Falls in the general elderly population: a 3- and 6- year prospective study of risk factors using data from the longitudinal population study ‘Good ageing in Skane’

Although accidental falls in the elderly have been the subject of extensive research during the past 20 years, it is still a major health problem in a rapidly ageing global population [1‐6]. Unintentional injuries are the fifth leading cause of death in older adults after cardiovascular, neoplastic, cerebrovascular and pulmonary causes. Falls are responsible for two-thirds of the fatalities resulting from unintentional injuries [2, 7]. About a third of community-dwelling people over 65 years fall each year, and the incidence increases with age [1]. Approximately 20% of accidental falls require medical attention, and 5% results in a fracture or other serious injuries [8]. The additional psychological and social consequences can be severe, with post-fall syndromes including fear, depression and activity avoidance [2, 7, 9, 10]. The high incidence of falls in the elderly, with substantial mortality and morbidity, underlines the importance of preventive interventions. Many preventive programmes and randomised controlled trials based on reported risk factors have been conducted and evaluated over the years [1, 11, 12]. Although some interventions have proven to be effective in reducing falls, there is still some uncertainty about the optimal approach and the efficacy of interventions aimed at preventing falls, probably due to the complex nature and aetiology of the clinical problem [2, 7, 8, 12‐15].

Risk factors for falls have been identified in epidemiological studies of varying quality [1]. Meta-studies show a notable heterogeneity in the selection, scope and methodology employed in previous studies, in which falls were assessed retrospectively, or small, unrepresentative samples of the general elderly population were studied with short follow-up periods [3, 5, 16, 17]. As retrospective and cross-sectional studies can identify risk factors, they are methodologically weaker than prospective studies, which have the ability to assess outcome after exposure to risks [18]. Three systematic reviews of studies on risk factors for falls in subjects over 64 years of age include 23 prospective studies, from 1988 to 2009 [3, 16, 17]. The follow-up period ranged from 16 days to a maximum of 16 months, and most had an one-year duration. An additional, recent systematic meta-analysis included 74 prospective studies from 1988 to 2009, in which risk factors for falls in community-dwelling older people were analysed [19]. The majority of these studies had a duration of one year or less, and sample sizes less than 500 subjects. Five studies had a duration of over 36 months (range 48–84 months), although they focused on one or few risk factors, and had skewed gender distributions.

As the cause of falls in the elderly is largely multifactorial, it is relevant to analyse a broad range of intrinsic and extrinsic risk factors for falls [7, 20]. Saari (2007) and Pluijm (2006) applied a multifactorial approach with 10- and 3-year follow-ups, respectively [21, 22]. Anstey (2006) employed an 8-year prospective design in studying the relation between cognitive performance and falls [23]. Besides these, to our knowledge, few long-term, multifactorial prospective cohort studies on risk factors for falls in the general elderly population have been carried out. Such long-term prospective studies are vital if we are to understand the processes behind falling, to be able to predict and define groups at risk, and to improve the design of future clinical trials [18]. As the first step in an effective intervention programme may be identifying risk factors for falls, their identification in a long-term perspective may be vital in future interventions aimed at a stable, long-term reduction of falls.

This study is based on a large cohort from the general elderly population, including the very elderly, with long-term follow-up assessments. The aim of this study was to identify risk factors predicting falls in a general elderly population after three and six years. The results are also presented in terms of thematic components, using a multifactorial approach in which a wide range of risk factors for falls was analysed.

The study population included 45.7% men and 54.3% women. Just over half the study population was in their sixties at baseline (55.5%). Approximately one in five was in their seventies or eighties (19.9% / 21.5%), and only 3.1% was in their nineties. Table 1 presents data on the incidence of falls and the basic characteristics of those defined as fallers. At the 3-year follow-up assessment, including the older subjects (78 years and older at baseline), the incidence of falls in the past six months was almost one in five (19.1%). At the 6-year follow-up assessment, almost one in seven of the participants had fallen once or more in the past six months (13.3%). Falls were almost four times more common in 90-year-olds than in 60-years olds at the 6-year assessment, and women fell almost twice as often as men throughout the study. The prevalence of risk factors at baseline, and their association with falls are presented in Table 2. Eleven factors were statistically significant in the age- and sex-adjusted regression model (indicated by boldface in the far right-hand column of Table 2), with ORs for falling ranging from 1.35 to 3.30. The use of neuroleptics (OR 3.30, 95% CI 1.15–9.43), heart failure with symptoms (OR 1.88, 95% CI 1.17–3.04), and low walking speed (OR 1.77, 95% CI 1.28–2.46) were the most prominent risk factors.

