JAGS 39:46-52, 1991
Falls and Injuries in Frail and Vigorous Cornmunitv Elderlv Persons J
J
Mark Speechley, PhD,* and Mary Tinetti, M D t
Identification of different types of falls and fallers among elderly persons might aid in the targeting of preventive efforts. In a representative sample of 336 community elderly, subjects were assigned to Frail, Vigorous, or Transition groups based on observed patterns of clustering among demographic, physical, and psychological variables. The frequency and circumstances of falls in these three groups were then ascertained. As expected, the observed incidence of falling in one year of follow-up was highest in the Frail group (52%) and lowest in the Vigorous group (17%). However, 22% (5123) of falls by
vigorous subjects, but only 6% (5189) of falls by frail subjects, resulted in a serious injury. Compared with frail subjects, vigorous fallers were somewhat more likely to fall during displacing activity (53% vs 31%), with an environmental hazard present (53% U S 29%, and on stairs (27% v s 6%). These findings suggest that fallrelated injuries can be a serious health problem for vigorous as well as frail elderly persons. Injury prevention, therefore, should be directed at all elderly persons but tailored to expected differences in fall circumstances. J
people age they become more diverse with increasing variability in their habits, life experiences, diseases, and disabilities. This heterogeneity among elderly persons can render assessment and treatment of health problems difficult. An additional challenge occurs when the health problem or event of interest has a multifactorial etiology. A fall is an example of a multifactorial event that results from interplay among intrinsic, situational, and environmental factors.’ Investigators have identified multiple chronic diseases and disabilities, such as cognitive impairment, upper and lower extremity disabilities, arthritis, visual impairment, and gait disorders as This strong association bepredisposing to tween chronic disabilities and falling may lead to the conclusion that intrinsic factors predominate in fall etiology and that falling is a health problem only for frail elderly persons. We recently found, however, that 15% of fallers in a representative sample of community
elderly persons had no more than one chronic intrinsic fall risk factor and would not be considered frail.’ Situational and environmental factors may play a great role among these more vigorous elderly fallers. It is important to identify the prevalence, circumstances, and sequelae of falls across the spectrum of elderly persons so that clinical assessment and preventive efforts can be targeted appropriately. A typology based on differences in characteristics of frail and vigorous elderly persons and on circumstances of their falls, would help in developing these clinical assessments and prevention programs. A central problem in constructing such a typology is the definition of Frailty and Vigor. A simple count of disabilities ignores both clustering among characteristics and the fact that not all disabilities predict fall likelihood. One solution is to examine the clustering among a wide array of characteristics to see if characteristics that are correlated statistically may be interpretable in a way that is clinically meaningful and useful. The aim of this study was to identify the prevalence, circumstances, and sequelae of falls across the functional spectrum of elderly persons. We accomplished this aim by first separating subjects into functional groups based on intercorrelations among 18 characteristics and then reviewing the fall situations of subjects in each group.
A
From the *Department of Epidemiology and Biostatistics, University of Western Ontario, London, Ontario, Canada; and the tDepartment of Medicine, Yale University School of Medicine, New Haven, Connecticut. Dr. Tinetti is a recipient of an NIA Academic Award (K08AG00292). This work was conducted while Dr. Speechley was a postdoctoral trainee at the Yale Health and Aging Project, one of four sites under the National Institute of Aging EPESE program. Address correspondence to Mark Speechley, PhD, Department of Epidemiology and Biostatistics, Kresge Building, The University of ’-In n t a n n I.ondon, Ontario, N6A 5C1 Canada.
.”
