| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Elena M. Andresen, PhD, College of Public Health and Health Professions, University of Florida Health Sciences Center, P.O. Box 100182, Gainesville, FL 32610-0182. E-mail: eandresen{at}phhp.ufl.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Balance • Strength • Mobility • Functional status • Quality of life
It is not surprising, therefore, that substantial clinical and research attention has been directed at minimizing falls, which are the principal cause of hip fractures (S. Tennstedt et al., 1998; T. L. Tennstedt, Lawrence, & Kasten, 2001; Tinetti, 1994, 2003). Most research on the antecedents of (or risk factors for) falls has focused on biomedical factors. Among the more commonly studied risk factors are dizziness, impaired balance, lack of postural control, abnormal gait, muscle weakness, and functional disability (Rubenstein & Josephson, 2002). More recent epidemiologic work has linked low falls efficacy and fear of falling to falls (Cumming, Salkeld, Thomas, & Szonyi, 2000; Friedman, Munoz, West, Rubin, & Fried, 2002; Lachman et al., 1998). Fear of falling and low falls efficacy are important not only for their association with falls, but because they can substantially reduce activity levels and quality of life even in the absence of a fall (Bruce, Devine, & Prince, 2002; Lachman et al., 1998; S. Tennstedt et al.; Tinetti, Mendes de Leon, Doucette, & Baker, 1994).
Much has been learned about falls, fear of falling, and falls efficacy among older White women, who are at greatest risk for hip fracture (e.g., Howland et al., 1998; Murphy, Dubin, & Gill, 2003). However, little is known about the prevalence and correlates of these factors among African American men and women (Tinetti, Doucette, & Claus, 1995), even though African Americans as a group have poorer health status than the majority White population (National Center for Health Statistics, 2002). Poorer health status also is associated with greater fear of falling (Murphy, Williams & Gill, 2002; Murphy et al., 2003). There are almost no data regarding these issues in population-based samples of middle-aged persons, and there are no longitudinal studies of which we are aware that prospectively examine risk factors for falls and worsening fear of falling among representative samples of African Americans. In addition, a comprehensive theoretical approach to understanding falling and fear of falling should include the social and environmental circumstances of adults (Andresen & Miller, 2005), and these factors have not been fully examined in longitudinal studies.
Data from the African American Health (AAH) cohort study can address these deficiencies. AAH is a multistage probability sample of 998 African Americans aged 59 to 65 years at baseline and living in St. Louis, Missouri. The recruitment and baseline comprehensive in-home assessments were conducted between September 2000 and July 2001, and the cohort was then followed with telephone assessments at the 1- and 2-year anniversaries of their initial evaluations. Using the baseline data, we have previously demonstrated that falls, fear of falling, and low falls efficacy were surprisingly common in this cohort (Wilson et al., 2005), suggesting that the etiologic timeline needs to be backed up to adequately understand these conditions and how they affect people in their later years. In this study, the AAH baseline data were examined cross-sectionally for the correlates of reported falls in the prior 2 years and fear of falling and low falls efficacy. The longitudinal information was examined prospectively in order to identify risk factors for incident falls and fear of falling in these middle-aged African American men and women over the subsequent 2 years.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
16 (Folstein, Folstein, & McHugh, 1975; Molloy et al., 1996). Participants received in-home baseline assessments that averaged 2.5 hours. Interviewers received intensive training on survey and physical performance testing. Telephone interviews (averaging 12 minutes in length) were conducted at 1 and 2 years after the in-home assessment. All potential risk factors for the prospective analyses are from baseline assessments.
Falls
Three questions measured falling. At baseline, participants were asked, "Have you fallen down in the last 2 years?" "How many times have you fallen down in the last 2 years?" and "In that fall (in any of those falls), did you injure yourself seriously enough to need medical treatment?" Falls requiring medical treatment were classified as injurious. Prospectively, annual telephone interviews asked these questions about the past year, and any fall during these 2 years was counted. We analyzed falling risk by using dichotomized measures of any fall or any injurious fall versus no fall.
