Physiological Factors Associated with Falls in an Elderly Population Stephen R. Lord, Phd,* Russell D. Clark, MBBS, FRACP, DTMbH,*t and Ian W.Webster, MD, FRACP* Objective: To determine whether a battery of 13 sensorimotor, vestibular, and visual tests discriminates between elderly fallers and elderly non-fallers. Design: One-year prospective study. Setting: Conducted at a 124-bed Hostel for Aged Persons, in Sydney, Australia. Participants: Ninety-five persons aged between 59 and 97 years (mean age 82.7 years) took part in the study. Of the 29 non-participants, four were ill, five were absent (on holidays, etc), and 20 declined. Residents were generally independent in activities of daily living although personal care assistance was available. Results: Eighty-four participants were available for followup. In the follow-up year, 40 subjects experienced no falls, 11 subjects fell one time only, 33 residents fell on two or more occasions. There was a total of 145 falls. Discriminant
function analysis identified proprioception in the lower limbs, visual contrast sensitivity, ankle dorsiflexion strength, reaction time, and sway with the eyes closed as the variables that significantly discriminated between subjects who experienced multiple falls and subjects who experienced no falls or one fall only. This procedure correctly classified 79% of subjects into multiple faller or non-multiple faller groups. Quadriceps strength was poorer in the multiple fallers compared with the non-fallers and once-only fallers, although the difference was not statistically significant. There was little difference in the mean scores for the tests of vestibular function in the non-fallers, once-only fallers, and multiple fallers. Conclusion: It appears that this approach highlights some key physiological factors that predispose elderly individuals to falls. J Am Geriatr SOC 391194-1200,1991
number of studies have found that falls in elderly persons are common and are associated with appreciable morbidity and mortality.'-* Three recent prospective community studies have developed relative risk rates for falls based upon a number of clinical Thus, the taking of psychoactive medications, the taking of more than four medications, the presence of stroke or Parkinsons disease, a past history of falls, the presence of impaired gait, muscle weakness, arthritis, foot problems, impaired cognition, abnormal neurological signs, and advanced age are all individually and cumulatively found to be important predictors for falls. We have taken a different approach to predicting and understanding falls by examining altered physiological control of postural stability. Our model of postural control (which is outlined fully elsewhere) emphasizes sensory, motor, and integration components? Our approach to postural instability and falls in elderly persons is "functional"as opposed to "diseaseoriented." It is suggested that regardless of any underlying medical condition elderly persons may suffer (whether diagnosed or not), impairments in stability will be manifested in reduced functioning in one or more of the sensori-motor systems upon which postural control depends. There is general agreement that postural control involves many body systems, but little work has been done on assessing to what extent demonstrated agerelated decline in sensory, motor, and integration sys-
tems have on the overall decline in postural control that occurs with age and whether impairment in these systems is associated with the increased incidence of falls experienced by elderly persons. We therefore developed a battery of simple, noninvasive tests to provide reliable measures of specific components of the "postural control process" which could be administered to elderly subjects in one session. By using this battery of simple tests, we have obtained concurrent examinations of sensory and motor factors involved in postural control as well as useful composite measures of postural stability. In this paper we examine which of these physiological factors are associated with falls in a sample of 95 persons living in a hostel for aged persons.
