Aging 3: 219-228, 1991
ORIGINAL ARTICLES
Prevalence, characteristics and factors associated with falls among the elderly living at home J.H. Downton and K. Andrews University Department of Geriatric Medicine, Hope Hospital, Salford, Manchester, U.K. ABSTRACT. Attempts to determine the underlying causes of falls have come to conflicting conclusions, partly because subject groups studied have not been representative of all elderly people. Two hundred and three randomly selected people of 75 years and over, living at home, were visited and questioned about falls experienced in the previous 12 months, and about factors that might be related to falling. Eighty-six subjects (42.4%) had suffered one or more falls during this time, and of fallers, 49 (59.3%) were injured, 9 of them seriously. Women were slightly more likely to have had falls and were more likely to have suffered injury, but no increase in frequency of falls with age was demonstrated. Only a minority of fallers (43.0%) sought medical attention following their fall. Falls outside the home accounted for 39.5% of falls and these were more likely to be due to simple trips or slips. Analysis of the factors related to falls showed a considerable overlap between fallers and non-fallers. Fallers had significantly greater dependency and cognitive impairment, more physical symptoms, and higher scores for anxiety and depression, but there was no association with postural hypotension, neurological abnormalities, or measurements relating to nutritional state. The factors found to be significant on discriminant analysis were combined to determine a "fall risk score". This type of easily calculated score might be of use to
medical and paramedical personnel for assessing the risk of falling among the elderly living at home. (Aging 3: 219-228, 1991)
INTRODUCTION Falls and their complications are recognised to be major health problems for elderly people. However, despite numerous studies over many years, the epidemiology of the problem is still unclear. Falls are reported to occur in 23% (1) to 60% (2) of old people; these discrepancies reflect the difficulties of studying the problem, ranging from defining what is a fall, to counting how often they occur. The study of falls is also open to much subjective interpretation and this has resulted in sometimes confusing classifications of falls, fallers, and events surrounding the falls (3). Why do old people fall? Humans have an intrinsic liability to fall because of their upright posture, and this liability seems to increase with age. This may be a reflection of neurological changes with age, gait and postural changes, medical and psychological ill health, drugs, or a combination of these and other factors; nonetheless, despite many studies, the fundamental "cause" of increased liability to falls in the elderly has not been identified. The most likely reason for this failure is that falls are multifactorial in
Key words: Assessment, elderly, falls, injury_ Correspondence: Dr. J.H. Downton, Department of Geriatric Medicine, St Thomas' Hospital, Stockport, SK3 8BL, United Kingdom_ Received August 27,1989; accepted September 30,1990.
Aging, Vol. 3, N. 3 219
J.H. Downton and K. Andrews
origin, and, in addition, potential factors may be inter-related. However, previous studies can be difficult to interpret because their findings are confounded by methodological problems. Physiological and pathological factors underlying falls are likely to vary depending on the group of elderly studied. One group for whom there has been a lack of clear data is old people at home. These account for the majority of the elderly, but are in many ways the most difficult to study, unlike the "captive population" of people in institutional care. We have therefore attempted to select a representative sample of old people living at home to try and elucidate some of these problems.
SUBJECTS AND METHODS Subjects were selected from the computerised age-sex registers of five General (Family) Practitioners within Salford District Health Authority. Patients aged 75 years and over, excluding those living in residential care, were considered as potential study subjects. In order to include a sufficient number of very elderly, the selection was stratified: all patients aged 85 and over were included in the study; and of those between 75 and 84 years, one in four was chosen by random number to be included. Interviews took place between November 1986 and May 1988. All subjects were visited at home by one investigator (J.HD) and a standard questionnaire was administered. This covered: social and demographic data; use of social services (home help, meals on wheels, etc); mobility and frequency of trips out; medical symptoms; continence; dependency; depression (using the Schwab Depression Inventory) (4); cognitive function (using the Information/Orientation component of the Clifton Assessment Schedule - CAS) (5); anxiety (using the General Health Questionnaire) (6); alcohol intake; past medical history; and any drugs taken regularly. Subjects were questioned about falls experienced in the 12 months prior to interview; a fall was defined as an event in which the subject inadvertently came to rest on the floor. This definition therefore excluded near-falls and falls onto bed or chair. Those who reported falls were classified as "fallers" and those who did
220 Aging, Vol. 3, N. 3
not report any falls in the previous 12 months were classified as "non-fallers". If any falls had occurred, more detailed information was collected about the most recent fall and the circumstances in which it occurred, about any injury, and about any request for medical attention from their general practitioner (GP) or from hospital. Injuries were classified as "trivial" (bruising or laceration not requiring sutures) or "serious" (fracture or laceration requiring sutures). Subjects were also asked about whether they felt well at the time of the fall, and whether they experienced any symptoms of "dizziness" on standing (postural dizziness). An attempt was made to classify the fall. If the subject was clearly able to describe a trip or slip, the fall was classified as a "trip" fall. Symptoms were sought which might indicate a specific cause for the fall (e.g., micturition syncope, postural hypotension etc). If neither of these situations were obtained, the fall was classified as "non-specific". Examination included pulse, sitting blood pressure and drop in systolic blood pressure one and two minutes after s:anding, visual acuity (using a modified Snellen chart), a limited neurological examination (biceps and patellar reflexes and plantar responses), and measurement of passive range of movement at knee and ankle joints using a goniometer. Nutritional status was assessed by measuring triceps skinfold thickness, and circumference of the upper arm midway between the shoulder and the elbow. These measurements were used to derive the arm muscle circumference and the arm muscle area (7).
