio [OR] = 0.56 per 10 points increase; 95% confidence interval [CI] 0.39–0.80; p = 0.002) was associated with ≥1 fall in the final multivariate model (area under the receiver operating characteristic curve [AUC] = 0.65; 95% CI 0.55–0.76; p = 0.005). Disability (Unified Parkinson’s Disease Rating Scale activities of daily living, OR = 1.20; 95% CI 1.07–1.34; p = 0.001) and levodopa equivalent dose (OR = 1.11 per 100 mg increase; 95% CI 0.95–1.30; p = 0.19) were associated with ≥2 falls in the final multivariate model (AUC = 0.72; 95% CI 0.60–0.84; p = 0.001). Recurrent fallers experienced their first fall earlier than single fallers (p < 0.05).
Conclusions:
Self-reported disability was the strongest single predictor of all falls and recurrent falls.
INTRODUCTION
The incidence of falls is high in individuals with Parkinson’s disease (PD), with reported rates ranging from 35% to 90% for at least one fall and from 18% to 65% for recurrent falls [1]. These falls result inserious consequences, such as fractures, fear of falling and increased caregiver burden [1]. Several risk factors for falling have been identified, including falling in the last year [2–5], disease severity [2–4], freezing of gait (FOG) [4–7], poor balance [4, 7], impaired mobility [6], fear of falling [8] and cognitive impairment [4, 7].
Understanding falls in people who have not recently fallen may give insight into why falls occur and guide falls prevention programs. However, few prospective studies have focused on people with PD without a history of falls [9–12]. These studies have reported a relationship between falls and impaired cognition [9], disease severity [3, 10], increased gait variability [11] and poor health-related quality of life [12]. Two studies have evaluated the predictive accuracy of risk factors, but in one the predictive accuracy was no better than chance [3] and the other reported a multivariate model including variables which were chosen based on their association with falling in the entire sample rather than in the subsample of individuals without previous falls [6]. Although providing valuable information about important risk factors for falling, these studies did not consider measures of balance, self-efficacy and disability concurrently.
There is increasing evidence supporting differences between people with PD who experience a single fall and those who are recurrent fallers in terms of balance confidence, gait measures and disability [13, 14]. However, predictors of falls and recurrent falls are yet to be investigated among people with PD without a history of falls.
We undertook this study to address the lack of information about risk factors for falls in people with PD who have not previously fallen, based on commonly-used clinical measures. The aims of this study were to: 1) determine predictors of all falls in people with PD who had not fallen in the past year; 2) determine predictors of recurrent falls in people with PD who had not fallen in the past year; and 3) explore the timing of falls for single and recurrent fallers. We hypothesized that a broader range of predictors would be associated with recurrent falls and that the first fall would occur earlier in recurrent fallers.
MATERIALS AND METHODS
Participants
Consecutive participants were recruited from the Movement Disorders Clinic at the State of Bahia Health Attention Center for the Elderly (CREASI), Brazil, between April 2010 and June 2013. The follow-up period lasted until June 2014. Eligibility criteria included a diagnosis of idiopathic PD [15], ability to walk without assistance of another person, with or without an assistive device, and no history of falls in the last 12 months. This subsample of people with PD who had not previously fallen was derived from a large prospective cohort study investigating predictors of recurrent falls which will be reported separately. Participants were excluded if they had neurological conditions other than PD or comorbidities that would affect locomotion or balance. In addition, participants with significant cognitive impairment were excluded based on Mini-Mental State Examination (MMSE) scores with cut-offs adjusted to education level (no education, 13; <8 years of education, 18; ≥8 years of education, 26) [16]. The study was approved by a local research ethics committee (CEP-SESAB, 042/2010). All participants provided written informed consent prior to assessment and were tested during the “on” phase of the medication cycle.
Baseline assessments
Demographic and clinical data including age, gender, comorbidities, PD duration, motor fluctuations, dyskinesia and medications taken, including levodopa equivalent dose (LED) [17] were recorded. Participants were classified as receiving polypharmacy when they were taking ≥4 medications other than anti-parkinsonian drugs [4]. In order to analyze cognitive impairment MMSE scores were adjusted according to level of education and capped at a maximum score of 30 points.