Am Geriatr SOC39:46-52, 1991
METHODS Description of Sample The sample in the present study has been described in detail elsewhere.’ Briefly, 0002-8614/91/$3.5(
IAGS-IANUARY 1991-VOL. 39, NO. 1
the sample is a subset of the Yale Health and Aging Project (YHAP) cohort from the New Haven site of the Established Populations for Epidemiologic Study of the Elderly (EPESE)program which numbered 2,812 when assembled in 1982.” The 336 persons in our study were selected from the 1,762 persons still living in 1985 who met the following eligibility criteria: age at least 75 years; living in a private residence or one of two designated elderly housing units; able to follow simple commands; able to walk within their homes. Of the 458 eligble subjects, 114 (25%) refused to participate and eight (2%) could not be found. Participants did not differ from eligible non-participants on any of several investigated factors. In 1985 all YHAP respondents underwent a followup face-to-face interview lasting approximately 1hour. The YHAP interview collected sociodemographic data in addition to information on chronic diseases, medications, depression, physical function, mental status, activities, alcohol and tobacco use, height, and weight. Eligible respondents who agreed to participate in a l-year falls substudy were administered a more detailed assessment by a trained nurse-researcher 1 to 3 weeks after the YHAP interview. This assessment collected information on self-reported musculoskeletal symptoms, dizziness, upper and lower extremity disability, mobility, fear of falling, and recent history of falling. The examination involved sitting-to-standing blood pressure change, vision and hearing, a standardized ll-item balance and 9-item gait assessment in which abnormalities were scored as present or absent, a thorough neuromuscular assessment, and a foot examination. Definition of Variables Most data were collected using standard instruments. Depression was defined as scores of 16 or greater on the Center for Epidemiologic Studies-Depression (CES-D) scale;” cognitive impairment was considered present if the subject made at least five errors on the Short Portable Mental Status Questionnaire (SPMSQ).13Using a Guttman scale for functional disability described elsewhere,’* we defined disability as gross movement im~airrnent’~ or disability with an activity of daily living.I6 Sedative medications mentioned by subjects (and verified by interviewers) included benzodiazepines, phenothiazines, and antidepressants. Vision impairment was defined as greater than 20% bilateral 10ss.I~Body mass index was the weight in kilograms divided by the square of the height in meters.” Estimates of ethanol consumption were derived from self-reports of use of beverage alcoh01.’~ Outcome Incident falls and surrounding circumstances were ascertained during telephone interviews conducted every 2 months. Falls due to intrinsic events such as stroke or syncope ( n = ll), or due to overwhelming extrinsic hazards such as being hit by a car that would cause a fall in young healthy persons (n =
FALL-RELATED INJURIES
47
10) were excluded. Recall was aided through the use of a fall diary which was reviewed and left with the subject or with a designated proxy respondent in cases where the nurse doubted the subject’s reliability. Serious injury was defined as any fracture or soft-tissue injuries requiring medical attention or resulting in activity restriction for more than 48 hours. Subjects who fell were asked about the circumstances and activities preceding the fall. Activities were categorized in terms of mild, moderate, or marked displacement of the center of gravity from the subject’s base of support by consensus of three physical therapists. Information was collected on 222 of 272 incident falls during the l-year follow-up, but we are concerned here mainly with information on the 108 first, or index, falls among subjects who fell because this event was closest in time to the risk factor assessment. Risk factors had similar associations for falls occurring early and later in the study period. Subjects were also asked about intervening illnesses or medication changes during the bimonthly telephone interviews. Defining Frailty and Vigorousness Many disabilities are at least moderately prevalent in elderly persons, yet not all are predictive of falling. Defining frailty in terms of a simple count of disabilities ignores both clustering that exists among these factors and the differential risk posed by these factors across the spectrum of functional ability. We used Principal Components Analysis (PCA)” to examine intercorrelations among characteristics and disabilities. This is not a conventional use of PCA. Earlier we had attempted to derive an a priori categorization using subjects’ balance and gait scores which resulted in findings that were often contradictory and difficult to interpret. For this reason, we used the present technique in what we emphasize must be considered an exploratory manner. An explanation of the statistical operations involved is beyond the scope of the present paper. The SAS/ STATz0manual has an excellent discussion of this and related methods together with suggestionsfor extended reading. To analyse these data we used PROC FACTOR, Methods = Principal, which standardizes variables originally based on different measurement scales. The eighteen variables fed into the initial PCA represented demographic, physical, psychological, social, and functional characteristics identified during the baseline interview and assessment. The program found statistical evidence for seven principal components, but we report loadings only for components One and Two because our original interest was to see if the two largest principal components (which together account for 23% of the observed variance in the data set) were descriptive of frailty and vigor. (We omit the tables of intercorrelation and the total factor structure to save space but will gladly furnish them on request).