From the 'School of Community Medicine, University of New South Wales, Kensington, and tSt. Vicents Hospital, Sydney, N.S.W., Australia. This work was carried out in and funded by the School of Community Medicine, University of New South Wales. Address correspondence and reprint requests to Stephen R. Lord, School of Community Medicine, University of New South Wales, PO Box 1, Kensington, N.S.W.,2033, Australia. IAGS 39:1194-1200, 1991 0 1991 by the American Geriafrics Society
MATERIALS AND METHODS Description of the Sample. All residents of a hostel for the aged, which housed 124 residents, were invited to take part in the study. Ninety-five persons aged between 59 and 97 years (mean age 82.7 years) agreed to take part, the age and sex distribution is shown in table 1. (Table 1) Of the twenty-nine nonpartiapants, four were ill, five were absent (on holidays etc), and 20 declined. Residents of the hostel lived in their own "motel style" rooms. The main criterion for hostel living was that residents must be sufficiently mobile and independent to use a common dining room for meals. Most residents (66%) left the hostel for varying periods everyday, and a number regularly used public transport. Residents were generally independent in activities of daily living although personal care assistance was available. The Follow-up Procedure All residents who undertook the balance tests were followed up for a period 0002-S6l4~9llS3.50
PHYSIOLOGICAL FACTORS AND FALLS IN THE ELDERLY
IAGS-DECEMBER 1991-VOL. 39, NO. 12
of 1 year to determine whether the physiological test measures were associated with falls. The following definition of a fall was used in this study: “A fall is an event which results in a person coming to rest unintentionally on the ground or other lower level, not as the result of a major intrinsic event (such as a stroke or syncope) or overwhelming h a ~ a r d . ” ~ Questionnaires were given to residents every month. The questionnaire contained questions seeking details on the number of falls in the past month, the location, the cause, and any injuries suffered. In addition to the monthly questionnaires, 4-monthly “room visit” follow-ups were made where questionnaires were missing or not clearly filled out. To minimize under-reporting of falls by elderly subjects with memory impairments, a falls record book was established by the nursing staff at the commencement of the study. Nurses recorded falls they witnessed, and cases where residents came to the nursing station for treatment of an injury because of a fall. This provided a further record of falls experienced by the hostel residents. Seventy-nine falls were recorded in the falls record book. The record book reports increased the number of falls reported from 110 to 145 and resulted in four subjects being classified as fallers rather than non-fallers. Outcome of residents at the end of the yearly followup period is shown in Table 2. Five subjects who left the hostel or died within the 12-month follow-up fell on one or more occasions while at the hostel. These subjects were classified as fallers and included in the prospective results. One other subject who left the hostel was followed up for the 12-month period. She experienced no falls. Thus, 84 subjects were available to follow-up. Analyses were performed that compared multiple fallers (those who fell on two or more occasions) with non-multiple fallers (those who did not fall or fell on one occasion only). Multiple falls were examined because it has been found that single falls are less predictable and may be due to accidents and overwhelming incident^.^ Multiple falling, on the other hand, may be a better indicator of those with a physiological predisposition to falling. Analyses were also performed that compared performance in the physiological tests in non-fallers and fallers (those who fell on one or more occasions). Apparatus and Procedures The test battery included nine tests of individual sensory and motor systems and four “composite”tests of reaction time and stability. The sensory and motor tests included two visual tests, a standard test of visual acuity, and a test TABLE 1. AGE AND SEX DISTRIBUTION OF THE SAMPLE Age Group Men Women Total 50-69 70-79 80-89 90-97
Total Mean (SD)
0 4 9 3 16 82.8 (7.4)
2 22 46 9 79 82.7 (6.5)
2 26 55 12 95
TABLE 2. OUTCOME OF RESIDENTS AT END OF FOLLOW-UP PERIOD Number Percentage __________
Residing at hostel Transferred to other
Moved to private
accommodation Transferred to nursing
