Falls among older adults in the South of Brazil: prevalence and determinants

ABSTRACT OBJECTIVE Evaluate the prevalence and the factors associated with the occurrence of falls among older adults. METHODS A cross-sectional study with a representative sample of 1,451 elderly residents in the urban area of Pelotas, RS, in 2014. A descriptive analysis of the data was performed and the prevalence of falls in the last year was presented. The analysis of demographic, socioeconomic, behavioral and health factors associated with the outcome was performed using Poisson regression with adjustment for robust variance according to the hierarchical model. The variables were adjusted to each other within each level and for the higher level. Those with p ≤ 0.20 were maintained in the model for confounding control and those with p < 0.05 were considered to be associated with the outcome. RESULTS The prevalence of falls among older adults in the last year was 28.1% (95%CI 25.9–30.5), and most occurred in the person’s own residence. Among the older adults who fell, 51.5% (95%CI 46.6–56.4) had a single fall and 12.1% (95%CI 8.9–15.3) had a fracture as a consequence, usually in the lower limbs. The prevalence of falls was higher in women, adults of advanced age, with lower income and schooling level, with functional incapacity for instrumental activities, and patients with diseases such as diabetes, heart disease, and arthritis. CONCLUSIONS The occurrence of falls reached almost a third of the older adults, and the prevalence was higher in specific segments of the population in question. About 12% of the older adults who fell fractured some bone. The factors associated with the occurrence of falls identified in this study may guide measures aimed at prevention in the older adult population.


INTRODUCTION
The rapid aging of the population in low-and middle-income countries results in a significant change in morbidity and mortality due to the increase in the prevalence of chronic non-communicable diseases 1,2 . In addition to increased comorbidities, the use of different medications, slow gait, and decreased visual acuity and hearing tend to generate a disabling process of the elderly over time 1,3 . As a consequence of these functional changes, there is an exponential increase in the occurrence of trauma due to external causes, especially falls.
According to the World Health Organization (WHO), a fall is the unintentional movement of the body to a level lower than the initial position, with the inability to correct it in a timely manner 4 . The fall is determined by multifactorial circumstances, in which intrinsic (physiological changes characteristic of aging, presence of morbidities, deficits in balance, vision, hearing or walking) and extrinsic (environmental risks due to poor lighting or slipperiness, risky behaviors, such as climbing chairs or stairs, and those related to daily activities) factors are involved. Most of the time, it results from the interaction between these factors 5 .
About 30% of older adults suffer falls in a one-year period in Brazil, according to the Ministry of Health 1 . Although falls are responsible for the increased risk of injury, emotional problems, and death in this population group, and therefore represents a serious public health problem, it does not receive due attention from Brazilian society 3,6-8 . Thus, evaluating the factors that most often put older adults at risk of falling can provide important information for the planning of public policies that aim to prevent the occurrence of this incident in this population.
The objective of this study was to estimate the prevalence of self-reported falls in older adults (60 years or older) in a medium-sized city in the southern region of Brazil, and to identify their main associated factors.

METHODS
A cross-sectional population-based study in the urban area of Pelotas in 2014, with the objective of understanding the health characteristics of the population aged 60 years and over. The research was conducted by the graduate program in Epidemiology of the Universidade Federal de Pelotas through the research consortium of master's students 9 .
The calculation of the sample size for the prevalence study considered a prevalence of falls in older adults of 34.8% 10 , a 95% confidence level, a margin of error of four percentage points, and a design effect of two. With an increase of 10% for losses, the minimum sample size required would be 1,198 older adults. The minimum sample required for the association study was estimated at 1,113 older adults, a result of the association between falls and gender, based on the following parameters: 95% confidence level, 80% power, 40/60 ratio between unexposed and exposed, fall prevalence of 26.5% among unexposed (male gender) 10 , prevalence ratio of 1.5, design effect of two, additional 10% for losses and refusals, and 15% for control of possible confounding factors.
The sampling process was carried out in two stages. Initially, the census tracts were ordered according to the average income of each inhabitant based on the census of the Brazilian Institute of Geography and Statistics (IBGE) of 2010 11 . The first stage of the sampling process comprised the random draw of the tracts, which resulted in 133 selected tracts. In the second stage, the selection of approximately 30 households in each tract was made through a systematic draw. At the end of the process, 4,123 households were selected for the survey. Based on a previous estimate of the same census, which found 0.43 older adult/domicile, we expected to find at least 12 older adults people per tract, which would result in a sample of approximately 1,700 older adults (all the older adults of the household were included in the study). This number was considered adequate for the outcome, since, according to the sample calculation, a minimum sample of 1,198 older adults would be necessary.