48
SPEECHLEY AND TINETTI
The second step in the analysis consisted of deciding which of the 18 variables to include in the definition of frailty and vigor since not all loaded heavily on the first two components, and patterns of loading were highly variable. All variables were coded with ascending values reflecting greater disability. (For race and sex, white and female were coded high). Thus, in an aged population with extensive disability, those characteristics most descriptive of frailty were expected to appear in the first component as positive values, with the degree of correlation they share with the component Frailty indicated by the distance of the loading above zero. Characteristics describing vigorousness would appear in the second component as negative values, since the second component is the “contrast” component and is calculated to be non-correlated with the first component. In this case the size of each loading indicates the correlation of that variable with the component Vigor. While any cutpoint in the loadings is arbitrary, we took values beyond f . 3 5 as suggesting a meaningful contribution to a principal component because this is the lower range of moderate coefficients in conventional correlation analysis and because it was a compromise between 0.5, which would exclude all but a few variables, and .25, which would include all but a few. As seen in Table 1, nine variables met the criterion for frailty with values larger than .35: age over 80, balance and gait abnormalities, infrequent walking for exercise, depressed, taking sedatives, decreased strength in shoulder, decreased strength in knee, lower extremity disability, and near vision loss. Suggestive of vigor are four characteristics in Component Two that contrast strongly with the frailty component (values smaller than -.35): age under 80, cognitively intact, frequent physical exercise other than walking, and relatively good near vision. Not surprisingly in a population over 75, some disability will be seen even among the most vigorous, as indicated by the presence (ie moderately large positive values) of dizziness, lower extremity disability, and foot problems. Because these are positive values, we excluded them from the definition of vigor, but discuss their possible meaning below. Subjects could have between 0 and 9 frail attributes and between 0 and 4 vigorous attributes. We conceived frailty as having many frail attributes and few vigorous attributes and conceived vigor as the opposite. We operationalized this by arraying the scores of the two distributions in a five by nine table (possible (observed) ranges of scores were 0-9 (0-8) for frail and 0-4 (0-4) for vigor). We maximized discrimination by considering subjects as frail only if they possessed at least four frail attributes (high frail) and no more than one vigorous attribute (low vigorous). Conversely, vigorous subjects needed to posses at least three vigor attributes and two
IAGS-JANUARY 1991-VOL. 39, NO. I
TABLE 1. PRINCIPAL COMPONENT ANALYSIS LOADINGS FOR THE FIRST TWO PRINCIPAL COMPONENTS* Principal Components Characteristics
Onet
Two*
,391 .338 .080
-.414 ,315 .231
,262 ,493
-.453 .191
.748 .352 .010
-.143 -.160 -.416
.244
-.242
Sociodemographic Age 2 80 years Female White Psychological Functioning Cognitive impairment5 Depressed Health and Functioning Balance and gait abnormalities Walk for exercise rarely or never Other physical exercise rarely or never Sick in bed during follow-up Medications Sedatives4 Physical Symptoms or Impairments Any dizziness Decreased shoulder strength Any lower extremity disability Decreased knee strength Serious foot problems Palmomental reflex present Near vision loss 2 20%4 Postural blood Dressure &OD 2 10% * See Methods for descriptions. t Loading > .35 taken as descriptive of Frailty.
,350
,097
.218 .491 .421 .535 .214 .289 .470 .122
.451 -.168 .347 .137 ,350 -.073 -.367 .167
$ Loading < -.35 taken as descriptive of Vigor. 5 See Methods for definition.
or fewer frail attributes. These criteria were based on simultaneously falling beyond the median values on both distributions (the median value is the value held by the person at the 50th percentile and is not necessarily the middle number in the range of scores. Median values were 3 for frail, 2 for vigor). For example, to be frail one had to have a score above the median value for frail (ie values above 3) and below the median for vigor (ie below 2). All subjects not meeting these criteria for frailty or vigor were assigned to the transition groupThe hypothesis that the prevalences of disabilities would be highest in the Frail group and lowest in the Vigorous group was tested using the chi-square test for trends in proportions which was also used in the analysis of differences in proportions of those who fell and in the circumstances surrounding their index fall. This test may be invalid for cognitive impairment and sedatives, due to cell sizes of zero in the vigorous column. Nonetheless, the presence of zero cases of cognitive impairment and sedative use in the vigorous group, and clearly increasing prevalence in the transition and frailty groups, does not threaten the hypothesis under test.
