Abstract

Background. Few studies have examined the relationship of race to falls. This study evaluated the association between potential risk factors and falls in a representative sample of 1049 African American and 1947 white participants of the second in-person wave of the Duke Established Populations for Epidemiologic Studies of the Elderly.

Methods. Information about sociodemographic characteristics, health-related behaviors, health status, visual function, and drug use was determined during baseline in-home interviews. Three years later, falls in the previous 12 months were assessed by self-report.

Results. One or more falls occurred in 22.2% of the participants. Nearly half the fallers reported more than one fall. Multivariable analysis revealed that African Americans were less likely than whites to have any fall (adjusted odds ratio [OR] 0.77, 95% confidence interval [CI] 0.62–0.94). Increased age and education, arthritis, diabetes, and history of broken bones were also significant (p < .05) independent risk factors for any fall. In multivariable analyses comparing those with two or more falls to those with none, again, increased age and education, arthritis, and diabetes were significant (p < .05) independent risk factors while smoking was protective. Race was not a significant predictor of multiple falls (adjusted OR 0.90, 95% CI 0.64–1.26).

Conclusions. Similar sociodemographic characteristics and health problems appear to be important risk factors for any and multiple falls in community-dwelling African American and white elderly residents, with white elders at greater risk of one-time falls.

FALLS are a common phenomenon among elderly adults and are associated with substantial morbidity and mortality (1)(2)(3)(4). Moreover, falls can frequently cause fear that may lead elders to restrict their physical activity and mobility (1)(2)(3)(4). A number of investigators have conducted cohort studies to determine the rate of falls in community-dwelling elders with estimates over the period of 1 year ranging from 25–40% (5)(6)(7)(8).

A number of factors have been shown to be associated with falls in the elderly population, including medication use, age, and certain diseases (e.g., arthritis) (1)(2)(3). There is limited information, however, about the association between falls and race. Means and colleagues conducted a cross-sectional descriptive study of falls in a small sample (n = 298) of white and African American women and found that both groups had similar fall histories (9). Tinetti and associates, in a cohort study, found in bivariate analyses that whites were more likely to fall than those of other races (8). Conflicting findings by Nevitt and colleagues and Studenski and associates have been reported as to whether race is associated with an increased risk of recurrent falls (6)(7).

Given such limited information regarding the impact of race on falls and the relevance of falls to the independent functioning of older persons, the objectives of the present study were to determine whether there is an association between race and falling at all and between race and multiple falls among community residents. In particular, we sought to determine whether race had a modifying effect on the relationship between other sociodemographic characteristics, health-related behaviors, visual function, health status, or medication use and falls.

Methods

Study Design, Sample, and Source of Data

This is a secondary analysis using data from the Duke site of the Established Populations for Epidemiologic Studies of the Elderly (EPESE), a 10-year prospective cohort study of community-dwelling elders (10). The purpose of the Duke EPESE was “to describe and identify predictors of mortality, hospitalization and placement in long term care facilities and to study risk factors for chronic diseases and loss of functioning” ((10), p. 1). An additional major goal of the Duke EPESE was to “study these associated factors among black and white older persons” ((10), p. 1). The study design is presented in detail elsewhere (10). Briefly, using a four-stage stratified household sampling design, a probability sample of 4162 community residents aged 65 years or older living in a five-county urban and rural area of the Piedmont region of North Carolina was selected. African Americans represented one third of the population aged 65 and older in the five counties. However, to enhance statistical precision for this minority group, African Americans were oversampled and comprised more than half of the study respondents. All levels of socioeconomic status were represented among both African American and white sample members. The manner of sample selection permits the development of weights, which take into account race, gender, age group, number of older persons in the household, location of residence, and nonresponse and allows projection of data from the sample to reflect the status of the same age population (approximately 28,000 people) in the sample's geographical region. This study was approved by the Duke University Medical Center Institutional Review Board, and informed consent was obtained from each participant prior to data collection.