Fear of Falling
To measure fear of falling, participants were asked, "Are you afraid of falling?" At baseline, those responding yes (25.8%) were then asked to clarify whether they felt somewhat afraid (13.8%) or very much afraid (12.0%). These questions were repeated at the follow-up telephone interviews, although the degree of fear was not ascertained. Therefore, we analyzed fear of falling by using a dichotomized measure of any fear versus no fear.
Falls Efficacy
Falls efficacy was measured at baseline with a 10-item scale (Tinetti, 1994). Participants were asked how confident (0 = not confident, 10 = completely confident) they were in performing common daily activities without falling. The activities included cleaning house, getting dressed, preparing meals, bathing, shopping, getting in or out of a chair, going up or down stairs, walking around the neighborhood, reaching into cabinets or closets, and answering the telephone. The falls-efficacy scale score was the sum of the 10 responses (range = 0–100; 
=.931). Given the skewed, leptokurtic distribution (598 participants had scores of 100), we dichotomized falls efficacy to contrast those in the lowest quintile (scores 
90, 18.9%) versus all others.
Risk Factors
We placed variables previously identified in the literature as risk factors for falls, fear of falling, or falls efficacy into five categories grouped as theoretical domains, along with other potential confounders: (a) demographics and socioeconomic status; (b) environmental factors; (c) disease and biomedical markers; (d) functional status and performance measures; and (e) walking frequency (a measure of opportunity for falling). We classified measures based on prior standards for this cohort (Miller et al., 2005; Miller et al., 2004) and in such a way as to maximize analytic contrasts (e.g., for continuous measures with skewed distributions). Demographic characteristics were age (continuous variable), gender, marital status, living arrangement, and race consciousness. Race consciousness was measured by a single question: "How often do you think about your race?" (never, once per year, once per month, once per week, once per day, once per hour, or constantly). We dichotomized race consciousness as once per year or never versus all other responses (Miller et al., 2004). In order to assess marital status, we used a set of dummy variables (single, divorced or separated, or widowed), with married as the reference category. We entered living alone as a dichotomous (yes, no) variable.
Socioeconomic status included education, subjective and objective income, and perceived financial barriers. We measured education in years of formal education (range = 0–25). In order to assess subjective income, we used a set of dummy variables (having a comfortable income, not having enough to get by), with having just enough to get by as the reference category. We collapsed annual income categories into a set of two dummy variables reflecting $20,000 or less or refusal to report income, with more than $20,000 as the reference category. We tapped perceived financial barriers by asking participants if they had been unable to see a doctor when they needed to due to cost at any time during the prior year. Environmental factors included self-reported neighborhood desirability, observer rating of the conditions of the resident's neighborhood block, and sampling stratum (inner city vs suburban). Neighborhood desirability was a 4-item scale of the participant's ratings of the neighborhood as a place to live, general feelings about the neighborhood, attachment to the neighborhood, and neighborhood safety from crime (4 = positive, 21 = negative; Andresen, Malmstrom, Miller, & Wolinsky, 2006; Miller et al., 2004). We contrasted the worst quartile with all others. An independent observer completed an assessment of the external appearance of the block on which the respondent lived. This assessment was a 5-item rating of the condition of the houses, amount of noise (from traffic, industry, etc.), air quality, condition of the streets, and condition of the yards and sidewalks in front of homes (5 = excellent, 20 = poor; Andresen et al., 2006; Krause, 1998). We used this as a continuous variable.
We included nine biomedical measures. These were self-reported history of cancer, diabetes, arthritis, or stroke; measured high systolic or diastolic blood pressure; body mass index (BMI); self-reported weight loss; and percent body fat. We classified high systolic blood pressure as 140 and high diastolic as 90 and measured this by using an automated sphygmomanometer. We measured BMI by using a set of two dummy variables. We defined severe underweight as BMI < 20 and obesity as BMI 30, with BMIs of 20–29 as the reference category. Weight loss was a binary contrast (1 = having lost 10 pounds in the past year). We measured body fat composition by using a set of two dummy variables derived from the percentage of body fat obtained using a Tanita (Arlington Heights, IL) bioelectrical impedance scale. One reflected the upper quartile of body fat ( 44%) and the other reflected not performing the test versus all others.