Assessment of activities of daily living was based on the Katz score (8), with 0 being fully independent, and 6 being completely dependent in the activities tested. Knee and ankle joint restriction measures were the deficits of right and left sides added together (assuming that full range of movement was 130° for each knee and 80° for each ankle). Other scores used during the analysis are shown in the Appendix. Results were analysed using the extended version of the Statistical Package for the Social Sciences (SPSS-X). The scores used were not normally distributed, 'and comparisons were therefore made using Mann Whitney U - Wil-
Falls in the elderly at home
coxon Rank Sum test. Categorical variables were compared using the X2 test with Yate's correction. Stepwise discriminant analysis was performed with fall group (i.e., faller or nonfaller) as the dependent variable, by transforming non-normally distributed scores into die hot omous categorical scores (e.g., trips out more frequently than weekly or not) which gave the best discrimination between fallers and nonfallers. Discriminant analysis was carried out for the whole group, and for men and women separately. The factors shown to be the best "predictors" of falling were then used to produce a "fall risk score".
RESULTS Of the potential study sample of 286 subjects 9 could not be traced and 10 refused to b~ interviewed. Two subjects were excluded because their GP felt they were not appropriate subjects (one was moribund and the other was deaf arid aphasic). A further 62 could not be included because by the time they were due to be interviewed they had died (33), moved away (12), or had mov~d into residential care (17). These were predominantly in the 85+ age grqup. The sample studied therefore consisted of 203 people aged 75 and over, of whom 143 (70.4%) were women and 60 (29.6%) were men. The mean age of the whole sample was 83.0 years (SO 5.0, range 75-97). Demographic data and comparison with those of 75 and over interview.ed for the General Household Survey (GHS) In 1980 (9) are shown in Table 1. Mobility was unrestricted in 103 subjects (50.7%), and a further 54 (26.6%) were mobile around their immediate neighbourhood. Only 46 (22.7%) had more restricted mobility. This is very similar to the GHS sample in which 22.2% were unable to get out of the house without help (9). Seventysix subjects (37.4%) had a home help, 10 (4.9%) had meals on wheels and 14 (6.9%) had district nurse' assistance with bathing. Only 18/203 (9.0%) scored seven points or less on the CAS information/ orientation test (the level taken as indicating significant cognitive impairment) (10). One hundred and thirty subjects (64.0%) were fully independent in activities of daily living. A further 61 (30.1%) required help only with bath-
Table 1 - Demographic data. This study Corrected for stratification
GHS 1980 (75+ years)
Women Single
70.4% 8.9%
69.9% 8.8%
Married Widowed
24.6% 66.0% 0.5%
27.2% 63.7% 0.3%
65.8% 10.1% 34.8% 54.2% 0.8%
Living alone 56.7% with spouse 2"1..7% with others 21.7%
55.3% 26.8% 17.9%
45.5% 30.1% 24.4%
Divorced
GHS = General Household Survey 1980 (9).
ing, leaving only 12 (5.9%) with varying degrees of more severe dependence. Forty-five subjects (22.2%) suffered occasional or frequent incontinence of urine. Eighty-six subjects (42.4%) had suffered one or more falls in the previous 12 months. The mean age of fallers was 83.6 years, and of nonfallers 82.5 years. This difference was not significant. Women were slightly more likely to be fallers than men (46.9% of women fell compared with 31.7% of men; X2 = 3.99, OF -'1, p = 0.046 before Yate's correction). Of those who had had a fall, 35/86 (40.7%) suffered no injury, 40 (46.5%) suffered a trivial injury, 2 subjects (2.3%) required sutures for lacerations, and 7 (8.1%) had fractures. (Two subjects could not remember whether they had injured themselves). Women were more likely to suffer injury than men: 64.6% of women fallers were injured com-
Table 2 - Time of falls. Time 06.00 - 12.00 12.00 - 18.00 18.00 - 00.00 00.00 - 06.00 Don't know
No. of falls (%) 22 (25.6) 47 10 4 3
(54.7) (11.6) (4.6) (3.5)
Aging, Vol. 3, N. 3
221
J.H. Downton and K. Andrews
Table 3 - Reason for falling. Reason
No. of falls (%)
Definite trip/slip Blackout Other specific "Dizziness" "Non-specific" Don't know
18 5 6 6 45 6
(20.9) (5.8) (7.0) (7.0) (52.3) (7.0)
pared with 36.8% of men ( X2 = 4.67, OF = 1, p = 0.03); and 7 out of the 9 serious injuries occurred in women. Most falls took place during the day, with 69 (80.3%) occurring between 6 a.m. and· 6 p.m. (Table 2). A substantial proportion of falls, 34/86 (39.5%), happened outside the home, mostly while out shopping, etc. Of falls inside the house, only small numbers occurred in the bathroom or on the stairs (3 and 2 falls respectively), the rest occurring mainly in the bedroom or living room. It was often difficult to classify why the fall had happened. Eighteen falls (20.9%) were due to a definite trip or slip, and in another 11 the history suggested a specific cause, such as cough syncope or epilepsy. In the majority of cases, subjects seemed to find themselves on the floor without being able to explain why it had happened, though some blamed "dizziness" (Table 3). Half the fallers were able to get up unaided, and only 6 were on the floor for longer than an hour.