Data Collection and Management

At baseline in 1986–1987, trained interviewers, using comprehensive structured questionnaires, gathered information from sample members in their homes. The information sought included sociodemographic characteristics, health-related behaviors, multiple aspects of the participants' physical health, and use of medications. For medications, at baseline, participants were asked whether, during the previous 2 weeks, they had taken any medicines prescribed by a doctor or any other medicines obtained from a store and, if so, to show the interviewer all these medications (11). The interviewer recorded the drug name, dosage form, and the number of dosage forms the respondent reported taking the previous day. Accuracy of medication data entry and management are high as detailed elsewhere (11). In addition, at the in-home interview 3 years later in 1989–1990, participants were asked whether they had a fall in the past 12 months. Those answering yes were asked how many times they had fallen in that time interval and whether they passed out or fainted with any of these falls. For this investigation, the patient cohort was refined to include all African American and white Duke EPESE participants from the baseline (1986–1987) and follow-up (1989–1990) in-person interviews who had complete information regarding baseline medication use and selected risk factors and falls information 3 years later (n = 2996).

Outcome Measures

Information obtained about falls at the 1989–1990 interview was used to create two dichotomous measures. One dichotomous measure indicated whether the participant had any fall in the previous 12 months, excluding those who had syncope (i.e., passed out or fainted). A second dichotomous measure distinguished those with two or more nonsyncopal falls from those with none in the previous 12 months.

Independent Variables

We examined potential risk factors previously reported for falls (1)(2)(3)(12). Sociodemographic factors were represented by dichotomous variables for race, gender, and urban/rural residence and by continuous variables for age, income, and education in years. Health-related behaviors were characterized by two continuous variables, one for smoking (measured in pack-years, which is the average number of packs of cigarettes smoked per day multiplied by the number of years of smoking) and one for alcohol use (average amount of wine, beer, and liquor measured in ounces per day). Weight and height were used to calculate body mass index (weight [kg]/height [m2]). Low (underweight) and high (overweight) body mass index were defined by age, gender, and race categories according to the 15th and 85th percentiles for body mass index from the National Health and Examination Survey II database (13). Health status factors were represented by continuous variables that included a modified (three-item) version of the Rosow-Breslau scale measured in terms of the number of disabilities reported (14), self-rated health (poor, fair, good, excellent, range 0–3), urinary incontinence (1 = never, 2 = hardly ever, 3 = some of the time, 4 = most of the time, and 5 = all the time), and sleep difficulties (sum of five sleep problems, range from 5–15), and dichotomous measures (present/absent) for self-reported health conditions (arthritis, stroke, history of broken bones, diabetes), as well as for severe depressive symptoms (nine or more symptoms on modified Center for Epidemiological Studies–Depression scale) (15), and cognitive impairment (measured by the Short Portable Mental Status Questionnaire) (16). Specifically, for this last dichotomous measure, for those with grade-school education, whites with four or more and blacks with five or more errors are categorized as being cognitively impaired. Whites and blacks with any high school education are permitted one fewer error and those with more than high school education are permitted two fewer errors. In addition, a continuous measure for upper and lower extremity function (range 0–5, disabilities adapted from Nagi) and a dichotomous measure on whether participants had cut down on usual activities were included (17). Visual function was measured by two dichotomous measures: self-reported ability to read ordinary newspaper print and ability to recognize a friend across the street (10). Exposure to drugs was determined from computerized files of participants' prescription drug data that was coded using an updated and modified version of the Drug Product Information Coding System and Iowa Nonprescription Drug Product Information Coding System (18)(19). Dichotomous variables were created for specific drugs/therapeutic classes including benzodiazepines, phenytoin, other central nervous system drugs (excluding benzodiazepines and phenytoin), nonsteroidal antiinflammatory drugs (NSAIDs), prednisone, diuretics, other cardiac drugs (excluding diuretics), and two categorical variables for the number of prescription and nonprescription drugs being taken (12).

Statistical Analyses

The individual was the unit of analysis. Because African Americans made up 54% of the sample but constitute 35% of the population, all data were weighted to make the sample representative of the population. Weighting also maintains the statistical precision afforded by oversampling African Americans. The analysis proceeded in three phases. In the first phase, the data were summarized by percentages and means and standard deviations as appropriate for all outcome measures and independent variables. In the second phase, multivariable logistic regression was used to estimate the effects of the independent variables on each of the dichotomous falls measures (20)(21)(22)(23). For the multivariable analyses, missing values were replaced with regression-predicted imputed scores. Specifically, three separate stepwise logistic regressions were conducted (p < .15 to remain in model) for health status, visual function, and drug use variables to select a sufficient set of predictors useful in making the relevant imputations while minimizing potential collinearity problems. In the final stage of analyses, the significant (p < .15) health status, visual function, and drug use variables were retained and entered simultaneously with all the sociodemographic and health-related behavior variables into logistic regression models to calculate adjusted odds ratios (ORs) and 95% confidence intervals (CIs) using SUDAAN, a specialized software program developed for the analysis of complex sampling designs and that adjusts for the effects of clustering and stratification (24). The underlying statistical assumptions of each model were evaluated, and collinearity was assessed. Significance tests reported for the multivariable models were adjusted to reflect the number of nonmissing cases prior to imputation. All two-way interactions with race were tested.