There were nine measures of functional status and performance. Four were simple counts from physical disability indices. The first involved seven basic activities of daily living (ADLs; having any difficulty with bathing, dressing, eating, getting into and out of bed or chairs, walking across a room, getting outside, and using the toilet; potential range = 0–7; Lawton & Brody, 1969). The second involved eight instrumental ADLs (IADLs; having any difficulty with preparing meals, shopping for groceries, managing money, making phone calls, doing light housework, doing heavy housework, getting to places outside of walking distance, and managing medications; potential range = 0–8; Lawton & Brody). The two remaining counts reflected physical performance (Nagi, 1976). Six items tapped lower-body limitations (difficulties in walking a quarter of a mile, walking up and down 10 steps without rest, standing for 2 hours, stooping, lifting 10 pounds, and pushing large objects; range = 0–6). Three items tapped upper-body limitations (difficulties reaching up over one's head, reaching out as if to shake hands, or grasping; range = 0–3).
The remaining functional status measures tapped vision, hearing, cognitive status, clinically relevant levels of depressive symptoms, and physical activities. We measured vision by using a 3-item scale derived from the Health and Retirement Study (2002; =.769), recoded to contrast being in the poorest quintile versus all others. We evaluated hearing by using the single-item Health and Retirement Study self-assessment, dichotomized to reflect fair or poor responses versus all others. We used the Mini-Mental State Examination (Folstein et al., 1975) as a dichotomous variable, contrasting the lowest tertile (scores 27) with individuals who scored 28–30 points. We evaluated clinically relevant levels of depressive symptoms by using the 11-item Center for Epidemiologic Studies–Depression scale. We defined a score of 9 or more on this scale as indicative of a clinically relevant level of depressive symptoms; a score of 9 on the abbreviated scale is equivalent to the standard criterion of 16 or more on the original 20-item version (Kohout, Berkman, Evans, & Cornoni-Huntley, 1993; Miller et al., 2004). We measured physical activities by a set of two dummy variables derived from the Yale Physical Activity Scale (YPAS; Dipietro, Caspersen, Ostfeld, & Nadel, 1993), an activity dimensions summary index. Being in the upper quintile and being in the lower quintile were compared to being in any other quintile.
Physical performance measures included tandem stands, one-leg stands, chair stands, grip strength, and peak expiratory flow (Wolinsky et al., 2005). For the one-leg stand, we allowed participants to choose their preferred leg on which to balance and required them to raise the other foot at least 2 inches off the ground. We categorized the data as three dummy variables: unable or untested combined with those who held the stand for 3 seconds, low performance (3–29 seconds), and high performance (30 seconds), with high performance as the reference category. Untested participants were those for whom the test was considered unsafe, and including them with the persons who had low performance was thus a conservative approach. We derived a set of two dummy variables from timings for tandem stands, eyes closed. One reflected holding the stand for the full 30 seconds, and the other reflected not attempting the test, versus all others. We measured chair stand performance by using a set of two dummy variables derived from the time needed to accomplish five complete rises. One reflected the slowest quartile ( 13.7 s), and the other reflected not performing the test, versus all others. We assessed grip strength by using the average of three Jamar (Jackson, MI) dynamometer attempts with the self-reported stronger hand (range = 3–76 kg), which we recoded into a set of three dummy variables reflecting the highest quintile ( 44.7 kg), the lowest quintile ( 25.0 kg) or not performing the test, versus the middle quintiles. We measured peak expiratory flow by using a standard flow meter (Assess Flow Meter, Respironics, Cedar Grove, NJ) with the participant standing for the assessment. We classified the participant's performance into one of three levels: unable or untested, low performance ( 392 L/min), and high performance (> 392 L/min), with the reference category being low performance.
Finally, we used questions from the YPAS in order to construct a measure of walking over the past 4 weeks, in two parts: the first reflected vigorous walking and the other, leisure walking. We estimated total minutes per week for the combined walking categories and categorized them as none, up to 29 minutes, 30–59 minutes, 60–89 minutes, 90–119 minutes, and 120 minutes or more. Respondents who said they did not walk were the reference category and were compared with the 30-minute interval categories of walkers.