Twenty-five fallers (29.1%) considered that they were in poor health at the time of their fall. This was not significantly associated with the type (trip or non-trip) or the place (inside or outside) of the fall. Slightly more outside than inside fallers suffered injury (50% us 33.3%), but this difference was not statistically significant. Outside fallers were more likely to have suffered trips than those who fell inside: 13 of the 34 outside fallers (38.2%) tripped compared with 5 of 49 (10.2%) inside fallers ( X2 = 6.6, OF = 1, p = 0.006). Of those who fell, 49/86 (57.0%) did not seek medical attention; the mean age of this group was the same as those who did (83.6 us 83.5 years). More men than women (41.2 % us 35.6%) sought medical attention, but this difference was not statistically significant. There was no association between the type of fall or place of fall and whether there was contact with GP or hospital. Comparisons of fallers with non-fallers (Tables 4-8) showed significant differences in CAS score, anxiety, depression, physical symptoms and knee joint restriction. Fallers were more likely to have an unstable knee joint, and to have had a previous stroke. There was no association between the number of drugs being taken, nor particular types of drugs (e.g., sedatives, diuretics etc) and a history of falls. A drop in systolic blood pressure on standing of 20 mmHg or more was found in 61 (31.3%) subjects, but postural hypotension was not associated with symptoms of postural dizziness, nor with falls. Results for the two age groups separately were similar.
Table 4 - Comparison of social factors in fallers and non-fallers.
Homehelp Meals on wheels Bath attendant Trips out less frequently than weekly ADL score (mean ± SD) Urinary incontinence Mobility score (mean ± SD) Alcohol consumption
222
Aging, Vol. 3, N. 3
Fallers
Non-fallers
p
42/86 (48.8%) 6/86 (7.0%) 7/86 (8.1%) 33/86 (38.4%) 0.62 ± 1.07 24/86 (27.9%) 1.14 ± 1.36 38/86 (44.2%)
34/117 (29.1%) 4/117 (3.4%) 7/117 (6.0%) 25/117 (21.4%) 0.45 ± 0.93 21/117 (17.9%) 0.77 ± 1.04 41/117 (35.0%)
0.006 NS NS 0.013 NS NS NS NS
Falls in the elderly at home
Table 5 - Comparison of cognitive and psychological factors in fallers and non-fallers. Fallers CAS score Anxiety score Depression score
Non-fallers
mean ±SD
p
10.5 ± 2.2 4.05 ± 3.22 8.95 ± 5.46
11.0 ± 1.2 2.91 ± 3.19 5.84 ± 5.91
0.035 0.0036 < 0.0001
Table 6 - Comparison of nutritional parameters in fallers and non-fallers. Fallers Triceps skinfold thickness (mm) Arm muscle circumference (cm) Arm muscle area (cm 2)
mean ±SD
16.1 ± 8.6 21.7 ± 2.2 38.0 ± 7.7
Non-fallers p
15.4 ± 8.8 21.9 ± 2.5 38.6 ± 8.9
NS NS NS
Table 7 - Comparison of medical history in fallers and non-fallers. Fallers Dyspnoea score (mean ± SD) Chest pain score (mean ± SD) "Arthritis" Postural dizziness Previous stroke Previous fracture Parkinson's disease Ischaemic heart disease Diabetes Number of drugs taken daily (mean ± SD)
0.99 ± 0.89 0.26 ± 0.56 51/86 (59.3%) 23/61 (37.7%) 14/86 (16.3%) 28/86 (32.6%) 3/86 (3.5%) 32/86 (37.2%) 7/86 (8.1%) 2.51 ± 2.11
Non·fallers 0.77 ± 0.88 0.08 ± 0.27 44/117 (37.6%) 32/95 (33.7%) 6/117 (5.1%) 26/117 (22.2%) 2/117 (1.7%) 37/117 (31.6%) 12/117 (10.3%) 2.08 ± 1.89
p
0.049 0.009 0.004 NS 0.017 NS NS NS NS NS
Table 8 - Comparison of medical examination in fallers and non-fallers. Fallers Systolic BP (mmHg) (mean ± SD) Systolic drop on standing (mmHg) (mean ± SD) Irregular pulse Visual acuity score (mean ± SD) Restriction of hip movement Knee instability Knee restriction (R + L) (mean± SD) Ankle restriction (R + L) (mean ± SD) Abnormal knee jerk Abnormal plantar response
179.7 ± 28.8 11.2 ± 11.3 23/86 (26.7%) 2.63 ± 1.78 8/86 (9.3%) 6/86 (7.0%) 71° ± 37° 81° ± 21° 21/85 (24.7%) 15/86 (17.4%)
Non-fallers 179.9 ± 34.2 13.9 ± 12.9 33/116 (28.4%) 2.28 ± 1.60 6/117 (5.1%) 1/117 (0.85%) 58° ± 26° 77° ± 21° 20/117 (17.1%) 14/115 (12.2%)
p
NS NS NS NS NS 0.049 0.012 NS NS NS
Aging, Vol. 3, N. 3 223
J.H. Downton and K. Andrews
Stepwise discriminant analysis was able to classify correctly only 71.8% of cases as "fallers" or "non-fallers" when all subjects were considered. Classification was less effective if women alone were considered (65.7% correctly classified), and better if men alone were considered (88.1% correctly classified). The discriminating factors ("predictors"), their discriminant function coefficients and the percentage each factor contributes to the discrimination are shown in Table 9. It can be seen that the "predictors" of falling and their relative importances are different for men and women. For all subjects, mobility has a negative coefficient meaning that poor mobility "protects", against falling. Living alone (rather than with spouse or relatives) is a "risk factor" for men but is "protective" for women. When these "predictors" were used to calculate a "fall risk" score, fallers and nonfallers had highly significant differences in scores, but there was a wide overlap in scores between fallers and non-fallers (Table 10). If a score of 5 or more is taken as indicating a subject who is likely to fall, then the score has
a specificity and sensitivity of 69%. The predictive value of a "positive" score (i.e., ~ 5) is 61.7% and that of a "negative" score (i.e., ~ 4) is 75.5%.