of contrast sensitivity; three tests of sensation in the leg, touch thresholds at the ankle, vibration sense at the knee, and a test of proprioception; two tests of vestibular function, the vestibular stepping test, and a test of vestibular optical stability; and two tests of muscle strength, quadriceps, and ankle dorsiflexion strength. The composite tests included a test of reaction time, tests of body sway on firm and compliant (foam rubber) surfaces, and a static and a dynamic balance test to measure postural stability. Visual acuity was measured binocularly using a standard Snellen scale with subjects wearing their best correction. Acuity was measured in terms of the logarithm of the minimum visual angle resolvable at a test distance of four meters. Contrast sensitivity was assessed using the “Melbourne Edge Test.”lo The test presented 20 circular patches containing edges with reducing contrast. Correct identification of the orientation of the edge allowed a measurement of contrast sensitivity expressed in decibel units, where dB = -lOloglo Contrast. Touch thresholds were measured with a SemmesWeinstein Pressure Aesthesiometer.” This instrument contained 20 nylon mono-filaments of equal length, but varying in diameter. The filaments were applied to the center of the lateral malleolus, and measurements were expressed in logarithms of milligrams pressure. Proprioception was tested using apparatus based on a design by De Domenico and McCloskey.” Subjects, with their eyes closed, attempted to simultaneously place the big toe of each foot at the same position but opposite side of a vertical perspex sheet. Any difference in matching was measured in degrees. Vibration sense was measured using an electronic device capable of generating a 200 Hz vibration of varying intensity. The vibration was applied to the tibia1 tuberosity and was measured in microns of motion perpendicular to the body surface. The Vestibular Stepping Test measured subjects’ ability to remain stationary and oriented in the one plane while “walkingon the spot” with eyes closed for 1 minute.? The distance travelled (measured in cm) from the starting point and the angle rotated (measured in degrees) were measured. The Vestibular-Optical Stability Test measured any difference between visual acuity at rest and while walking on a treadmill, recorded in logarithms of visual angle. Quadriceps strength was measured in the sitting
LORD ET AL
position. A strap (which was connected to a spring gauge) was placed around the subject's stronger leg. A measure of maximal quadriceps strength was obtained when the subject attempted to extend her leg against the pull of the gauge. Strength was corrected for body size by dividing the measure by the height of the subject. Ankle dorsiflexion strength was also measured in the sitting position. A strap (which was connected to a spring gauge) was placed over the subject's stronger foot just proximal to the commencement of the little toe. A measure of maximal ankle dorsiflexion strength was obtained when the subject attempted to raise her foot against the pull of the gauge. Strength was corrected for body size by dividing the measure by the height of the subject. Reaction time was assessed with a simple reaction time paradigm, using a light as the stimulus and depression of a switch (by the hand) as the response. A reaction time paradigm using hand movement as the response (rather than a movement by the foot) was used in an attempt to obtain a measure of reaction time that emphasized the decision-time component rather than the movement-time component. This was done to gain an indirect measure of central nervous system processing speed. Reaction time was measured in milliseconds. Sway was measured using a swaymeter that measured displacements of the body at the level of the waist. The device consisted of a rod attached to the subject at waist level by a firm belt. The rod was 40cm long and extended behind the subject. A pen attached to the end of the rod recorded the movements of the subject on a sheet of graph paper (with a millimeter square grid) which was fastened to the top of an adjustable height table. Testing was performed on a firm surface (a small pile carpeted floor) and on a piece of foam rubber ( l m by l m by 15 cm thick) with the subject standing in the center. The same test was repeated on both surfaces with the subject's eyes closed. The foam rubber was used to reduce proprioceptive input from the ankles and cutaneous inputs from the soles of the feet so that subjects would be required to rely on visual and vestibular cues to maintain a steady stance. Total sway (number of square millimeter squares traversed by the pen) in the 30second periods were recorded for the four test conditions. The Static Balance Test measured the ability of subjects to maintain balance while standing on a firm and a compliant surface. Subjects were classified into three grades: Grade 1 - unable to maintain balance for 30 seconds on a firm surface (with eyes open or closed) or unable to maintain balance on the compliant surface for any period without support; Grade 2 - capable of undertaking the tests of the firm surface but unable to maintain balance for 30 seconds on the compliant surface; and Grade 3 - capable of maintaining balance while standing on the firm and compliant surfaces (eyes open and closed) for 30-second periods without difficulty. Grades in this test were given while subjects were undertaking the sway tests. The Dynamic Balance Test was an assessment of the