Analytic Approach
Descriptive statistics are presented using the weighted data. We constructed six multiple logistic regression models for the cumulative falls and injurious falls in the 2 years prior to baseline, baseline fear of falling and falls efficacy, and cumulative falls and fear of falling during the 2 years following the baseline interview. We modeled any falls and any injurious falls separately for the combined 2-year period prior to the baseline interview. We compared separately participants who were somewhat afraid or very much afraid of falling with those who expressed no fear of falling. We compared the lowest quintile on the falls-efficacy scale with all others at baseline. Finally, we evaluated all statistical assumptions by using standard procedures (Selvin, 1991).
Model building relied on four procedures. Forced entry of all of the covariates was used first, followed by forward and backward stepwise procedures. The results of these first three approaches were remarkably consistent. We included any covariate that had a statistically significant effect in any of these three modeling procedures within a given outcome measure as a possible covariate in the next step. Then, we reviewed the six models for measures that were common to multiple models or met clinical relevance criteria (i.e., they were measures that were commonly cited in the literature and for which there was some evidence of importance). For example, body weight, weight loss, and percent body fat all contributed statistically in some way to models, and we selected BMI to represent this risk factor in all models (BMI 30 and BMI < 20 vs middle BMI reference) due to its overall prominence in the models and in the literature. Clinically relevant depressive symptoms had been retained in five of six models and this variable was retained in all models. Vision impairment has strong clinical importance, and this variable was retained in all models. Age, gender, and education were included in all models. We ran the final models for all six outcomes with this consistent set of covariates by using the forced-entry method, and model statistics and parameter estimates were compared to the best-fit models based on the first three methods. In prospective models (falls and fear of falling in 2 years after baseline), we retained baseline falls and fear of falling because of prior research on their importance. We ran the final models first without the walking frequency and then, as a sensitivity analysis, with the walking frequency to represent fall-risk opportunity (we excluded the rest of the YPAS measure from the final models). The C statistic, based on area under the curve (predicted probability), was consistently similar between the final model and the best of the best-fit models. Therefore, we present here the final unweighted models using the consistent set of covariates in all models and adjusting for walking frequency.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
Although the level and significance of ORs differed for the baseline and prospective fear-of-falling models, the pattern of results was very similar between the two models: The single predictor variable with a noticeable difference was low neighborhood desirability. This increased fear of falling in the prospective model (OR = 1.49; p .05) but was not significant and in the opposite direction in the cross-sectional model. Older age was protective for fear of falling at baseline and in the prospective model, and higher education was protective in the baseline model but showed only a trend toward protection in the prospective model. The largest single predictor of having fear of falling during the subsequent 2 years was having fear of falling at baseline (OR = 8.14). For low falls efficacy at baseline, higher levels of education and higher levels of weekly walking were protective, whereas diabetes and low scores on the lower-body functional limitations measure, ADLs, IADLs, and depressive symptoms were all associated with increased risk.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Higher levels of education also were associated with falling, which is counterintuitive. In post hoc analysis, we examined the possibility that this relationship was confounded by work status, but none of the ORs changed significantly. At this point, we do not have a good explanation for this finding. Hopefully, additional research in this and other samples may help clarify this anomaly.
We were concerned that the contribution of prior falls, as well as baseline fear of falling and low falls efficacy, might have adversely constrained the potential effects of other risk factors in the prospective models. This was a special concern with regard to risk factors that yielded fairly modest ORs in this study as compared with prior reports (e.g., poorer vision, arthritis). Accordingly, we removed prior falls and baseline fear of falling and low falls efficacy from the prospective models, but this had virtually no effect on the effect size estimates for the remaining functional and comorbidity risk factors (data not shown).
The risk factors for fear of falling were also consistent with prior reports. Female gender (Murphy et al., 2002), low education, some measures of lower-body function, and depressive symptoms were all independently associated with cross-sectionally or prospectively modeled fear of falling. Other factors varied somewhat between the cross-sectional and prospective models for fear of falling (e.g., stroke, diabetes, living alone, low neighborhood desirability). The large effect of baseline fear of falling on future fear of falling was partly based on the stability of this characteristic: Of the 289 respondents who reported fear of falling at baseline, nearly 80% (n = 230) were afraid of falling 2 years later.