FALL-RELATED INJURIES
IAGS-JANUARY 1991-VOL. 39, NO. 1
49
RESULTS
as likely as those in the Vigorous group to live in senior housing and to leave their neighborhood less than daily Sixty-seven subjects met criteria for frailty based on and about 3 times as likely to be functionally disabled, simultaneous presence of four or more frail descriptors based on a Guttman scale of self-reported items. Those and one or no vigorous items. Conversely, 87 subjects in the Vigorous group are over twice as likely as the met criteria for vigorousness. The remaining 182 sub- frail subjects to drink alcohol regularly. The percentage jects were considered transitional, possessing charac- of subjects who fell during the year of followup interistics of both vigor and frailty. creased from 17% in the vigorous to 32% in the tranPrevalences of several characteristics, listed in Table sitional to 52% in the frail indicating, as expected, a 2, illustrate the differences among the three groups. strong relationship between frail-vigorous status and The resulting trends in prevalence were of course ex- risk of falling (P < .001, Chi-square test for trend). pected for the variables used for the original allocation Table 3 lists differences in the circumstances and into Frail and Vigorous groups. One test of the ade- consequences of the falls among the 108 persons who quacy of our categorization scheme, however, was to fell within the 1-year follow-up period. Compared with see if trends in prevalence among those variables not falls among the frail, falls in the Vigorous group were originally used increased linearly across the groups. For significantly more likely to occur on stairs and away example, the subjects in the Frail group are about twice from home. While not statistically significant, the
TABLE 2. DISTRIBUTION OF SUBJECT CHARACTERISTICS BY FRAIL, TRANSITIONAL OR VIGOROUS STATUS* Characteristics
Frail n = 67
Transitional Vigorous n = 182 n = 87
%
Sociodemographic Age 5 80 yearst 99 Mean age SD 86.3 (4.5) Female 60 White 91 Living in housing 24 for the elderly Psychological functioning Cognitive impair- 21 ment*t Depressed* 25 Health and functioning Leaves neighbor- 85 hood
Falls and Injuries in Frail and Vigorous Cornmunitv Elderlv Persons J
J
Mark Speechley, PhD,* and Mary Tinetti, M D t
Identification of different types of falls and fallers among elderly persons might aid in the targeting of preventive efforts. In a representative sample of 336 community elderly, subjects were assigned to Frail, Vigorous, or Transition groups based on observed patterns of clustering among demographic, physical, and psychological variables. The frequency and circumstances of falls in these three groups were then ascertained. As expected, the observed incidence of falling in one year of follow-up was highest in the Frail group (52%) and lowest in the Vigorous group (17%). However, 22% (5123) of falls by
vigorous subjects, but only 6% (5189) of falls by frail subjects, resulted in a serious injury. Compared with frail subjects, vigorous fallers were somewhat more likely to fall during displacing activity (53% vs 31%), with an environmental hazard present (53% U S 29%, and on stairs (27% v s 6%). These findings suggest that fallrelated injuries can be a serious health problem for vigorous as well as frail elderly persons. Injury prevention, therefore, should be directed at all elderly persons but tailored to expected differences in fall circumstances. J
people age they become more diverse with increasing variability in their habits, life experiences, diseases, and disabilities. This heterogeneity among elderly persons can render assessment and treatment of health problems difficult. An additional challenge occurs when the health problem or event of interest has a multifactorial etiology. A fall is an example of a multifactorial event that results from interplay among intrinsic, situational, and environmental factors.’ Investigators have identified multiple chronic diseases and disabilities, such as cognitive impairment, upper and lower extremity disabilities, arthritis, visual impairment, and gait disorders as This strong association bepredisposing to tween chronic disabilities and falling may lead to the conclusion that intrinsic factors predominate in fall etiology and that falling is a health problem only for frail elderly persons. We recently found, however, that 15% of fallers in a representative sample of community
elderly persons had no more than one chronic intrinsic fall risk factor and would not be considered frail.’ Situational and environmental factors may play a great role among these more vigorous elderly fallers. It is important to identify the prevalence, circumstances, and sequelae of falls across the spectrum of elderly persons so that clinical assessment and preventive efforts can be targeted appropriately. A typology based on differences in characteristics of frail and vigorous elderly persons and on circumstances of their falls, would help in developing these clinical assessments and prevention programs. A central problem in constructing such a typology is the definition of Frailty and Vigor. A simple count of disabilities ignores both clustering among characteristics and the fact that not all disabilities predict fall likelihood. One solution is to examine the clustering among a wide array of characteristics to see if characteristics that are correlated statistically may be interpretable in a way that is clinically meaningful and useful. The aim of this study was to identify the prevalence, circumstances, and sequelae of falls across the functional spectrum of elderly persons. We accomplished this aim by first separating subjects into functional groups based on intercorrelations among 18 characteristics and then reviewing the fall situations of subjects in each group.