JAGS-DECEMBER 1991-VOL. 39, NO. 12
subject's ability to maintain balance while performing the vestibular stepping test - walking on the spot for 1 minute with the eyes closed. Subjects were classified into five grades, from Grade 1, which indicated that the subject's balance was so poor that the test could not be administered to "Grade 5", which indicated that the subject could perform the task with no difficulty maintaining good balance. Test Reliability Most of the tests have been assessed for test-retest reliability in pilot studies. Total sway with the eyes open and eyes closed, quadriceps strength, visual acuity, and the sum of distance and angle recorded in the vestibular stepping test had reliability co-efficients greater than 0.70. Vibration sense, reaction time, vestibular stepping distance, and vestibular stepping angle had reliability co-efficients between 0.50 and 0.70. The remaining tests-proprioception, touch, and vestibular optical stability-showed moderate to poor test-retest reliability. Full descriptions of the apparatus and procedures, along with test-retest reliability scores (and confidence intervals) for the test measures are reported elsewhere? The 13 tests were administered to the subjects by one experimenter (S.L.) who was blind to faller status, ie, the number of falls suffered in the year before testing. Testing for each subject was usually completed within 1 hour. Cognitive Status and Medication Use Cognitive status and medication use were also assessed as it has been shown that these factors are independent risk factors for falls in elderly populations. Cognitive status was assessed clinically by the supervising nurse at the time of test administration. The subjects were classified as either having no significant cognitive impairment or significant impairment as characterized by disorientation, confusion, and memory impairment which had been persistent over the 12 months prior to initial assessment. A record of medication use for each resident was taken at the time of testing. Statistical Analysis The sensory, motor, and body sway test measures were coded as continuous variables, while the static and dynamic balance test scores were ordinally scaled. For variables with right skewed distributions (such as the vestibular stepping test scores and the,proprioception, quadriceps strength, reaction time, and sway measures) logs of variables were analyzed. The data were analyzed using the SPSSX computer pa~kage.'~ Analysis of variance was used to assess differences in the means of the visual, vestibular, sensori-motor, and sway test measures between the faller and nonfaller groups while controlling for the effects of age. Log-linear models were used to assess the associations between falls status and the two clinical measures of stability (static and dynamic balance). In these analyses, models with all two-way interactions: stability measure * falls, stability measure * age, and falls * age were compared to models that differed only in that they omitted the stability measure * falls effects. Changes in the likelihood-ratio Chi squares (or partial Chi squares) were then assessed for significance. Stepwise discriminant function analysis (a statistical
PHYSIOLOGICAL FACTORS AND FALLS IN THE ELDERLY
IAGS-DECEMBER 1991-VOL. 39, NO. 12
technique which highlights the variables that are most important in the identification of groups) was used to determine which postural control variables discriminated between the multiple fallers and non-multiple fallers. In the discriminant function procedure, age was included as a possible "predictor" variable. The standardized co-efficients for the variables included in the discriminant function (expressed in z score form) are presented. As the units of each measure have been standardized, the co-efficients give an indication of the relative importance of each variable in explaining the variance in the dependent variable. After deriving the discriminant function, validation of the predictors was camed out using the jack-knife procedure. This procedure involves leaving out each of the cases in turn and calculating the function based on the remaining (n-1) cases. Classification of the remaining case is then made. This technique provides a more conservative estimate of the misclassification rate.
RESULTS Incidence of Falls Forty subjects reported having no falls in the follow-up year, while 44 subjects re-
ported one or more falls. There was a total of 145 falls overall. Of those who fell, 11 fell once only, 14 fell on two occasions, while 19 fell three or more times. The mean age of the non-fallers, once-only fallers, and multiple fallers were 81.4 years, 82.7 years, and 84.6 years, respectively (f=2.76, df=2.81, P=0.07).