The data collection was performed from January to August 2014 in the household of the older adult, by trained interviewers, through a questionnaire previously tested and built in a digital format, which was applied with the help of netbooks.
The outcome was the prevalence of falls in the older adult in the last year, investigated by the following question: "Have you fallen down from <month in the previous year> to today?". Older adults who reported an affirmative answer had their past-year history of falls investigated by two additional questions about the number of falls that occurred and whether a limb fractured because of the fall.
Demographic characteristics examined as independent variables: gender ( female or male), age (categorized as 60-69, 70-79 and 80 years old or more) and marital status (with partner or without partner). The socioeconomic characteristics were classified according to an instrument of the Brazilian Association of Research Companies (ABEP) 11 and categorized into classes A/B, C, D/E), schooling level in complete years of study (none, 1 to 3, 4 to 7, 8 to 10, 12 or more), currently employed (no or yes). Regarding health variables, the following morbidities were studied: systemic arterial hypertension, diabetes, heart disease, arthritis, osteoporosis, Parkinson's disease, nephropathy, glaucoma, and emphysema. We also investigated the occurrence of stroke at some point in life. Considering the presence of these diseases, a morbidity score was created, categorized as "none", "1", "2" or "≥ 3" morbidities.
The presence of sarcopenia was assessed according to the criteria proposed by the European Working Group on Sarcopenia in Older People 12 . The loss of muscle mass was determined by measuring with an inextensible tape (Cescorf, Brazil) the circumference of the calves of the older adult, based on cut-off points established in the population studied (≤ 34 cm for men and ≤ 33 cm for women were indicative loss of muscle mass) 13 . Muscle strength was estimated based on hand grip strength, evaluated by digital manual dynamometry (Jamar Digital Plus + Hand Dynamometer and Simmons Preston, Canada). Cut-off points for loss of strength were < 30 kg for men and < 20 kg for women 14 . Finally, the muscular performance was evaluated by the 4 meter gait speed test, considering as "loss" a path velocity < 0.8 m/s 14 . The combination of these tests allowed the categorization of individuals into sarcopenic (presence of loss of muscle mass associated with loss of muscle strength or performance) or non-sarcopenic. The detailed methodology of collecting the above data can be verified in a previous article originating from the same study 13 .
The use of medication was evaluated in the 15 days prior to the interview. To better qualify the information, respondents were asked to show the packaging or prescription of the medicines used. We carried out a classification by pharmacological groups according to the ATC classification (Anatomical Therapeutic Chemical Classification System) recommended by WHO 15 . Among the drugs, we classified those potentially related to the occurrence of falls among the elderly: psychoanalytics, psycholeptics, antiepileptics, calcium channel blockers, diuretics, muscle relaxants and, among the medications used for cardiac therapy, digoxin.
Functional ability was assessed from basic activities (eating, bathing, dressing, using the toilet, lying down and getting up from the bed or chair, and controlling the functions of urinating or evacuating) using the Katz index 16 and instrumental activities (using the telephone, going to distant places using transportation, shopping, tidying up, washing clothes, preparing meals, taking medication, and taking care of money) measured by the Lawton scale 17 . For both instruments, the participants were classified as independent (did not need help to perform any activity) or dependent (they needed partial or total help to perform at least one activity). Regarding the behavioral characteristics, self-reported alcohol dependence (measured by the CAGE method) 18 was evaluated positive for the person's perception of their dependence on alcoholic beverages (up to one positive answer = negative; two or more positive answers = positive) and the level of physical activity, measured by the International Physical Activity Questionnaire (IPAQ) 19 , classified as active those who reported practicing ≥ 150 weekly minutes of physical activities in leisure and travel. The gross and adjusted prevalence ratios were obtained by means of Poisson regression, with robust adjustment for the variance, with the calculation of the p-value for heterogeneity or linear trend for ordinal variables. The adjusted analysis was performed respecting the hierarchical levels: in the first, the demographic, socioeconomic and occupational variables (sex, age, marital status, education, economy class, work situation) were added; in the second, the behavioral and healthrelated variables (physical activity level, self-reported alcohol dependence, sarcopenia and morbidity score) were added; in the third level, we added the drugs that could potentially cause falls and in the fourth and last level we added the variables of disabilities. The variables were adjusted to each other within each level and for the higher level. Those with p ≤ 0.20 were maintained in the model for confounding control and those with p < 0.05 were considered to be associated with the outcome.