Results

The characteristics of the cohort at baseline are shown in Table 1 . When asked 3 years later, 22.2% of the 2996 sample members reported one or more falls in the previous 12 months. Specifically, 20.2% of the African American and 23.2% of the whites had one or more falls in the previous year. Among fallers, 45.0% had multiple (2+) falls in the previous 12 months. The percent of frequent fallers was similar for African Americans and whites.

Table 2 shows the factors associated with having at least one fall in the previous 12 months. All of the visual function variables and some of the health status factors (i.e., cognitive impairment, history of stroke, Rosow-Breslau disabilities) were not significant (p > .15) in intermediate stepwise models, and, therefore, they were not included in the final model. Multivariable analysis revealed that increased age and post high school education, arthritis, diabetes, and history of broken bones increased the risk of having a fall. Conversely, African Americans were significantly less likely than whites to have a fall. None of the drug use variables was significantly associated with having a fall. No significant race interactions were found.

Table 3 shows the factors associated with multiple (two or more) falls in the previous 12 months. All of the visual function variables and some of the health status factors (i.e., cognitive impairment, history of stroke, Rosow-Breslau disabilities) were not significant (p > .15) in intermediate stepwise models, and, therefore, they were not included in the final model. Multivariable analyses again showed that increased age and post high school education, arthritis, and diabetes were significant risk factors. In addition, smoking was a protective factor for multiple falls. Neither race nor history of broken bones was a significant risk factor for multiple falls. No significant race interactions were found.

Discussion

We found that more than one in five community residents 68 years of age and older in a five-county area reported a fall in the previous year. This rate is somewhat lower than the 25–40% rates reported in two recent studies of community-dwelling elders that used prospective falls monitoring (5)(8). This may be due in part to the fact that the sample in the current study was “healthier” because those who died after baseline (1986–1987) or were too sick to be interviewed in 1989–1990 and required a proxy interview were not included. The lower fall rate could also be related to the fact that retrospective questions about falls may result in an underestimation (25). However, our 45% rate of recurrent falls among fallers is consistent with other studies (2)(5)(8).

This is one of the first studies that had a large sample of African Americans, the largest minority group among U.S. elders, that has reported the impact of race on fall occurrence. Specifically, we found that, in controlled analyses, African Americans had a 23% reduced risk of experiencing a fall in the preceding year. What are some possible explanations for this finding? One is that it could be due to selection bias where healthier blacks survived to the second in-person wave that was 3 years from the baseline interview. This is unlikely, however, as previous studies by our group have shown that those who dropped out between the first and second in-person interviews due to death did not differ based on race (26). It is also possible that misclassification could have occurred, and blacks may have been more likely to underreport falls. However, evidence from the prospective study by Studenski and colleagues of 306 elderly men, of whom 27.1% were black, does not support this contention (7). Specifically, in that study, they used 30-day calendars for elders to record falls, postcards for patients to mail in should they experience a fall, and phone calls from the investigators to ascertain and follow-up on reported falls. Using these state-of-the-art methods to measure falls, they found that blacks had similar reporting of recurrent falls as did whites (7). In addition, it is possible that the association we found between race and falls is confounded by a third factor. We believe, however, that this is unlikely, as we controlled for a number of well-established risk factors (i.e., arthritis, depression, cognitive impairment, visual impairment, age, physical functional status, drug use) that could potentially confound the association between race and falls. Moreover, we tested all two-way interactions between race and the other independent variables in the final model and found no evidence of significant effect modification. The finding from our study that African Americans compared with whites had a reduced risk of any fall is also consistent with the findings of the study conducted by Tinetti and colleagues (8). In the Tinetti study, bivariate analyses suggested that non-whites were 38% less likely to fall than whites (8). However, this association was neither maintained nor statistically significant in multivariable analyses that controlled for other factors. It is also interesting to note that our finding of no racial difference in the risk of multiple falls is consistent with the findings of Studenski and colleagues (7) using other data, where race was not an important predictor of recurrent falls. However, our finding that race is not associated with multiple falls is in contrast to the findings of Nevitt and associates (6), where nonwhites were 59% less likely to have recurrent falls than whites. The reasons for these disparate findings are not clear. One possible explanation is that there may be racial differences in lower extremity muscle strength. Means and colleagues reported that elderly black women compared with white women had poorer lower muscle strength (9). This contrasts with the findings of Rantanen and associates, who found that black women had greater hip flexion strength than white women (27).