Physiological Measures and Multiple Falls. Table 3 shows the mean scores plus standard deviations for each continuously scored test measure for the non-fallers, once-only fallers, multiple fallers, and the total sample. (Table 3) Due to poor balance, 34 subjects could not undertake the vestibular optical stability test, 19 subjects could not undertake the vestibular stepping test, 18 subjects could not undertake the sway-on-thefoam-with-eyes-open test, and 24 subjects could not undertake the sway-on-the-foam-with-eyes-closed test. Those who fell on two or more occasions in the year after testing had reduced visual contrast sensitivity, decreased proprioception in the lower limbs, slower reaction times, increased sway on the floor with eyes closed, and increased sway on the foam with eyes open
TABLE 3. MEAN VALUES AND STANDARD DEVIATIONS FOR THE TEST MEASURES Once-Only Multiple Non-Fallers Fallers Fallers Total Vision Visual acuity" Contrast sensitivityb Peripheral sensation Proprioception' Touchd Vibration sensee Vestibular function Vest optical stability' Vest stepping distances Vest stepping angleh Vest sum (dist + angle) Strength Quadriceps strength/height' Ankle strength/height' Reaction time' Sway-eyes openk Sway-eyes closedk Sway on foam-eyes openk Sway on foam-eyes closedk
2.39 (1.03) 17.5 (3.2)
6.44 (14.5) 16.9 (2.6)
5.11 (7.58) 14.5 (4.6)**
3.99 (7.10) 16.3 (4.0)
2.2 (1.5) 3.93 (0.48) 32.3 (24.3)
2.6 (2.3) 4.36 (0.36) 45.2 (35.1)
3.5 (2.3)** 4.24 (0.39) 42.6 (31.6)
2.7 (2.0) 4.11 (0.46) 38.1 (29.0)
0.20 (0.27) 113 (79) 52 (51) 165 (90)
0.11 (0.14) 128 (43) 54 (58) 183 (101)
0.29 (0.33) 105 (48) 56 (73) 161 (89)
0.21 (0.27) 112 (68) 54 (59) 166 (90)
9.08 (3.48) 2.80 (0.81) 267 (64) 98 (48) 109 (59) 187 (71)
9.58 (2.77) 2.75 (0.69 289 (59) 109 (59) 147 (121) 179 (52)
7.80 (3.05) 2.46 (0.75) 376 (205)** 125 (80) 178 (154)* 307 (167)**
8.65 (3.27) 2.66 (0.78) 313 (146) 110 (64) 140 (116) 228 (126)
Low scores in the contrast sensitivity and strength tests and high scores in all other tests indicate impaired performance. * --significant difference betuleen subjects w h o did not fall or fell one time only and subjects w h o had multiple (2+) falls after controlling for age (P < 0.05). ** -significant difference between subjects who did not fall or fell one time only and subjects w h o had multiple (2+) falls after controlling for age
(P < 0.01). --smallest visual angle correctly reported at four meters (minutes). -dB log contrast. -degrees difference. -loglo 0.1 milligrams pressure. -microns of motion perpendicular to body surface (peak to peak). ' -difference in smallest visual angle correctly reported. at 4 meters on treadmill and smallest visual angle. whilst seated (minutes). -distance moved from initial position in one minute (centimeters). -angle of rotation in one minute (degrees). ' -(kg/cm) X 100. I -milliseconds. '-millimeter squares traversed by pen on swaymeter in 30 seconds.
correctly reported at four meters
LORD ET AL
and closed compared with those who did not fall or fell on one occasion only. Quadriceps and ankle dorsiflexion strength were poorer in the multiple fallers compared with the nonfallers and once-only fallers, although the differences were not statistically significant. There was little difference in the mean scores for the tests of vestibular function in the non-fallers, once-only fallers, and multiple fallers. Multiple fallers performed poorly, compared with the non-multiple fallers in the two clinical stability tests: static and dynamic balance (partial Chi2=7.2, df=2, P=0.028 and partial Chi2=15.1,df=4, P=0.005, respectively). The discriminant analysis revealed that contrast sensitivity, proprioception in the lower limbs, ankle dorsiflexion strength, reaction time, and sway with eyes closed discriminated significantly between the multiple faller and non-multiple faller groups, as indicated by the final Wilk’s lambda of 0.71 (P=O.OOO) and a canonical correlation for the discriminant function of 0.54. The standardized canonical correlation co-efficients were -0.36 for contrast sensitivity, 0.50 for proprioception, -0.23 for ankle dorsiflexion strength, 0.55 for reaction time, and 0.39 for sway with eyes closed. These variables correctly classified 79% of the cases before validation using the jack-knife procedure and 75% after validation.