DISCUSSION Any attempt to study old people in the community is beset with problems. The relatively high geographical mobility of old people (ll) and the fact that there is an inevitable mortality means that it is difficult to ensure a "typical" sample of community elderly, particularly if data collection takes more than a short time. There are also differences in social mix and in proportions of elderly between different areas of the country (and also between countries). This study has attempted to look at a representative selection of people aged 75 and over living in their own homes by using the computerised age-sex registers of five local GPs. The likely attrition of the oldest subjects was taken into account by stratifying the selection to include sufficient numbers of very elderly.
Table 9 - Stepwise discriminant analysis: discriminant junction coefficients. Predictors Poor mobility T rips out less frequently than weekly Depression Previous stroke Symptoms of arthritis Female sex Meals on wheels Unstable knees Alcohol consumption Visual acuity worse than 6/9 Systolic BP Homehelp Urinary incontinence Lives alone Fear of falls Bath nurse Dizziness Abnormal kneejerk Correctly classified
224 Aging, Vol. 3, N. 3
All subjects
Women
·0.60 (14.7%)
·0.73 (20.9%)
0.54 (13.1%)
0.50 (14.1%) 0.30 (8.2%) 0.30 (8.6%) 0.49 (13.9%)
0.50 (12.3%) 0.47 (11.4%) 0.37 (8.9%) 0.35 (8.6%) 0.33 (7.9%) 0.29 (7.0%) 0.24 (5.8%) 0.24 (5.8%) 0.19 (4.6%)
0.32 (9.0%)
·0.38 0.56 0.69 0.95
(5.9%) (8.7%) (10.7%) (14.8%)
0.31 (4.8%) 0.87 (13.6%) 0.41 (6.4%)
0.40 (11.3%) 0.25 (7.2%) ·0.42 (6.9%)
71.8%
, Men
65.7%
0.33 (5.2%) 0.72 (11.2%) ·0.54 (8.4%) 0.34 (5.4%) ·0.31 (4.8%) 88.1%
Falls in the elderly at home
Table 10 - Fall risk ~coresfor fallers and non-fallers. Fall risk score
Fallers
Non-fallers
1 2 3
0 0 4 22 28 27 2 0
1 10 32 37 28 7 1 0 0
4 5 6 7 8 9
x2 = 45.7, DF = 8, p < O.QOOOL Compared with data from the General Household Survey (9), the study sample has a higher percentage of widow(er)s and larger numbers living alone, partly reflecting the larger proportion of very elderly resulting from the stratification procedure. (In the population of the practices used, people of 85 and over made up 25% of over-75s whereas they made up 47% of the eventual study sample.) Because of the difficulties in selecting a truly random sample, and the dependence of information on the subject's memory for events that have occurred a variable amount of time earlier, any extrapolations from a study of this type have to be made cautiously. In addition, some findings depended on a subjective assessment by the interviewer (for example, the reason for the fall). However, since all subjects were seen by a single interviewer, the problem of multiple, different subjective judgements was avoided. One of the aims of the study was to cover as many as possible of the factors which might be associated with falling. In order to maintain the good will of the subjects during a fairly lengthy interview, assessment of individual factors, and particularly the number of "invasive" assessments, necessarily had to be limited. This meant that, for example, neurological examination and assessment of orthostatic hypotension were not carried out in an ideal fashion. However, the very low refusal rate means that the results are more likely to be generalisable to other community populations.