Four factors (advanced age, having arthritis or diabetes, and post high school education) were found to be independently related to having both any and multiple falls. The findings for advanced age and arthritis are consistent with those in a new guideline report as being among the factors most commonly associated with falls (28). Diabetes has been found to be a risk factor in some but not all studies of falls (2)(29). Perhaps the finding with diabetes may be related to peripheral neuropathy in lower extremities. One study found that among those with diabetes, persons with sensory neuropathy, when compared to those without sensory neuropathy, had greater postural instability as measured by force platform (30). Further attention to examining diabetes as a risk factor for falls and the impact that improved treatment may have on falls is warranted. To the best of our knowledge, the associations between education with any and multiple falls and smoking with multiple falls have not been identified previously, and the possible reasons for the associations are unclear (31).

It is interesting to note that a number of risk factors examined (i.e., depression, cognitive impairment, visual impairment, lower extremity difficulties, and drug use) were not associated with falls. Lack of association for cognition, depression, and visual problems may be due to low power given the small number of participants with problems in these areas. The lack of an association with drug use is a more complex issue. We did find in bivariate analyses that both NSAID (crude OR 1.52; 95% CI 1.20–1.91) and benzodiazepine (crude OR 1.50; 95% CI 1.14–1.96) use was associated with any fall. However, after controlling for health status factors that included some of the most common indications for these agents (arthritis and sleep problems), the associations diminished and were no longer statistically significant. The lack of association for NSAIDs was consistent with that found in a recent meta-analysis (32). In contrast to previous studies, we controlled for confounding by indication (i.e., arthritis) in our multivariable analyses. Moreover, for benzodiazepine use, we did find that the point estimate for multiple falls (adjusted OR 1.49) was consistent with that found in a recent meta-analysis (33) and held for those taking either long or short half-life benzodiazepines (data not shown). It is possible that drug use may only be an important risk factor in “frailer elders” (e.g., institutionalized elders).

There are several potential limitations to our study. The study involved only nonproxy participants with complete drug and fall information because proxies were not asked about falls. Another potential limitation was that the fall rate we determined may be an underestimate. However, it was encouraging to find that the rate of recurrent fallers was similar to that found in other studies. A third limitation is that we were not able to examine the issue of injurious falls because the Duke EPESE survey did not ask about this element. Injurious falls are important because it has been reported that 2% to 6% of falls result in fracture. Moreover, falls account for nearly 90% of all fractures in people aged 65 years and older (31). However, our group has previously published information that, compared with white women, African American women had a reduced risk of experiencing nonvertebral fractures (34). A fourth limitation was that we were not able to examine the impact of certain environmental factors, use of assistive devices, or muscle strength or balance because the Duke EPESE did not ask participants about these issues. We did, however, examine a number of important potential risk factors including arthritis, depression, cognitive impairment, visual impairment, age, physical functional status, drug use, and race. A fifth limitation is that, given the low incidence of recurrent falls, we may have had limited power to detect any association with race in the logistic regression model. However, the calculated point estimate is likely to be a first approximation of the true magnitude of the association. Finally, this is a study of community-dwelling elders living in the southeastern United States and may not be representative of other populations elsewhere.

Despite these potential limitations, we found that race was associated with any fall but not with recurrent falls. Moreover, certain sociodemographic characteristics (age, post high school education) and health problems (arthritis and diabetes) were consistently associated with any and multiple falls in nonproxy community-dwelling elderly residents. Future prospective cohort studies examining the importance of potential racial differences in falls are warranted. If our findings are replicated, additional studies should be conducted to determine why there are racial differences and to find out what to do about them.

Table 1.