Due to the complex design, the analyses were weighted considering the census tract (cluster). For the Poisson regression analysis, the svy command was used to correct for complex sampling.
The analyses were conducted in the Stata 12.1  program.   Self-reported alcohol dependence and drug use were not significantly associated with the outcome in the gross analysis. In the adjusted analysis, the occurrence of falls among older adults was associated with gender, age, economic class, schooling, current work, morbidity and functional capacity by the Lawton scale. The women presented a prevalence of falls almost 1.5 times greater when compared to the men. Regarding age, a linear trend was observed, and the highest prevalence of falls was among those aged 80 years or older (PR = 1.27, 95%CI 0.97-1.66), when compared with those aged 60 to 69 years old. Older adults in the C and D/E classes had a higher prevalence of falls when compared to those in the A/B economic class, with a linear trend. For schooling, a linear trend was also observed, and older adults with no schooling presented a 47.0% greater prevalence of falls when compared to those with a high schooling level. The prevalence of falls was 50.0% lower among older adults with a chronic disease when compared to those without chronic disease. However, protection decreased with increasing numbers of chronic diseases. Older adults who were not employed had 30.0% greater prevalence of falls compared to those who were working. Older adults who were dependent for instrumental activities of daily living (Lawton) also had a higher prevalence of falls compared to independent participants (PR = 1.38, 95%CI 1.10-1.73) ( Table 3).
Diabetes, heart disease, stroke, arthritis, osteoporosis and Parkinson' s disease were associated with the occurrence of the outcome in the gross analysis. Associations with osteoporosis, sarcopenia and Parkinson' s disease lost statistical significance in the adjusted analysis. The prevalence of falls was higher among older adults with diabetes (PR = 1.24, 95%CI 1.03-1.50), heart problem (PR = 1.24, 95%CI 1.02-1.51), who had a stroke (PR = 1.44, 95%CI 1.14-1.81) and had arthritis (PR = 1.25, 95%CI 1.06-1.48), when compared to those without these diseases (Table 4).

DISCUSSION
One in three older adults suffered at least one fall in the last year and 12.0% of them fractured some bone as a result of that fall. The injury occurred mainly in their own residence. The prevalence of falls was higher in women, individuals of more advanced age, lower educational level and economic class, who were not employed and who presented functional incapacity for instrumental activities. The prevalence of falls was also higher in patients with diabetes, heart disease, arthritis and those who reported having had a stroke.
The prevalence of falls in the last year found in this study (28.1%) is similar to that found in other Brazilian studies, whose prevalence ranges from 28% to 37.5% 10,20-22 , as well as those found in studies conducted in countries in Latin America and the Caribbean, where prevalence ranged from 21.0% to 34.0% 23,24 . Differences in prevalence among studies should be interpreted with caution, since they may be due to the delineation of studies and the methodologies adopted or because they are point estimates that have error margins 25 .
As in other studies, the majority of the sample was female 10,25-29 , which can be explained by the fact that women live longer than men 30 . In addition, the prevalence of falls in this and other studies was higher among women 10,25-29 , but the mechanisms underlying this association are still unclear. Some factors pointed out by other authors concern the difference in the body composition of the women when compared to the men. Women have lean mass and muscle strength in smaller amounts and greater loss of bone mass due to the reduction of estrogen levels. This increases the likelihood of osteoporosis in this group 25,29,31 and, consequently, the risk of falls.
Age was positively associated with the outcome, which confirms findings in the national and international literature 10,25,[32][33][34] . This relationship occurs because the biological aging process involves structural and functional changes, such as decreased muscle strength and elasticity, impairment of joint stability and dynamics, and changes in the sensory and nervous system. Such changes compromise postural control and are capable of altering gait and balance, culminating, consequently, in an elevated risk of this outcome 28,35 .
Economic class and schooling were inversely associated with the risk of falls, an association also found in other studies 10,36 . A possible explanation for this association would be the greater difficulty of individuals with low socioeconomic status to access health services and consequently prevent factors that could lead to falls 36 .