PD severity was ascertained from the motor section of the Unified Parkinson’s Disease Rating Scale (UPDRS) [18] and the modified Hoehn and Yahr stage (H&Y) [19]. Self-reported disability was ascertained by the UPDRS activities of daily living (ADL) section [18] and Schwab and England scale (S&E) [20]. PD motor subtypes [21] and the postural instability and gait difficulty (PIGD) score (sum of UPDRS items 13, 14, 15, 29 and 30) were determined. Participants were classified as having FOG if they scored ≥1 point on item 14 of the UPDRS ADL. Self-efficacy measures included the Activities-specific Balance Confidence scale (ABC) and the Falls Efficacy Scale-International (FES-I) [22]. The following balance measures were used: Berg balance scale (BBS), Functional reach test (FRT), Timed up and go (TUG) and Dynamic gait index (DGI) [23]. Health-related quality of life was measured by the 8-item Parkinson’s Disease Questionnaire (PDQ-8) [24].
Procedures
All assessments were performed on the same day as participants were recruited. The order was standardized for all participants and the assessment session took approximately 60 minutes. Rest breaks were given if 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 other level, not as the result of a major intrinsic event or overwhelming hazard” [25]. Following baseline assessment, participants were followed-up for 12 months during which they completed a falls diary and received monthly telephone calls to verify the recorded information.
Statistical analysis
Descriptive statistics for demographic and clinical variables as well as univariate and multivariate analyses were performed using IBM SPSS version 21 (IBM Corp, Armonk, New York). Univariate logistic regression analyses were performed to identify associations between potential predictors and falls during the follow-up period. To evaluate predictors of having at least one fall in the next 12 months, we categorized participants as non-fallers or fallers (≥1 fall). To evaluate predictors of having two or more falls in the next 12 months, we categorized participants as non-recurrent fallers (0 or 1 fall) or recurrent fallers (≥2 falls).
Predictor variables were grouped into ten domains (i.e. demographic, PD severity, PD symptoms, comorbidities, medications, cognition, disability, balance/mobility, self-efficacy and quality of life) and predictors with p < 0.10 in univariate analyses were chosen to be entered into multivariate models. Model 1 investigated predictors of ≥1 fall while model 2 investigated predictors of ≥2 falls. If there was more than one significant predictor in a specific domain, the one with the lowest p-value was selected to enter into the multivariate model. Pearson’s correlation coefficients were used to assess the correlation between the variables. Highly correlated variables (r≥0.75) were not simultaneously entered into the multivariate models. Backward stepwise logistic regression was performed, with p-to-remove set at 0.20.
Receiver Operating Characteristic (ROC) curves were constructed. An area under the curve (AUC) of 0.5 was considered no better than chance, >0.5 to 0.7 was considered low accuracy, >0.7 to 0.9 was considered moderate accuracy, and >0.9 was considered high accuracy [26]. The AUCs of the full multivariate model and the final multivariate model were compared. Kaplan-Meier survival analysis was used to compare the time to first fall between single and recurrent fallers and to explore the timing of second fall in recurrent fallers. A significance level of 0.05 was set for all statistical tests.
RESULTS
Participant characteristics and fall frequency
One hundred and thirty individuals with mild to moderate PD, aged 70.3 years (SD 6.7), were enrolled in this study (Table 1) and all participants completed the 12-month follow-up period. Overall participants had minor balance impairment and disability at baseline. Forty participants (31%) fell at least once during the follow-up period and 21 (16%) were recurrent fallers. Figure 1A shows the distribution of the number of falls (range: 0 to 35). Participants reported 114 falls in total with 10 of the recurrent fallers (48%) experiencing two falls in the follow-up period.
Univariate analysis
Univariate associations between potential predictors and all falls (non-fallers versus fallers) and recurrent falls (non-recurrent fallers versus recurrent fallers) are provided in Table 2. Fallers had greater disability (UPDRS ADL and S&E), reduced balance confidence (ABC) and greater concern about falling (FES-I) than non-fallers (p < 0.05).
More variables were significantly associated with recurrent falls. Recurrent fallers had greater disability (UPDRS ADL and S&E), reduced balance confidence (ABC) and greater concern about falling (FES-I), as well as poorer UPDRS motor scores and longer PD duration than non-recurrent fallers. Recurrent fallers also had worse FRT and BBS and more FOG than non-recurrent fallers (p < 0.05).
Some predictor variables within the same domain were highly correlated (r = 0.75–0.79; p < 0.001). Specifically, the UPDRS ADL and S&E scale were highly correlated, as were the BBS, DGI and TUG.