The characteristics of the sample demonstrate that the stereotype of old people at home being frail, confused and requiring care is misleading; most subjects were mobile, independent and mentally clear. Despite this, falls were very common, affecting almost half of the subjects during the 12 months prior to interview. This study confirms that women seem more likely to fall than men (12-14) but it does not show the expected increasing incidence of falls with age though the selection was stratified to include adequate numbers of very elderly. It is possible that over the course of a study of 18 months duration there was selective loss of the frailest (and perhaps most likely to fall) elderly through death or entry into residential care. However, Sheldon (12) collected data over a similar length of time and found a very marked increasing prevalence of falling with increasing age, though he studied relatively few very elderly. Injuries due to falls were also comr:non, though not usually serious. The figures quoted here are almost certainly an underestimate because only the most recent fall was considered when enquiring about injury. Women were more likely to suffer injury, and although the larger number of fractures can be explained by the higher prevalence of osteoporosis in women, the reason for the higher number of injuries overall is not clear. It may reflect different thresholds of concern about the consequences of falls, since it could be interpreted as a subjective judgement of how serious the fall was. However, more men than women sought medical attention after the index fall. It proved extremely difficult to classify why a fall had occurred. A small proportion of subjects gave a clear story of a trip or slip (such as tripping over uneven paving stones or slipping on ice), but more commonly the description of the circumstances was very vaguean
ORIGINAL ARTICLES
Prevalence, characteristics and factors associated with falls among the elderly living at home J.H. Downton and K. Andrews University Department of Geriatric Medicine, Hope Hospital, Salford, Manchester, U.K. ABSTRACT. Attempts to determine the underlying causes of falls have come to conflicting conclusions, partly because subject groups studied have not been representative of all elderly people. Two hundred and three randomly selected people of 75 years and over, living at home, were visited and questioned about falls experienced in the previous 12 months, and about factors that might be related to falling. Eighty-six subjects (42.4%) had suffered one or more falls during this time, and of fallers, 49 (59.3%) were injured, 9 of them seriously. Women were slightly more likely to have had falls and were more likely to have suffered injury, but no increase in frequency of falls with age was demonstrated. Only a minority of fallers (43.0%) sought medical attention following their fall. Falls outside the home accounted for 39.5% of falls and these were more likely to be due to simple trips or slips. Analysis of the factors related to falls showed a considerable overlap between fallers and non-fallers. Fallers had significantly greater dependency and cognitive impairment, more physical symptoms, and higher scores for anxiety and depression, but there was no association with postural hypotension, neurological abnormalities, or measurements relating to nutritional state. The factors found to be significant on discriminant analysis were combined to determine a "fall risk score". This type of easily calculated score might be of use to
medical and paramedical personnel for assessing the risk of falling among the elderly living at home. (Aging 3: 219-228, 1991)
INTRODUCTION Falls and their complications are recognised to be major health problems for elderly people. However, despite numerous studies over many years, the epidemiology of the problem is still unclear. Falls are reported to occur in 23% (1) to 60% (2) of old people; these discrepancies reflect the difficulties of studying the problem, ranging from defining what is a fall, to counting how often they occur. The study of falls is also open to much subjective interpretation and this has resulted in sometimes confusing classifications of falls, fallers, and events surrounding the falls (3). Why do old people fall? Humans have an intrinsic liability to fall because of their upright posture, and this liability seems to increase with age. This may be a reflection of neurological changes with age, gait and postural changes, medical and psychological ill health, drugs, or a combination of these and other factors; nonetheless, despite many studies, the fundamental "cause" of increased liability to falls in the elderly has not been identified. The most likely reason for this failure is that falls are multifactorial in
Key words: Assessment, elderly, falls, injury_ Correspondence: Dr. J.H. Downton, Department of Geriatric Medicine, St Thomas' Hospital, Stockport, SK3 8BL, United Kingdom_ Received August 27,1989; accepted September 30,1990.
Aging, Vol. 3, N. 3 219
J.H. Downton and K. Andrews
origin, and, in addition, potential factors may be inter-related. However, previous studies can be difficult to interpret because their findings are confounded by methodological problems. Physiological and pathological factors underlying falls are likely to vary depending on the group of elderly studied. One group for whom there has been a lack of clear data is old people at home. These account for the majority of the elderly, but are in many ways the most difficult to study, unlike the "captive population" of people in institutional care. We have therefore attempted to select a representative sample of old people living at home to try and elucidate some of these problems.
SUBJECTS AND METHODS Subjects were selected from the computerised age-sex registers of five General (Family) Practitioners within Salford District Health Authority. Patients aged 75 years and over, excluding those living in residential care, were considered as potential study subjects. In order to include a sufficient number of very elderly, the selection was stratified: all patients aged 85 and over were included in the study; and of those between 75 and 84 years, one in four was chosen by random number to be included. Interviews took place between November 1986 and May 1988. All subjects were visited at home by one investigator (J.HD) and a standard questionnaire was administered. This covered: social and demographic data; use of social services (home help, meals on wheels, etc); mobility and frequency of trips out; medical symptoms; continence; dependency; depression (using the Schwab Depression Inventory) (4); cognitive function (using the Information/Orientation component of the Clifton Assessment Schedule - CAS) (5); anxiety (using the General Health Questionnaire) (6); alcohol intake; past medical history; and any drugs taken regularly. Subjects were questioned about falls experienced in the 12 months prior to interview; a fall was defined as an event in which the subject inadvertently came to rest on the floor. This definition therefore excluded near-falls and falls onto bed or chair. Those who reported falls were classified as "fallers" and those who did
220 Aging, Vol. 3, N. 3
not report any falls in the previous 12 months were classified as "non-fallers". If any falls had occurred, more detailed information was collected about the most recent fall and the circumstances in which it occurred, about any injury, and about any request for medical attention from their general practitioner (GP) or from hospital. Injuries were classified as "trivial" (bruising or laceration not requiring sutures) or "serious" (fracture or laceration requiring sutures). Subjects were also asked about whether they felt well at the time of the fall, and whether they experienced any symptoms of "dizziness" on standing (postural dizziness). An attempt was made to classify the fall. If the subject was clearly able to describe a trip or slip, the fall was classified as a "trip" fall. Symptoms were sought which might indicate a specific cause for the fall (e.g., micturition syncope, postural hypotension etc). If neither of these situations were obtained, the fall was classified as "non-specific". Examination included pulse, sitting blood pressure and drop in systolic blood pressure one and two minutes after s:anding, visual acuity (using a modified Snellen chart), a limited neurological examination (biceps and patellar reflexes and plantar responses), and measurement of passive range of movement at knee and ankle joints using a goniometer. Nutritional status was assessed by measuring triceps skinfold thickness, and circumference of the upper arm midway between the shoulder and the elbow. These measurements were used to derive the arm muscle circumference and the arm muscle area (7).