Descriptive Characteristics of Duke EPESE Baseline Sample (n = 2996)

Variables Weighted Mean or Proportion SD 
Sociodemographics   
Age (years) 72.34 5.82 
Female .64  
African American .35  
Rural residence .44  
Education (years) 9.93 7.21 
Income (dollars/year) 13,201 11,760 
Health-Related Behaviors   
Body mass index 25.80 5.04 
Alcohol use (ounces/day) 0.09 0.30 
Smoking (pack-years) 17.04 29.27 
Health Status   
Rosow-Breslau disabilities .69 1.02 
Cognitive impairment .07  
Depression symptoms .06  
Nagi disabilities 0.86 1.26 
Arthritis .53  
Diabetes .14  
Stroke .05  
History of broken bones .23  
Sleep problems 8.20 2.29 
Incontinence 1.94 1.08 
Self-rated health 2.38 0.88 
Reduced activities .31  
Visual Function   
Unable to read newspaper print .09  
Unable to recognize friend across street .07  
Drug Use   
Diuretic .32  
Other cardiac drugs .52  
Benzodiazepine .11  
Phenytoin .01  
Other central nervous system drugs .11  
Nonsteroidal antiinflammatory drugs .16  
Prednisone .01  
Number of prescription drugs   
0–1 .45  
2–4 .42  
5+ .13  
Number of over-the-counter drugs   
0–1 .69  
2–3 .27  
4+ .04  
Variables Weighted Mean or Proportion SD 
Sociodemographics   
Age (years) 72.34 5.82 
Female .64  
African American .35  
Rural residence .44  
Education (years) 9.93 7.21 
Income (dollars/year) 13,201 11,760 
Health-Related Behaviors   
Body mass index 25.80 5.04 
Alcohol use (ounces/day) 0.09 0.30 
Smoking (pack-years) 17.04 29.27 
Health Status   
Rosow-Breslau disabilities .69 1.02 
Cognitive impairment .07  
Depression symptoms .06  
Nagi disabilities 0.86 1.26 
Arthritis .53  
Diabetes .14  
Stroke .05  
History of broken bones .23  
Sleep problems 8.20 2.29 
Incontinence 1.94 1.08 
Self-rated health 2.38 0.88 
Reduced activities .31  
Visual Function   
Unable to read newspaper print .09  
Unable to recognize friend across street .07  
Drug Use   
Diuretic .32  
Other cardiac drugs .52  
Benzodiazepine .11  
Phenytoin .01  
Other central nervous system drugs .11  
Nonsteroidal antiinflammatory drugs .16  
Prednisone .01  
Number of prescription drugs   
0–1 .45  
2–4 .42  
5+ .13  
Number of over-the-counter drugs   
0–1 .69  
2–3 .27  
4+ .04  

Note: EPESE = Established Populations for Epidemiologic Studies of the Elderly.

Table 2.

Risk Factors for Any Falls in Community-Dwelling Elders: Duke EPESE (n = 2996)