Multivariate analysis
The predictors entered into multivariate model 1 (≥1 fall as the outcome) were: disease severity (UPDRS motor), FOG, disability (S&E), dynamic balance during gait (DGI) and balance confidence (ABC). The full model had low accuracy in predicting all falls (AUC = 0.68; 95% confidence interval [CI] 0.58–0.78; p = 0.001). Only disability (Odds Ratio [OR] = 0.56 per 10 points increase; 95% CI 0.39–0.80; p = 0.002) remained as a predictor of all falls after backward stepwise regression. This single-predictor model also had low accuracy in predicting all falls (AUC = 0.65; 95% CI 0.55–0.76; p = 0.005)(Table 3) and was not significantly different from the full model (p = 0.32).
The predictors entered into multivariate model 2 (≥2 falls as the outcome) included disease severity (UPDRS motor), FOG, LED, disability (UPDRS ADL), anteroposterior stability (FRT) and balance confidence (ABC). The full model was moderately accurate in predicting recurrent falls (AUC = 0.75; 95% CI 0.64–0.87; p < 0.001). Disability (OR = 1.20; 95% CI 1.07–1.34; p = 0.001) and LED (OR = 1.11 per 100 mg increase; 95% CI 0.95–1.30; p = 0.19) remained as predictors of recurrent falls after backward stepwise regression. This two-predictor model was also moderately accurate in predicting recurrent falls (AUC = 0.72; 95% CI 0.60–0.84; p = 0.001) (Table 3) and was not significantly different from the full model (p = 0.21).
Time to first and second fall
Overall the median time to first fall was 2.1 months (95% CI 0.0–7.2). Twenty fallers (50%) had experienced their first fall within the first 3 months of follow-up, increasing to 26 fallers (65%) within the first 6 months. The median time to first fall was shorter (log rank test: p = 0.002; Wilcoxon test: p = 0.005) in recurrent fallers (1.1 months, 95% CI 0.7–1.5) than for single fallers (6.5 months, 95% CI 3.1–10.0). Nine (47%) single fallers had experienced their fall within the first 6 months of follow-up, while 17 (81%) recurrent fallers had experienced their first fall at this point (Fig. 1B).
The median time to second fall in the recurrent fallers was 7.1 months (95% CI 5.0–9.1). Six (29%) had experienced their second fall within the first 3 months of follow-up, increasing to nine (43%) within 6 months and 15 (71%) within 9 months (Fig. 1C).
DISCUSSION
This is the largest prospective study to date to focus on predictors of falls in people with PD without significant cognitive impairment who had not fallen in the past year. Nearly one third of participants fell in the follow-up period and half of these fell recurrently. Self-reported disability, reduced balance confidence and greater concern about falling were associated with all falls and recurrent falls. In addition, PD severity, disease duration, impaired balance and FOG were found to be associated with recurrent falls. Multivariate analysis revealed that disability was the only independent predictor of all falls and recurrent falls.
Self-reported disability, as measured by the UPDRS ADL and S&E, was associated with all falls and recurrent falls. Some tasks of the UPDRS ADL can be affected by postural instability, such as dressing, hygiene and walking. The S&E rates overall functional independence and slowness in performing daily activities, which may also reflect balance impairment. In contrast, other studies reporting disability measures in people with PD without a history of falls have not found an association with falls [6, 12]. This may be due to the small sample size in Kerr et al. [6] and the younger group in Gazibara et al. [12].
Self-efficacy measures reflecting balance confidence (ABC) and concern about falling (FES-I) were also associated with all falls and recurrent falls in univariate analyses. Both these measures may be considered constructs similar to fear of falling [22], which has been reported to be related to postural instability and disability in people with PD [27]. These findings are in line with Pickering et al’s [3] suggestion that fear of falling may be an early predictor of falls in people with PD without a history of falls. Since disability, balance confidence and concern about falling were the strongest predictors of all falls in our study, these variables should be assessed as potential predictors of first falls in people with PD. In addition, since these risk factors are potentially modifiable with exercise interventions targeting balance impairment [28, 29], they provide targets for early fall prevention programs.