For fear of falling, there was a counterintuitive result for age, which was protective. Reasons for this relationship are speculative. In post hoc analyses, we considered the possibility that age and gender interacted, and that the relationship with age might differ for men and women. However, we observed no statistical interaction. In addition, older age was not predictive of falling. It is possible that the somewhat younger age of this cohort (aged 49–65 at baseline) and the more restricted age range partly explain these anomalous relationships. It also is possible that this is due to a cohort effect (Wilson et al., 2005): That is, middle-aged African Americans in this study were not different than the older adults in prior research just because of their younger age, but because they had had different secular and environmental experiences (e.g., living with greater levels of obesity than their predecessors). Additional research may better explain these anomalous relationships.
Risk factors for low falls efficacy were also as expected, including the associations with more lower-body, ADL, and IADL difficulties and the finding that those who walked more frequently were less likely to have low falls efficacy. Thus, lack of confidence in the ability to perform common daily activities without falling is largely driven in this population by difficulties in performing those activities in the first place. However, future researchers also should examine a more sensitive measure of opportunity for falling. This study accounted for lower-body functioning but did not include a measure of behaviors that might provide more opportunity to fall other than walking frequency.
The results of this study contribute substantially to previous reports of the correlates and risk factors among primarily White older adults. Despite their younger ages, falling and fear of falling are common reports among members of the AAH cohort study. Environmental circumstances (self-reported neighborhood desirability and observer rating of the neighborhood) were not consistently linked to fear of falling or falling itself as we had theorized. These general measures may not be sensitive to specific aspects of environmental barriers (e.g., crime, poor sidewalks), and future researchers should examine environmental barriers in more detail. Future interventions may need to account for both individual and environmental social circumstances in order to be effective (Andresen & Miller, 2005).
In summary, this study confirmed a number of risk factors and correlates for falling that have been reported for older adults and White populations in a very important minority subgroup—African Americans—and documented that these phenomena occur in late-middle as well as in older age. In addition to the age and race of the population studied, the strengths of the study include the retrospective and prospective design and the comprehensive set of explanatory factors examined, including the environmental context. Next steps in this area could usefully focus on how early fear of falling and low falls efficacy develop in various population groups, what the life trajectory of these phenomena is from earlier middle age through old age, and what interventions can effectively improve the natural trajectories.
In the meantime, these results, considered as a whole, suggest that clinical interventions to improve falls and fear of falling should focus on increasing physical activity and lower-body strength, reducing ADL and IADL limitations, improving vision, and decreasing depression. Each of these risk factors is amenable to clinical detection and potentially amenable to broad-based public health education and intervention. In order to be successful, it is essential that such interventions also include sound psychological approaches for building patients' beliefs that exercise will improve their health and well-being and that they can exercise safely via recognition and effective management of falls risks and successful coping if falls do occur (Lachman et al., 1997; Tennstedt et al., 1998; Tennstedt et al., 2001). The serious health and economic sequelae of falls and fear of falling continue to reinforce the importance of clinical and public health attention to these issues.
|
Footnotes |
|---|
1 Rehabilitation Outcomes Research Center, Research Services, Department of Veterans Affairs Medical Center, Gainesville, FL.
2 College of Public Health and Health Professions, University of Florida, Gainesville.
3 Center for Research in the Implementation of Innovative Strategies in Practice, Iowa City Veteran's Affairs Medical Center, IA.
4 Department of Health Management and Policy, The University of Iowa, Iowa City.
5 Division of Biostatistics, Washington University School of Medicine, St. Louis, MO.
6 Departments of Internal Medicine and 7Psychiatry, School of Medicine, Saint Louis University, St. Louis, MO.
8 Indiana University Center for Aging Research and Regenstrief Institute, Inc., Indianapolis.
Decision Editor: Linda S. Noelker, PhD
Received for publication April 5, 2005. Accepted for publication December 12, 2005.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  F. D. Wolinsky, T. R. Miller, T. K. Malmstrom, J. P. Miller, M. Schootman, E. M. Andresen, and D. K. Miller Self-Rated Health: Changes, Trajectories, and Their Antecedents Among African Americans J Aging Health, March 1, 2008; 20(2): 143 - 158. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