Variables Adjusted OR 95% CI 
Sociodemographics   
Age (per year) 1.04 1.02–1.06* 
Female 1.28 0.96–1.69 
African American 0.77 0.62–0.94* 
Rural residence 1.06 0.86–1.30 
Education (years)   
≤8 reference — 
9–11 0.98 0.74–1.29 
12 0.83 0.53–1.30 
13+ 1.49 1.05–2.12* 
Income 0.99 0.97–1.00 
Health-Related Behaviors   
Underweight 1.04 0.73–1.48 
Overweight 1.11 0.80–1.55 
Normal weight reference — 
Alcohol use (ounces per day) 1.26 0.44–3.44 
Smoking (pack-years) 0.92 0.85–1.00 
Health Status   
Depression 1.03 0.98–1.07 
Nagi disabilities 1.06 0.97–1.16 
Arthritis 1.59 1.28–1.98* 
Diabetes 1.36 1.04–1.79* 
Broken bones 1.37 1.06–1.73* 
Sleep problems 1.04 0.99–1.09 
Incontinence 1.04 0.79–1.37 
Self-rated health 1.11 0.95–1.29 
Cut down on activities 1.21 0.94–1.56 
Drug Use   
Benzodiazepine 0.99 0.71–1.39 
Nonsteroidal antiinflammatory drugs 1.02 0.78–1.32 
Diuretic 1.07 0.80–1.42 
Other cardiac drugs 0.84 0.60–1.18 
Phenytoin 0.95 0.40–2.27 
Other central nervous system drugs 1.20 0.85–1.70 
Prednisone 0.96 0.44–2.09 
Number of prescription drugs 1.05 0.96–1.15 
Number of over-the-counter drugs 0.96 0.44–2.09 
Variables Adjusted OR 95% CI 
Sociodemographics   
Age (per year) 1.04 1.02–1.06* 
Female 1.28 0.96–1.69 
African American 0.77 0.62–0.94* 
Rural residence 1.06 0.86–1.30 
Education (years)   
≤8 reference — 
9–11 0.98 0.74–1.29 
12 0.83 0.53–1.30 
13+ 1.49 1.05–2.12* 
Income 0.99 0.97–1.00 
Health-Related Behaviors   
Underweight 1.04 0.73–1.48 
Overweight 1.11 0.80–1.55 
Normal weight reference — 
Alcohol use (ounces per day) 1.26 0.44–3.44 
Smoking (pack-years) 0.92 0.85–1.00 
Health Status   
Depression 1.03 0.98–1.07 
Nagi disabilities 1.06 0.97–1.16 
Arthritis 1.59 1.28–1.98* 
Diabetes 1.36 1.04–1.79* 
Broken bones 1.37 1.06–1.73* 
Sleep problems 1.04 0.99–1.09 
Incontinence 1.04 0.79–1.37 
Self-rated health 1.11 0.95–1.29 
Cut down on activities 1.21 0.94–1.56 
Drug Use   
Benzodiazepine 0.99 0.71–1.39 
Nonsteroidal antiinflammatory drugs 1.02 0.78–1.32 
Diuretic 1.07 0.80–1.42 
Other cardiac drugs 0.84 0.60–1.18 
Phenytoin 0.95 0.40–2.27 
Other central nervous system drugs 1.20 0.85–1.70 
Prednisone 0.96 0.44–2.09 
Number of prescription drugs 1.05 0.96–1.15 
Number of over-the-counter drugs 0.96 0.44–2.09 

Note: EPESE = Established Populations for Epidemiologic Studies of the Elderly; OR = odds ratio; CI = confidence interval.

*

p < .05.

Table 3.

Risk Factors for Two or More Falls in Community-Dwelling Elders: Duke EPESE (n = 2634)

Variables Adjusted OR 95% CI 
Sociodemographics   
Age (per year) 1.04 1.02–1.06* 
Female 0.91 0.63–1.30 
African American 0.90 0.64–1.26 
Rural residence 1.04 0.77–1.42 
Education (years)   
≤8 reference — 
9–11 0.84 0.57–1.24 
12 0.95 0.51–1.77 
13+ 1.69 1.10–2.60* 
Income 0.98 0.96–1.00 
Health-Related Behaviors   
Underweight 0.88 0.59–1.31 
Overweight 0.98 0.63–1.51 
Normal weight reference — 
Alcohol use (ounces per day) 0.87 0.20–3.70 
Smoking (pack-years) 0.89 0.80–0.99* 
Health Status   
Depression 1.05 0.99–1.10 
Nagi disabilities 1.12 0.98–1.27 
Arthritis 2.19 1.57–3.05* 
Diabetes 1.46 1.02–2.08* 
Broken bones 1.35 0.97–1.86 
Sleep problems 1.06 0.99–1.13 
Incontinence 1.10 0.81–1.50 
Self-rated health 1.25 0.99–1.56 
Cut down on activities 1.33 0.95–1.85 
Drug Use   
Benzodiazepine 1.49 0.98–2.28 
Nonsteroidal antiinflammatory drugs 1.10 0.79–1.53 
Diuretic 0.94 0.67–1.31 
Other cardiac drugs 1.15 0.74–1.79 
Phenytoin 1.54 0.56–4.23 
Other central nervous system drugs 1.00 0.60–1.67 
Prednisone 1.18 0.42–3.30 
Number of prescription drugs 1.00 0.91–1.10 
Number of over-the-counter drugs 1.12 0.97–1.29 
Variables Adjusted OR 95% CI 
Sociodemographics   
Age (per year) 1.04 1.02–1.06* 
Female 0.91 0.63–1.30 
African American 0.90 0.64–1.26 
Rural residence 1.04 0.77–1.42 
Education (years)   
≤8 reference — 
9–11 0.84 0.57–1.24 
12 0.95 0.51–1.77 
13+ 1.69 1.10–2.60* 
Income 0.98 0.96–1.00 
Health-Related Behaviors   
Underweight 0.88 0.59–1.31 
Overweight 0.98 0.63–1.51 
Normal weight reference — 
Alcohol use (ounces per day) 0.87 0.20–3.70 
Smoking (pack-years) 0.89 0.80–0.99* 
Health Status   
Depression 1.05 0.99–1.10 
Nagi disabilities 1.12 0.98–1.27 
Arthritis 2.19 1.57–3.05* 
Diabetes 1.46 1.02–2.08* 
Broken bones 1.35 0.97–1.86 
Sleep problems 1.06 0.99–1.13 
Incontinence 1.10 0.81–1.50 
Self-rated health 1.25 0.99–1.56 
Cut down on activities 1.33 0.95–1.85 
Drug Use   
Benzodiazepine 1.49 0.98–2.28 
Nonsteroidal antiinflammatory drugs 1.10 0.79–1.53 
Diuretic 0.94 0.67–1.31 
Other cardiac drugs 1.15 0.74–1.79 
Phenytoin 1.54 0.56–4.23 
Other central nervous system drugs 1.00 0.60–1.67 
Prednisone 1.18 0.42–3.30 
Number of prescription drugs 1.00 0.91–1.10 
Number of over-the-counter drugs 1.12 0.97–1.29 