The prevalence of falls was higher in unemployed older adults. The literature shows an inverse association between income and risk of falls 6,[36][37][38] . It is likely that employed older adults will have a better financial situation, as well as being healthier, as they are fit for work and therefore less likely to fall 6 .
Physical activity is indicated for the prevention of falls among older adults, since certain activities, such as those involving strength and balance, promote increased muscle and bone strength, coordination, walking speed, functional ability and quality of life 39 . However, the present study did not observe an independent association, perhaps because it did not distinguish the types of activity performed by the participants, presenting only whether they were active or not.
Likewise, there was no statistically significant association between falls and sarcopenia after adjustment for several factors, including gender. Baumgartner, in 1998, when stratifying the occurrence of falls by gender, found an increased risk for falls among sarcopenic older adults, but this association was also not statistically significant for females 40 . Nevertheless, even when the analysis was stratified by gender, no association was found between sarcopenia and occurrence of falls (data not shown in the table). Muscular weakness is cited as one of the risk factors for falls in a guideline drawn up by different medical societies 41 and corroborated by Sayer in an analysis of the Hertfordshire cohort 42 . However, muscle weakness is not a necessary or sufficient criterion for sarcopenia -only a collaborator. Perhaps the combination with the other diagnostic factors for the syndrome (loss of muscle mass and muscular performance) influences this association to the point of taking away the statistical significance. Therefore, although theoretically the association between falls and sarcopenia was plausible, it was not verified in this study.
The occurrence of falls was associated with functional capacity, assessed through instrumental activities (Lawton scale), while the association was not observed for basic activities (Katz scale). This result may be due to the fact that basic activities are more closely related to personal care and instrumental activities assess the ability of the older adults to perform activities to manage their life and home independently. Some of these activities involve getting around (using transportation, shopping, tidying up, washing clothes), and failure to perform them efficiently can increase the risk of falls. Perracini et al. report that older adults with difficulty to perform between one and three activities of daily living have 2.37 times more chance of suffering falls 27 . However, the relationship between falls and functional disability is subject to reverse causality bias. Falls among older adults can result in functional disability, either due to the occurrence of fractures that make it impossible to perform the activities or the fear of suffering a new fall 43 . Older adults, with functional limitations that hamper daily activities, which may occur due to lack of balance or motor coordination, are at increased risk of falls.
Elderly people with arthritis, depressive symptoms, orthostatic hypotension, and cognitive, visual, balance, gait, or muscular strength deficits are at increased risk for falls 44,45 . As for diabetes, this relationship is well established in several studies 46,47 , with complications due to peripheral neuropathy, reduced vision and decreased renal function as the main causes. Balance, strength, and gait issues are likely intermediate factors in any association between diabetes complications and increased risk of falls 46 . Strokes may lead to hemiplegia or paresis of the lower limbs, affecting the gait of the individual, who assumes an unstable erect position with an impaired support base. As a consequence, visual dysfunction and spatial-visual injury may also occur, thus influencing the older adult's balance and mobility 47 . In relation to heart problems, certain arrhythmias, such as atrioventricular blocks, sinus node changes and bradycardia may also cause falls 48 . In addition, the fall may be the first sign of an acute asymptomatic myocardial infarction 49 .
Regarding the use of medication considered to potentially cause falls, this study found no association, although several studies have demonstrated the plausibility of this relationship [50][51][52][53] . However, in this study, only drugs that could potentially cause falls were evaluated, whereas in studies that show an association, any type of medication is considered, which could explain in part the divergence in the results.
This study has limitations. The prevalence of falls found may be underestimated, because, due to survival bias, older adults who suffered falls and who had more serious health complications caused by the fall may already have died. The memory bias is also not discarded, considering the study population. However, hopefully this was not such a big problem since factors such as the occurrence of falls tend to be strongly recalled because of the impact they usually have on health. As strengths of the study, it is worth mentioning the quality of the sample method, which allowed the analysis of a representative sample of older adults from the urban area of Pelotas. In addition, quality control was performed in the various stages of the study, such as interviewer training, instrument testing, data quality control, and ensuring greater credibility of the analyzed data. The study was also able to evaluate a series of factors that, according to previous literature, could be related to the occurrence of falls, allowing us to draw a profile of characteristics predisposing to the occurrence of this injury.
The high prevalence of falls among older adults makes it clear that steps should be taken to prevent them. Prevention strategies should be focused on vulnerable groups, i.e. at higher risk for falls, such as older people with low socioeconomic status who are not working and are dependent for functional capacity.