Physiological Measures and One or More Falls The significant differences found in the test measures between the non-multiple and multiple (2+ falls) fallers were also found when comparing those who experienced no falls at all in the follow-up period, and those who fell on one or more occasion, with two exceptions. Sway on the foam with eyes closed was not significantly different between fallers and nonfallers, and touch thresholds were significantly higher in the fallers compared with the non-fallers (f=10.6, df=1,80, P=.002). Visual acuity and vibration sense were superior in those who experienced no falls compared with those who fell on one or more occasion, although the differences were not statistically significant after controlling for age.
IAGS-DECEMBER 1991-VOL. 39, NO. 12
recurrent falls (three or more falls) in the 2 months before transfer. No falls were reported in the other four residents transferred to nursing homes in the period between initial assessment and transfer. Cognitive Impairment and Medication Use Thirteen of the 84 residents available to follow-up were identified as having cognitive impairments. Those with cognitive impairment were more likely to suffer multiple falls-10 of 13 residents with cognitive impairment fell two or more times compared with only 23 of the 71 subjects with no cognitive impairment (Relative risk = 2.37, 95% CI=1.52-3.72). Twenty-nine residents were taking one or more psychoactive medications (sedatives, anti-anxiety agents, anti-psychotics, and anti-depressants) at initial assessment. No significant association was found between psycho-active drugs and falling-15 of the 29 residents taking psychoactive drugs suffered multiple falls compared with 18 of the 55 subjects not taking psychoactive medications (Relative risk = 1.58, 95% CI=0.942.65). Similarly, there was no significant association between use of drugs with a hypotensive effect (psychoactive drugs, anti-hypertensives, and diuretics) and falls. Twenty-five of the 54 residents taking “hypotensive” drugs fell compared with eight of the 30 subjects not taking “hypotensive”medications (Relative risk = 1.74, 95% CI=0.90-3.36).
Recent research into causes of falls in elderly persons has approached the problem from a clinical perspective and has highlighted a number of important risk fact o r ~ . ~ ,In ~ - contrast, ’ the present study placed a major emphasis on gaining quantitative measures of sensory, motor and balance factors. It was found that multiple fallers had reduced vision, decreased peripheral sensation, slower reaction times, and decreased stability compared with non-multiple fallers. The finding of an association between reduced proprioception and falls is in accord with Brocklehurst et al,15 and the finding that body sway is increased in fallers is in line with a number of published studCharacteristics of Residents Who Died or Were ies.6,15-18 Unlike the studies by Whipple et all9 and Transferred to Nursing Homes At the end of the Studenski et a1,20however, no significant univariate 12-month follow-up, those who were still residing in associations were found between measures of strength the hostel and those who moved to another hostel or in the lower limbs and falls. This may be because the to private accommodation were classified as one studies by Whipple et a1 and Studenski et a1 were group-those with a ’good’ outcome. Subjects who retrospective in design, while the present study was were transferred to a nursing home and those who prospective in nature. While prospective studies have many advantages died were classified into a second group-those with a ‘poor’ outcome. Age was not a significant predictor over retrospective studies, they have an inherent disof outcome, although the mean age of those with a advantage in that a percentage of subjects are inevitapoor outcome was slightly higher (84.9 years) than the bly lost to follow-up. In the present study, 12% could mean age of those with a good outcome (82.3 years), not be followed up for a full year. This was not (a (t=1.05, P=0.31). Residents with a poor outcome had random loss but primarily a selective loss through significantlyinferior contrast sensitivity and quadriceps death and transfers to nursing homes. These subjects strength compared with residents with a good outcome. had significantly inferior strength compared with those Subjects with a poor outcome also performed signifi- available to follow-up 12 months after balance testing. The analysis in the prospective study was therefore cantly worse in the static and dynamic balance tests (all differences significant at P