Assessment of activities of daily living was based on the Katz score (8), with 0 being fully independent, and 6 being completely dependent in the activities tested. Knee and ankle joint restriction measures were the deficits of right and left sides added together (assuming that full range of movement was 130° for each knee and 80° for each ankle). Other scores used during the analysis are shown in the Appendix. Results were analysed using the extended version of the Statistical Package for the Social Sciences (SPSS-X). The scores used were not normally distributed, 'and comparisons were therefore made using Mann Whitney U - Wil-
Falls in the elderly at home
coxon Rank Sum test. Categorical variables were compared using the X2 test with Yate's correction. Stepwise discriminant analysis was performed with fall group (i.e., faller or nonfaller) as the dependent variable, by transforming non-normally distributed scores into die hot omous categorical scores (e.g., trips out more frequently than weekly or not) which gave the best discrimination between fallers and nonfallers. Discriminant analysis was carried out for the whole group, and for men and women separately. The factors shown to be the best "predictors" of falling were then used to produce a "fall risk score".
RESULTS Of the potential study sample of 286 subjects 9 could not be traced and 10 refused to b~ interviewed. Two subjects were excluded because their GP felt they were not appropriate subjects (one was moribund and the other was deaf arid aphasic). A further 62 could not be included because by the time they were due to be interviewed they had died (33), moved away (12), or had mov~d into residential care (17). These were predominantly in the 85+ age grqup. The sample studied therefore consisted of 203 people aged 75 and over, of whom 143 (70.4%) were women and 60 (29.6%) were men. The mean age of the whole sample was 83.0 years (SO 5.0, range 75-97). Demographic data and comparison with those of 75 and over interview.ed for the General Household Survey (GHS) In 1980 (9) are shown in Table 1. Mobility was unrestricted in 103 subjects (50.7%), and a further 54 (26.6%) were mobile around their immediate neighbourhood. Only 46 (22.7%) had more restricted mobility. This is very similar to the GHS sample in which 22.2% were unable to get out of the house without help (9). Seventysix subjects (37.4%) had a home help, 10 (4.9%) had meals on wheels and 14 (6.9%) had district nurse' assistance with bathing. Only 18/203 (9.0%) scored seven points or less on the CAS information/ orientation test (the level taken as indicating significant cognitive impairment) (10). One hundred and thirty subjects (64.0%) were fully independent in activities of daily living. A further 61 (30.1%) required help only with bath-
Table 1 - Demographic data. This study Corrected for stratification
GHS 1980 (75+ years)
Women Single
70.4% 8.9%
69.9% 8.8%
Married Widowed
24.6% 66.0% 0.5%
27.2% 63.7% 0.3%
65.8% 10.1% 34.8% 54.2% 0.8%
Living alone 56.7% with spouse 2"1..7% with others 21.7%
55.3% 26.8% 17.9%
45.5% 30.1% 24.4%
Divorced
GHS = General Household Survey 1980 (9).
ing, leaving only 12 (5.9%) with varying degrees of more severe dependence. Forty-five subjects (22.2%) suffered occasional or frequent incontinence of urine. Eighty-six subjects (42.4%) had suffered one or more falls in the previous 12 months. The mean age of fallers was 83.6 years, and of nonfallers 82.5 years. This difference was not significant. Women were slightly more likely to be fallers than men (46.9% of women fell compared with 31.7% of men; X2 = 3.99, OF -'1, p = 0.046 before Yate's correction). Of those who had had a fall, 35/86 (40.7%) suffered no injury, 40 (46.5%) suffered a trivial injury, 2 subjects (2.3%) required sutures for lacerations, and 7 (8.1%) had fractures. (Two subjects could not remember whether they had injured themselves). Women were more likely to suffer injury than men: 64.6% of women fallers were injured com-
Table 2 - Time of falls. Time 06.00 - 12.00 12.00 - 18.00 18.00 - 00.00 00.00 - 06.00 Don't know
No. of falls (%) 22 (25.6) 47 10 4 3
(54.7) (11.6) (4.6) (3.5)
Aging, Vol. 3, N. 3
221
J.H. Downton and K. Andrews
Table 3 - Reason for falling. Reason
No. of falls (%)
Definite trip/slip Blackout Other specific "Dizziness" "Non-specific" Don't know
18 5 6 6 45 6
(20.9) (5.8) (7.0) (7.0) (52.3) (7.0)
pared with 36.8% of men ( X2 = 4.67, OF = 1, p = 0.03); and 7 out of the 9 serious injuries occurred in women. Most falls took place during the day, with 69 (80.3%) occurring between 6 a.m. and· 6 p.m. (Table 2). A substantial proportion of falls, 34/86 (39.5%), happened outside the home, mostly while out shopping, etc. Of falls inside the house, only small numbers occurred in the bathroom or on the stairs (3 and 2 falls respectively), the rest occurring mainly in the bedroom or living room. It was often difficult to classify why the fall had happened. Eighteen falls (20.9%) were due to a definite trip or slip, and in another 11 the history suggested a specific cause, such as cough syncope or epilepsy. In the majority of cases, subjects seemed to find themselves on the floor without being able to explain why it had happened, though some blamed "dizziness" (Table 3). Half the fallers were able to get up unaided, and only 6 were on the floor for longer than an hour.