A
From the *Department of Epidemiology and Biostatistics, University of Western Ontario, London, Ontario, Canada; and the tDepartment of Medicine, Yale University School of Medicine, New Haven, Connecticut. Dr. Tinetti is a recipient of an NIA Academic Award (K08AG00292). This work was conducted while Dr. Speechley was a postdoctoral trainee at the Yale Health and Aging Project, one of four sites under the National Institute of Aging EPESE program. Address correspondence to Mark Speechley, PhD, Department of Epidemiology and Biostatistics, Kresge Building, The University of ’-In n t a n n I.ondon, Ontario, N6A 5C1 Canada.
.”
Am Geriatr SOC39:46-52, 1991
METHODS Description of Sample The sample in the present study has been described in detail elsewhere.’ Briefly, 0002-8614/91/$3.5(
IAGS-IANUARY 1991-VOL. 39, NO. 1
the sample is a subset of the Yale Health and Aging Project (YHAP) cohort from the New Haven site of the Established Populations for Epidemiologic Study of the Elderly (EPESE)program which numbered 2,812 when assembled in 1982.” The 336 persons in our study were selected from the 1,762 persons still living in 1985 who met the following eligibility criteria: age at least 75 years; living in a private residence or one of two designated elderly housing units; able to follow simple commands; able to walk within their homes. Of the 458 eligble subjects, 114 (25%) refused to participate and eight (2%) could not be found. Participants did not differ from eligible non-participants on any of several investigated factors. In 1985 all YHAP respondents underwent a followup face-to-face interview lasting approximately 1hour. The YHAP interview collected sociodemographic data in addition to information on chronic diseases, medications, depression, physical function, mental status, activities, alcohol and tobacco use, height, and weight. Eligible respondents who agreed to participate in a l-year falls substudy were administered a more detailed assessment by a trained nurse-researcher 1 to 3 weeks after the YHAP interview. This assessment collected information on self-reported musculoskeletal symptoms, dizziness, upper and lower extremity disability, mobility, fear of falling, and recent history of falling. The examination involved sitting-to-standing blood pressure change, vision and hearing, a standardized ll-item balance and 9-item gait assessment in which abnormalities were scored as present or absent, a thorough neuromuscular assessment, and a foot examination. Definition of Variables Most data were collected using standard instruments. Depression was defined as scores of 16 or greater on the Center for Epidemiologic Studies-Depression (CES-D) scale;” cognitive impairment was considered present if the subject made at least five errors on the Short Portable Mental Status Questionnaire (SPMSQ).13Using a Guttman scale for functional disability described elsewhere,’* we defined disability as gross movement im~airrnent’~ or disability with an activity of daily living.I6 Sedative medications mentioned by subjects (and verified by interviewers) included benzodiazepines, phenothiazines, and antidepressants. Vision impairment was defined as greater than 20% bilateral 10ss.I~Body mass index was the weight in kilograms divided by the square of the height in meters.” Estimates of ethanol consumption were derived from self-reports of use of beverage alcoh01.’~ Outcome Incident falls and surrounding circumstances were ascertained during telephone interviews conducted every 2 months. Falls due to intrinsic events such as stroke or syncope ( n = ll), or due to overwhelming extrinsic hazards such as being hit by a car that would cause a fall in young healthy persons (n =
FALL-RELATED INJURIES
47
10) were excluded. Recall was aided through the use of a fall diary which was reviewed and left with the subject or with a designated proxy respondent in cases where the nurse doubted the subject’s reliability. Serious injury was defined as any fracture or soft-tissue injuries requiring medical attention or resulting in activity restriction for more than 48 hours. Subjects who fell were asked about the circumstances and activities preceding the fall. Activities were categorized in terms of mild, moderate, or marked displacement of the center of gravity from the subject’s base of support by consensus of three physical therapists. Information was collected on 222 of 272 incident falls during the l-year follow-up, but we are concerned here mainly with information on the 108 first, or index, falls among subjects who fell because this event was closest in time to the risk factor assessment. Risk factors had similar associations for falls occurring early and later in the study period. Subjects were also asked about intervening illnesses or medication changes during the bimonthly telephone interviews. Defining Frailty and Vigorousness Many disabilities are at least moderately prevalent in elderly persons, yet not all are predictive of falling. Defining frailty in terms of a simple count of disabilities ignores both clustering that exists among these factors and the differential risk posed by these factors across the spectrum of functional ability. We used Principal Components Analysis (PCA)” to examine intercorrelations among characteristics and disabilities. This is not a conventional use of PCA. Earlier we had attempted to derive an a priori categorization using subjects’ balance and gait scores which resulted in findings that were often contradictory and difficult to interpret. For this reason, we used the present technique in what we emphasize must be considered an exploratory manner. An explanation of the statistical operations involved is beyond the scope of the present paper. The SAS/ STATz0manual has an excellent discussion of this and related methods together with suggestionsfor extended reading. To analyse these data we used PROC FACTOR, Methods = Principal, which standardizes variables originally based on different measurement scales. The eighteen variables fed into the initial PCA represented demographic, physical, psychological, social, and functional characteristics identified during the baseline interview and assessment. The program found statistical evidence for seven principal components, but we report loadings only for components One and Two because our original interest was to see if the two largest principal components (which together account for 23% of the observed variance in the data set) were descriptive of frailty and vigor. (We omit the tables of intercorrelation and the total factor structure to save space but will gladly furnish them on request).
48
SPEECHLEY AND TINETTI
The second step in the analysis consisted of deciding which of the 18 variables to include in the definition of frailty and vigor since not all loaded heavily on the first two components, and patterns of loading were highly variable. All variables were coded with ascending values reflecting greater disability. (For race and sex, white and female were coded high). Thus, in an aged population with extensive disability, those characteristics most descriptive of frailty were expected to appear in the first component as positive values, with the degree of correlation they share with the component Frailty indicated by the distance of the loading above zero. Characteristics describing vigorousness would appear in the second component as negative values, since the second component is the “contrast” component and is calculated to be non-correlated with the first component. In this case the size of each loading indicates the correlation of that variable with the component Vigor. While any cutpoint in the loadings is arbitrary, we took values beyond f . 3 5 as suggesting a meaningful contribution to a principal component because this is the lower range of moderate coefficients in conventional correlation analysis and because it was a compromise between 0.5, which would exclude all but a few variables, and .25, which would include all but a few. As seen in Table 1, nine variables met the criterion for frailty with values larger than .35: age over 80, balance and gait abnormalities, infrequent walking for exercise, depressed, taking sedatives, decreased strength in shoulder, decreased strength in knee, lower extremity disability, and near vision loss. Suggestive of vigor are four characteristics in Component Two that contrast strongly with the frailty component (values smaller than -.35): age under 80, cognitively intact, frequent physical exercise other than walking, and relatively good near vision. Not surprisingly in a population over 75, some disability will be seen even among the most vigorous, as indicated by the presence (ie moderately large positive values) of dizziness, lower extremity disability, and foot problems. Because these are positive values, we excluded them from the definition of vigor, but discuss their possible meaning below. Subjects could have between 0 and 9 frail attributes and between 0 and 4 vigorous attributes. We conceived frailty as having many frail attributes and few vigorous attributes and conceived vigor as the opposite. We operationalized this by arraying the scores of the two distributions in a five by nine table (possible (observed) ranges of scores were 0-9 (0-8) for frail and 0-4 (0-4) for vigor). We maximized discrimination by considering subjects as frail only if they possessed at least four frail attributes (high frail) and no more than one vigorous attribute (low vigorous). Conversely, vigorous subjects needed to posses at least three vigor attributes and two
IAGS-JANUARY 1991-VOL. 39, NO. I
TABLE 1. PRINCIPAL COMPONENT ANALYSIS LOADINGS FOR THE FIRST TWO PRINCIPAL COMPONENTS* Principal Components Characteristics
Onet
Two*
,391 .338 .080
-.414 ,315 .231
,262 ,493
-.453 .191
.748 .352 .010
-.143 -.160 -.416
.244
-.242
Sociodemographic Age 2 80 years Female White Psychological Functioning Cognitive impairment5 Depressed Health and Functioning Balance and gait abnormalities Walk for exercise rarely or never Other physical exercise rarely or never Sick in bed during follow-up Medications Sedatives4 Physical Symptoms or Impairments Any dizziness Decreased shoulder strength Any lower extremity disability Decreased knee strength Serious foot problems Palmomental reflex present Near vision loss 2 20%4 Postural blood Dressure &OD 2 10% * See Methods for descriptions. t Loading > .35 taken as descriptive of Frailty.