The eleven statistically significant factors were included in the PCA. Examination of the correlation matrix obtained from the PCA revealed the presence of nine coefficients of 0.3 and above, the value indicating a moderate positive relationship according to our definition. The analysis revealed four independent components with eigenvalues exceeding 1, explaining 51.7% of the variance. Inspection of the screen plot revealed a clear break after the third component. As the fourth component consisted of a single variable, a three-component solution was chosen (Table 3). Direct oblimin rotation was used to calculate the factor loadings of the variables and no correlation between the components over 0.3 was found. The first component, called reduced mobility, included the factors indoor adaptation, osteoarthritis of hip and/or knee, low walking speed and impaired mobility. It explained 21.3% of the variance and had an OR for falling of 2.12 (95% CI 1.54–2.91). The second component, consisting of heart disease, heart failure with symptoms and the use of fall risk drugs, was called heart dysfunction, and explained 11.2% of the variance with an OR for falling of 1.66 (95% CI 1.26–2.20). The final component, called functional impairment, included dependence in pADL and nocturia, and explained 9.8% of the variance, with an OR for falling of 1.38 (95% CI 1.01–1.88).

The attrition analysis revealed that there were significantly relatively more 80- and 90-year-olds in the group of non-participants than in the included study population (Table 4). Risk factors such as heart disease, tendency to fall, the use of fall risk drugs and low walking speed were significantly more prevalent in the non-participants.

We found that the component reduced mobility was a prominent predictor of falling. This is in accordance with previous findings, where several studies have revealed a relationship between low walking speed and falls in the elderly [43, 44]. Knee pain and lower back pain have also been significantly associated with multiple falls, predominantly in women [44, 45]. Osteoarthritis and pain in the musculoskeletal system may reduce the ability of the individual to maintain an upright position in non-ideal surroundings, and treatment with opioids may cause additional adverse effects in the elderly, such as dizziness. Randomised trials and systematic reviews have shown that exercise, strength and balance training can significantly reduce both non-injurious and injurious falls in the elderly [1, 8, 13].

The association between heart dysfunction and future falls is partly a confirmatory finding, where the most common cardiovascular disorders associated with falls are carotid sinus hypersensitivity, vasovagal syndrome and brady- and tachyarrhythmias [46]. However, heart failure with symptoms was a prominent individual predictor, which may not have been widely reported. This may be clinically relevant, as heart failure is a cardiovascular disease with increasing incidence and prevalence, and in an ageing population it can be expected to rise significantly in the coming decades [47, 48]. According to a previous review, there is a relative lack of published clinical trials in which the efficacy of heart failure treatment has been assessed exclusively in elderly individuals [47]. In a recent geriatric study, falls and urinary incontinence were reported as being common comorbidities in individuals with heart failure approaching the end of life, and there is limited knowledge on the best way to manage these individuals [49]. The elevated risk of falling in the elderly with heart failure may be explained by its pharmacotherapy (e.g. diuretics, ACE inhibitors, beta-blockers) and their association with falls, although these categories of pharmaceuticals did not individually predict falls in our study. Response to pharmacotherapy for heart failure varies in elderly individuals, who are susceptible to adverse events such as orthostatic hypotension, dehydration, electrolyte disturbance, incontinence and drug-drug interactions [47]. One common side effect of diuretics is hyponatraemia, which is associated with impairment of gait and attention. Stable mild chronic hyponatraemia is generally considered asymptomatic, although Renneboog (2006) observed a high number of falls in comparison to controls in hyponatraemic subjects considered clinically asymptomatic. Furthermore, individuals with moderate chronic hyponatraemia fell dramatically more frequently than patients with normal serum sodium levels [50].

The component functional impairment, including dependence in pADL and nocturia, predicted falls, and has previously been reported to be a risk factor [2, 4, 51]. In the limited studies of fall risk factors in the very elderly, dependence in pADL was found to be one of the most important independent predisposing factors in a population-based study including individuals 85–103 years old [4]. In a recent study, the elevated risk of falling in adults over 85 years appeared to be due to the overall deterioration of health with age. Among those with excellent overall health, the risk of falling in adults over 85 years was no higher than in those 65–84 years of age [52]. The association between nocturia and falls may be explained by incontinence as a symptom of a general deterioration of health. Night-time visits to the toilet in darkness may also contribute to the risk of falling [53]. There may also be an association between nocturia and the treatment for heart failure, as these groups both showed an elevated risk of falling. As diuretic drugs can cause urge incontinence, orthostatic hypotension and hyponatraemia, their relation to nocturia and night-time falls may be underestimated.

The use of neuroleptics, although the prevalence is low, is found to be a prominent individual risk factor for falls in the present study. Although they have not been as widely studied in this field as other psychotropic and psychoactive drugs, such as SSRIs and sedatives, an association between neuroleptics and falls has been reported previously [27, 39, 54]. Unnecessary or incorrect use of drugs is one of the most easily correctable risk factors for falls in the elderly, and it was shown in a randomised trial that gradual withdrawal of psychotropic drugs reduced the risk of falling by 66% [55].

Three main components predicting falls were identified in a general elderly population after three and six years: reduced mobility, heart dysfunction, and functional impairment including nocturia. Furthermore, the use of neuroleptic drugs, although low in prevalence, was a prominent individual risk factor. Heart failure with symptoms was a significant risk factor for falls, and may be of clinical importance as the prevalence in the elderly is increasing worldwide. There is need for further research on the relation between heart failure and falls in the elderly, as the treatment of the condition is poorly documented in this demographic. The findings of this study may be valuable for future trials aimed at sustainable, long-term reduction of falls in the elderly.