Twenty-five fallers (29.1%) considered that they were in poor health at the time of their fall. This was not significantly associated with the type (trip or non-trip) or the place (inside or outside) of the fall. Slightly more outside than inside fallers suffered injury (50% us 33.3%), but this difference was not statistically significant. Outside fallers were more likely to have suffered trips than those who fell inside: 13 of the 34 outside fallers (38.2%) tripped compared with 5 of 49 (10.2%) inside fallers ( X2 = 6.6, OF = 1, p = 0.006). Of those who fell, 49/86 (57.0%) did not seek medical attention; the mean age of this group was the same as those who did (83.6 us 83.5 years). More men than women (41.2 % us 35.6%) sought medical attention, but this difference was not statistically significant. There was no association between the type of fall or place of fall and whether there was contact with GP or hospital. Comparisons of fallers with non-fallers (Tables 4-8) showed significant differences in CAS score, anxiety, depression, physical symptoms and knee joint restriction. Fallers were more likely to have an unstable knee joint, and to have had a previous stroke. There was no association between the number of drugs being taken, nor particular types of drugs (e.g., sedatives, diuretics etc) and a history of falls. A drop in systolic blood pressure on standing of 20 mmHg or more was found in 61 (31.3%) subjects, but postural hypotension was not associated with symptoms of postural dizziness, nor with falls. Results for the two age groups separately were similar.
Table 4 - Comparison of social factors in fallers and non-fallers.
Homehelp Meals on wheels Bath attendant Trips out less frequently than weekly ADL score (mean ± SD) Urinary incontinence Mobility score (mean ± SD) Alcohol consumption
222
Aging, Vol. 3, N. 3
Fallers
Non-fallers
p
42/86 (48.8%) 6/86 (7.0%) 7/86 (8.1%) 33/86 (38.4%) 0.62 ± 1.07 24/86 (27.9%) 1.14 ± 1.36 38/86 (44.2%)
34/117 (29.1%) 4/117 (3.4%) 7/117 (6.0%) 25/117 (21.4%) 0.45 ± 0.93 21/117 (17.9%) 0.77 ± 1.04 41/117 (35.0%)
0.006 NS NS 0.013 NS NS NS NS
Falls in the elderly at home
Table 5 - Comparison of cognitive and psychological factors in fallers and non-fallers. Fallers CAS score Anxiety score Depression score
Non-fallers
mean ±SD
p
10.5 ± 2.2 4.05 ± 3.22 8.95 ± 5.46
11.0 ± 1.2 2.91 ± 3.19 5.84 ± 5.91
0.035 0.0036 < 0.0001
Table 6 - Comparison of nutritional parameters in fallers and non-fallers. Fallers Triceps skinfold thickness (mm) Arm muscle circumference (cm) Arm muscle area (cm 2)
mean ±SD
16.1 ± 8.6 21.7 ± 2.2 38.0 ± 7.7
Non-fallers p
15.4 ± 8.8 21.9 ± 2.5 38.6 ± 8.9
NS NS NS
Table 7 - Comparison of medical history in fallers and non-fallers. Fallers Dyspnoea score (mean ± SD) Chest pain score (mean ± SD) "Arthritis" Postural dizziness Previous stroke Previous fracture Parkinson's disease Ischaemic heart disease Diabetes Number of drugs taken daily (mean ± SD)
0.99 ± 0.89 0.26 ± 0.56 51/86 (59.3%) 23/61 (37.7%) 14/86 (16.3%) 28/86 (32.6%) 3/86 (3.5%) 32/86 (37.2%) 7/86 (8.1%) 2.51 ± 2.11
Non·fallers 0.77 ± 0.88 0.08 ± 0.27 44/117 (37.6%) 32/95 (33.7%) 6/117 (5.1%) 26/117 (22.2%) 2/117 (1.7%) 37/117 (31.6%) 12/117 (10.3%) 2.08 ± 1.89
p
0.049 0.009 0.004 NS 0.017 NS NS NS NS NS
Table 8 - Comparison of medical examination in fallers and non-fallers. Fallers Systolic BP (mmHg) (mean ± SD) Systolic drop on standing (mmHg) (mean ± SD) Irregular pulse Visual acuity score (mean ± SD) Restriction of hip movement Knee instability Knee restriction (R + L) (mean± SD) Ankle restriction (R + L) (mean ± SD) Abnormal knee jerk Abnormal plantar response
179.7 ± 28.8 11.2 ± 11.3 23/86 (26.7%) 2.63 ± 1.78 8/86 (9.3%) 6/86 (7.0%) 71° ± 37° 81° ± 21° 21/85 (24.7%) 15/86 (17.4%)
Non-fallers 179.9 ± 34.2 13.9 ± 12.9 33/116 (28.4%) 2.28 ± 1.60 6/117 (5.1%) 1/117 (0.85%) 58° ± 26° 77° ± 21° 20/117 (17.1%) 14/115 (12.2%)
p
NS NS NS NS NS 0.049 0.012 NS NS NS
Aging, Vol. 3, N. 3 223
J.H. Downton and K. Andrews
Stepwise discriminant analysis was able to classify correctly only 71.8% of cases as "fallers" or "non-fallers" when all subjects were considered. Classification was less effective if women alone were considered (65.7% correctly classified), and better if men alone were considered (88.1% correctly classified). The discriminating factors ("predictors"), their discriminant function coefficients and the percentage each factor contributes to the discrimination are shown in Table 9. It can be seen that the "predictors" of falling and their relative importances are different for men and women. For all subjects, mobility has a negative coefficient meaning that poor mobility "protects", against falling. Living alone (rather than with spouse or relatives) is a "risk factor" for men but is "protective" for women. When these "predictors" were used to calculate a "fall risk" score, fallers and nonfallers had highly significant differences in scores, but there was a wide overlap in scores between fallers and non-fallers (Table 10). If a score of 5 or more is taken as indicating a subject who is likely to fall, then the score has
a specificity and sensitivity of 69%. The predictive value of a "positive" score (i.e., ~ 5) is 61.7% and that of a "negative" score (i.e., ~ 4) is 75.5%.