,350
,097
.218 .491 .421 .535 .214 .289 .470 .122
.451 -.168 .347 .137 ,350 -.073 -.367 .167
$ Loading < -.35 taken as descriptive of Vigor. 5 See Methods for definition.
or fewer frail attributes. These criteria were based on simultaneously falling beyond the median values on both distributions (the median value is the value held by the person at the 50th percentile and is not necessarily the middle number in the range of scores. Median values were 3 for frail, 2 for vigor). For example, to be frail one had to have a score above the median value for frail (ie values above 3) and below the median for vigor (ie below 2). All subjects not meeting these criteria for frailty or vigor were assigned to the transition groupThe hypothesis that the prevalences of disabilities would be highest in the Frail group and lowest in the Vigorous group was tested using the chi-square test for trends in proportions which was also used in the analysis of differences in proportions of those who fell and in the circumstances surrounding their index fall. This test may be invalid for cognitive impairment and sedatives, due to cell sizes of zero in the vigorous column. Nonetheless, the presence of zero cases of cognitive impairment and sedative use in the vigorous group, and clearly increasing prevalence in the transition and frailty groups, does not threaten the hypothesis under test.
FALL-RELATED INJURIES
IAGS-JANUARY 1991-VOL. 39, NO. 1
49
RESULTS
as likely as those in the Vigorous group to live in senior housing and to leave their neighborhood less than daily Sixty-seven subjects met criteria for frailty based on and about 3 times as likely to be functionally disabled, simultaneous presence of four or more frail descriptors based on a Guttman scale of self-reported items. Those and one or no vigorous items. Conversely, 87 subjects in the Vigorous group are over twice as likely as the met criteria for vigorousness. The remaining 182 sub- frail subjects to drink alcohol regularly. The percentage jects were considered transitional, possessing charac- of subjects who fell during the year of followup interistics of both vigor and frailty. creased from 17% in the vigorous to 32% in the tranPrevalences of several characteristics, listed in Table sitional to 52% in the frail indicating, as expected, a 2, illustrate the differences among the three groups. strong relationship between frail-vigorous status and The resulting trends in prevalence were of course ex- risk of falling (P < .001, Chi-square test for trend). pected for the variables used for the original allocation Table 3 lists differences in the circumstances and into Frail and Vigorous groups. One test of the ade- consequences of the falls among the 108 persons who quacy of our categorization scheme, however, was to fell within the 1-year follow-up period. Compared with see if trends in prevalence among those variables not falls among the frail, falls in the Vigorous group were originally used increased linearly across the groups. For significantly more likely to occur on stairs and away example, the subjects in the Frail group are about twice from home. While not statistically significant, the
TABLE 2. DISTRIBUTION OF SUBJECT CHARACTERISTICS BY FRAIL, TRANSITIONAL OR VIGOROUS STATUS* Characteristics
Frail n = 67
Transitional Vigorous n = 182 n = 87
%
Sociodemographic Age 5 80 yearst 99 Mean age SD 86.3 (4.5) Female 60 White 91 Living in housing 24 for the elderly Psychological functioning Cognitive impair- 21 ment*t Depressed* 25 Health and functioning Leaves neighbor- 85 hood