Note: EPESE = Established Populations for Epidemiologic Studies of the Elderly; OR = odds ratio; CI = confidence interval.

*

p < .05.

The data on which this publication was based were obtained pursuant to Contract N01-AG-1-2101 from the National Institute on Aging in support of the Established Populations for Epidemiologic Studies of the Elderly (Duke). The content of this publication does not necessarily reflect the views or policies of the U.S. Department of Health and Human Services.

This work was also supported by National Institute on Aging Research Grant R01-AG15432 and the VFW Endowed Chair in Pharmacotherapy for the Elderly, College of Pharmacy, University of Minnesota.

This paper was presented in part at the 52nd Annual Scientific Meeting of The Gerontological Society of America, November 1999, San Francisco, CA.

We thank Arline Bohannon for her assistance in early phases of this project and Brenda Davis for her help in preparing the manuscript.

References

1
Tinetti ME,
1989
. Prevention of falls among the elderly.
N Engl J Med
 
320
:
1055
-1059.
2
Studenski S, Wolter L,
1998
. Instability and falls. Duthie E, Katz P, , ed.
The Practice of Geriatrics
  3rd ed.
131
-144. WB Saunders, Philadelphia.
3
Tideiksaar R. Falling in Old Age: Prevention and Management. 2nd ed. New York: Springer; 1997.
4
Grisso JA, Schwarz DF, Wolfson V, Polansky M, LaPann K,
1992
. The impact of falls in an inner-city elderly African-American population.
J Am Geriatr Soc
 
40
:
673
-678.
5
Campbell AJ, Borrie MJ, Spears GF,
1989
. Risk factors for falls in a community-based prospective study of people 70 years and older.
J Gerontol Med Sci
 
44
:
M112
-M117.
6
Nevitt MC, Cummings SR, Kidd S, Black D,
1989
. Risk factors for recurrent nonsyncopal falls: a prospective study.
JAMA
 
261
:
2663
-2668.
7
Studenski S, Duncan P, Chandler J, et al.
1994
. Predicting falls: the role of mobility and nonphysical factors.
J Am Geriatr Soc
 
42
:
297
-302.
8
Tinetti ME, Speechley M, Ginter SF,
1988
. Risk factors for falls among elderly persons living in the community.
N Engl J Med
 
319
:
1701
-1707.
9
Means KM, O'Sullivan PS, Rodell DE,
2000
. Balance, mobility, and falls among elderly African American women.
Am J Phys Med Rehabil
 
79
:
30
-39.
10
Cornoni-Huntley J, Blazer DG, Lafferty ME, Everett DF, Brock DB, Farmer ME, eds. Established Populations for Epidemiologic Studies of the Elderly. Vol 2. Resource Data Book. Bethesda, MD: National Institute on Aging; 1990. NIH publication 90-495.
11
Hanlon JT, Fillenbaum GG, Burchett B, et al.
1992
. Drug use patterns in black and nonblack community dwelling elderly.
Ann Pharmacother
 
26
:
679
-685.
12
Hanlon JT, Cutson T, Ruby CM,
1996
. Drug-related falls in the elderly.
Top Geriatr Rehabil
 
11
:
38
-54.
13
Rowland ML,
1989
. A nomogram for computing body mass index.
Dietetic Currents
 
16
:
5
-12.
14
Rosow I, Breslau N,
1966
. A Guttman health scale for the aged.
J Gerontol.
 