DISCUSSION Any attempt to study old people in the community is beset with problems. The relatively high geographical mobility of old people (ll) and the fact that there is an inevitable mortality means that it is difficult to ensure a "typical" sample of community elderly, particularly if data collection takes more than a short time. There are also differences in social mix and in proportions of elderly between different areas of the country (and also between countries). This study has attempted to look at a representative selection of people aged 75 and over living in their own homes by using the computerised age-sex registers of five local GPs. The likely attrition of the oldest subjects was taken into account by stratifying the selection to include sufficient numbers of very elderly.
Table 9 - Stepwise discriminant analysis: discriminant junction coefficients. Predictors Poor mobility T rips out less frequently than weekly Depression Previous stroke Symptoms of arthritis Female sex Meals on wheels Unstable knees Alcohol consumption Visual acuity worse than 6/9 Systolic BP Homehelp Urinary incontinence Lives alone Fear of falls Bath nurse Dizziness Abnormal kneejerk Correctly classified
224 Aging, Vol. 3, N. 3
All subjects
Women
·0.60 (14.7%)
·0.73 (20.9%)
0.54 (13.1%)
0.50 (14.1%) 0.30 (8.2%) 0.30 (8.6%) 0.49 (13.9%)
0.50 (12.3%) 0.47 (11.4%) 0.37 (8.9%) 0.35 (8.6%) 0.33 (7.9%) 0.29 (7.0%) 0.24 (5.8%) 0.24 (5.8%) 0.19 (4.6%)
0.32 (9.0%)
·0.38 0.56 0.69 0.95
(5.9%) (8.7%) (10.7%) (14.8%)
0.31 (4.8%) 0.87 (13.6%) 0.41 (6.4%)
0.40 (11.3%) 0.25 (7.2%) ·0.42 (6.9%)
71.8%
, Men
65.7%
0.33 (5.2%) 0.72 (11.2%) ·0.54 (8.4%) 0.34 (5.4%) ·0.31 (4.8%) 88.1%
Falls in the elderly at home
Table 10 - Fall risk ~coresfor fallers and non-fallers. Fall risk score
Fallers
Non-fallers
1 2 3
0 0 4 22 28 27 2 0
1 10 32 37 28 7 1 0 0
4 5 6 7 8 9
x2 = 45.7, DF = 8, p < O.QOOOL Compared with data from the General Household Survey (9), the study sample has a higher percentage of widow(er)s and larger numbers living alone, partly reflecting the larger proportion of very elderly resulting from the stratification procedure. (In the population of the practices used, people of 85 and over made up 25% of over-75s whereas they made up 47% of the eventual study sample.) Because of the difficulties in selecting a truly random sample, and the dependence of information on the subject's memory for events that have occurred a variable amount of time earlier, any extrapolations from a study of this type have to be made cautiously. In addition, some findings depended on a subjective assessment by the interviewer (for example, the reason for the fall). However, since all subjects were seen by a single interviewer, the problem of multiple, different subjective judgements was avoided. One of the aims of the study was to cover as many as possible of the factors which might be associated with falling. In order to maintain the good will of the subjects during a fairly lengthy interview, assessment of individual factors, and particularly the number of "invasive" assessments, necessarily had to be limited. This meant that, for example, neurological examination and assessment of orthostatic hypotension were not carried out in an ideal fashion. However, the very low refusal rate means that the results are more likely to be generalisable to other community populations.
The characteristics of the sample demonstrate that the stereotype of old people at home being frail, confused and requiring care is misleading; most subjects were mobile, independent and mentally clear. Despite this, falls were very common, affecting almost half of the subjects during the 12 months prior to interview. This study confirms that women seem more likely to fall than men (12-14) but it does not show the expected increasing incidence of falls with age though the selection was stratified to include adequate numbers of very elderly. It is possible that over the course of a study of 18 months duration there was selective loss of the frailest (and perhaps most likely to fall) elderly through death or entry into residential care. However, Sheldon (12) collected data over a similar length of time and found a very marked increasing prevalence of falling with increasing age, though he studied relatively few very elderly. Injuries due to falls were also comr:non, though not usually serious. The figures quoted here are almost certainly an underestimate because only the most recent fall was considered when enquiring about injury. Women were more likely to suffer injury, and although the larger number of fractures can be explained by the higher prevalence of osteoporosis in women, the reason for the higher number of injuries overall is not clear. It may reflect different thresholds of concern about the consequences of falls, since it could be interpreted as a subjective judgement of how serious the fall was. However, more men than women sought medical attention after the index fall. It proved extremely difficult to classify why a fall had occurred. A small proportion of subjects gave a clear story of a trip or slip (such as tripping over uneven paving stones or slipping on ice), but more commonly the description of the circumstances was very vaguean