Given the associations between balance confidence and falls, it was surprising that none of the balance and mobility measures investigated in this study were strongly associated with all falls. This finding might be explained by the high baseline performance on balance and mobility measures in our sample. In addition, some balance and mobility measures, for example the BBS, are known to have a ceiling effect [30]. Nevertheless, FRT and BBS were associated with recurrent falls, as well as disease severity. While other studies of people with PD without a history of falls report mixed results for associations between falls and balance/mobility measures [6, 11] as well as disease severity [3, 9–12] these studies do not differentiate between all falls and recurrent falls. The results of the present study suggest that people with PD presenting with disability, reduced balance confidence and/or greater concern about falling can be considered to be at risk of recurrent falls, even when they have minor balance and mobility deficits and do not have a history of falling in the past year.
FOG was a significant predictor of recurrent falls but not all falls. Furthermore, its predictive strength appeared to be less than measures of disability and balance confidence/concern about falling. This may be explained by the low FOG prevalence and severity in our sample, which is expected since FOG has been reported to be associated with longer disease duration and more advanced stage of PD [31, 32]. Similarly, some previous work in those without a history of falls did not report an association between FOG and falls [6, 11], while one study with a four-year follow-up did find an association [10]. Taken together, these results suggest that while FOG is a key risk factor for falls in people with PD [4–7], it appears to be a stronger risk factor when people with a history of falls are included in the sample and in recurrent fallers.
The results of the multivariate analysis show that all falls could be predicted with low accuracy and disability was the only predictor retained in the model. In contrast, recurrent falls could be predicted with moderate accuracy with disability being the only independent predictor, but with LED retained in the model. These findings show that recurrent falls can be predicted with greater accuracy than all falls in people with PD who do not have a history of falls in the past year. This is likely due to recurrent fallers having a larger number of potentially interacting risk factors.
Most fallers in the present study had experienced their first fall within 6 months of follow-up. As expected, recurrent fallers experienced their first fall sooner than single fallers. Furthermore, most recurrent fallers experienced their first fall within 6 months of follow-up while less than half the single fallers had fallen at this point. This finding may be explained by the greater number and strength of predictors for recurrent falls. In line with our study, Mactier et al. [14] reported time to first fall to be shorter for recurrent fallers, however, they reported that 78% of fallers fell within the first 4 months of the study. This difference is likely to be because Mactier et al. [14] included participants with PD with and without a history of falls.
A period of 6 months was able to detect most first falls in this study, but not second falls, as the median time to second fall was 7 months. Furthermore, less than half the recurrent fallers had experienced their second fall within 6 months of follow-up, while by 9 months most had fallen for the second time. Future prospective studies designed to investigate predictors of recurrent falls in people with PD without a history of falls, as well as trials designed to identify the ability of interventions to prevent recurrent falls in this subgroup, should consider a follow-up period of at least 9 months.
The main strength of this study is that disease-specific, self-efficacy and balance measures were simultaneously analyzed for their potential to predict falls and recurrent falls in a large sample of people with PD who had not fallen in the previous 12 months. Falls were recorded by participants through falls diary, which is considered the best currently-available method for collecting fall incidence data [33]. While activity monitors are increasingly able to detect abnormalities of gait [11, 35], to date there is no validated activity monitor which has been shown to accurately detect fall events. We acknowledge that the smaller number of fallers and recurrent fallers might have influenced the number and strength of predictors that were found. Furthermore, there are some potential predictors of falls that were not included in the current study. This study excluded people with significant cognitive impairment, based on the MMSE, which limits the generalizability of our findings. However, there is evidence that alternative measures, such as the Montreal Cognitive Assessment, may be a better predictor of mild cognitive impairment in PD [9]. Moreover, measures of impulsivity and physical activity were not included. Therefore, the results of this study should be interpreted with caution until the models have been externally validated and other potential predictors investigated, such as cognitive impairment.
CONCLUSIONS
Self-report disability was found to be an independent predictor for all falls and recurrent falls in people with PD who had not fallen in the prior 12 months, and should be considered in the assessment of fall risk in people with PD without a history of falls. Notably, recurrent fallers experienced their first fall earlier than single fallers, and most recurrent fallers had fallen for the second time within 9 months. Further research is required to validate the models produced by this study and to explore other potential predictors of falls and recurrent falls in people with PD who have not previously fallen.
CONFLICTS OF INTEREST
The authors have no conflict of interest to report.
Footnotes
ACKNOWLEDGMENTS
The authors thank Nádja Negreiros for her assistance with recruitment and data collection and the people with PD and their families for their participation. This work was supported by CAPES Foundation (Ministry of Education of Brazil), research scholarship awarded to LRSA (document number BEX 9676/14–7).