21
:
556
-559.
15
Blazer D, Burchett B, Service C, George LK,
1991
. The association of age and depression among the elderly: an epidemiologic exploration.
J Gerontol Med Sci
 
46
:
M210
-M215.
16
Pfeiffer E,
1975
. A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients.
J Am Geriatr Soc
 
23
:
433
-441.
17
Nagi SZ,
1976
. An epidemiology of disability among adults in the United States.
Milbank Mem Fund Q
 
54
:
439
-468.
18
DeVito CA, Aldridge GW, Wilson A, et al.
1979
. Framework and development of a comprehensive drug product coding system.
Contemp Pharm Pract
 
2
:
62
-65.
19
Helling DK, Lemke JH, Semla TP, Wallace RB, Lipson DP, Cornoni-Huntley J,
1987
. Medication use characteristics in the elderly: the Iowa 65+ Rural Health Study.
J Am Geriatr Soc
 
35
:
4
-12.
20
Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic Research: Principles and Quantitative Methods. New York: Van Nostrand Reinhold; 1982.
21
Kleinbaum DG, Kupper LL, Muller KE. Applied Regression Analysis and Other Multivariable Methods. 2nd ed. Boston: PWS-Kent Co.; 1988.
22
Neter J, Wasserman W, Kutner MH. Applied Linear Regression Models. Homewood, IL: Richard D. Irwin Inc.; 1989.
23
Hosmer D, Lemeshow S. Applied Logistic Regression. New York: Wiley; 1989.
24
SUDAAN: Professional Software for Survey Data Analysis. Research Triangle Park, NC: Research Triangle Institute; 1998.
25
Cumming SR, Nevitt MC, Kidd S,
1988
. Forgetting falls: the limited accuracy of recall of falls in the elderly.
J Am Geriatr Soc
 
36
:
613
-616.
26
Fillenbaum GG, Horner R, Hanlon J, et al.
1996
. Factors predicting change in prescription and non-prescription drug use in a community-residing black and white elderly population.
J Clin Epidemiol
 
49
:
587
-593.
27
Rantanen T, Guralnik JM, Leveille S, et al.
1988
. Racial differences in muscle strength in disabled older women.
J Gerontol Biol Sci
 
53
:
B355
-B361.
28
American Geriatrics Society, British Geriatrics Society, and American Academy of Orthopaedic Surgeons Panel on Fall Prevention
2001
. Guideline for the prevention of falls in older persons.
J Am Geriatr Soc
 
49
:
664
-672.
29
Stone JK, Wyman JF,
2000
. Falls. Stone JK, Wyman JF, Salisbury SA, , ed.
Clinical Gerontological Nursing: A Guide to Advanced Practice
  2nd ed.
341
-366. WB Saunders, Philadelphia.
30
Simoneau GG, Ulbrecht JS, Derr JA, et al.
1994
. Postural instability in patients with diabetic sensory neuropathy.
Diabetes Care
 
17
:
1411
-1421.
31
Fink H, Wyman J, Hanlon JT. Falls. In: Tallis R, Fillit H, eds. Brocklehurst's Textbook of Geriatric Medicine. 6th ed. London, UK: Churchill Livingstone; 2002. In press.
32
Leipzig RM, Cumming RG, Tinetti ME,
1999
. Drugs and falls in older people: a systematic review and meta-analysis: II. Cardiac and analgesic drugs.
J Am Geriatr Soc
 
47
:
40
-50.
33
Leipzig RM, Cumming RG, Tinetti ME,
1999
. Drugs and falls in older people: a systematic review and meta-analysis: I. Psychotropic drugs.
J Am Geriatr Soc
 
47
:
30
-39.
34
Bohannon A, Hanlon JT, Landerman LR, Gold DT,
1999
. Relationship between race and other potential risk factors and nonvertebral fractures in community dwelling elderly women.
Am J Epidemiol
 
149
:
1002
-1009.