The Winchester falls project: a randomised controlled trial of secondary prevention of falls in older people

Transcription

1 Age and Ageing 2008; 1 8 doi: /ageing/afn192 Age and Ageing Advance Access published October 1, 2008 C The Author Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org The Winchester falls project: a randomised controlled trial of secondary prevention of falls in older people CLAIRE L. SPICE 1,WENDY MOROTTI 2,STEVE GEORGE 3,THOMAS H. S. DENT 4, JIM ROSE 5,SCOTT HARRIS 6,CHRISTOPHER J. GORDON 7 1 Consultant, Department of Medicine for Older People, Portsmouth Hospitals NHS Trust, Queen Alexandra Hospital, Southwick Hill, Cosham, PO6 3LY, UK 2 Project Research Nurse, Mid Hampshire Primary Care Trust, Tidbury Farm, Bullington Cross, Sutton Scotney, Hampshire, SO21 3QQ 3 Director of University of Southampton Clinical Trials Unit MP131, Southampton General Hospital, Tremona Road, Southampton, SO16 6YD, UK 4 Mid Hampshire Primary Care Trust, Tidbury Farm, Bullington Cross, Sutton Scotney, Hampshire, SO21 3QQ 5 GP Principal. Derrydown Clinic, St Mary Bourne, Andover, UK 6 Statistician, Public Health Sciences and Medical Statistics Group, University of Southampton School of Medicine, Southampton General Hospital, Tremona Road, Southampton, SO16 6YD, UK 7 Consultant Physician, Medicine and Elderly Care, Royal Hampshire County Hospital, Romsey Road, Winchester SO22 5DG, UK Address correspondence to: Claire Spice, Department Medicine for Older People, Portsmouth Hospitals NHS Trust, Queen Alexandra Hospital, Southwick Hill, Cosham, PO6 3LY, UK. claire.spice@porthosp.nhs.uk This study was a joint initiative between the Royal Hampshire County Hospital and Mid-Hampshire Primary Care Trust with the work being carried out within the community and at the Royal Hampshire County Hospital. Abstract Background: the mortality and morbidity of falls in older people is significant, with recurrent fallers being at an increased risk. The most effective way to reduce falls in this group is not clear. Objective: to determine the effectiveness of two interventions, one based in primary care and the other in secondary care, at preventing further falls in recurrent fallers. Design: cluster randomised controlled trial. Participants: sixty-five years or over, living in the community, two or more falls in the previous year and not presenting to an emergency department with index fall. Setting: Mid Hampshire, UK. Intervention: eighteen general practices were randomly allocated to one of three groups. The primary care group was assessed by nurses in the community, using a risk factor review and subsequent targeted referral to other professionals. The secondary care group received a multi-disciplinary assessment in a day hospital followed by identified appropriate interventions. The control group received usual care. Follow-up was for 1 year. Results: five hundred and five participants were recruited. Follow-up was completed in 83% (421/505). The proportion of participants who fell again was significantly lower in the secondary care group (75%, 158/210) compared to the control group [84%, 133/159, adjusted odds ratio (OR) 0.52 (95% CI ) P = 0.002]. The primary care group showed similar results to the control group [87%, 118/136, adjusted OR 1.17 (95% CI ) P = 0.673]. Conclusion: a structured multi-disciplinary assessment of recurrent fallers significantly reduced the number experiencing further falls, but a community-based nurse-led assessment with targeted referral to other professionals did not. Keywords: elderly, older people, prevention, randomised controlled trial, recurrent falls Introduction Falls are a major cause of death, injury, functional decline, hospital admission, psychological trauma and institutionalisation in older people [1 4]. Recurrent fallers are at higher risk [1, 5]. The most effective type of interventions, the optimal setting and the best groups to target were not clear [6]. 1

2 C. L. Spice et al. The structured bi-disciplinary assessment of fallers presenting to an emergency department significantly reduced subsequent falls [7]. A community-delivered intervention targeting those with falls risk factors was beneficial [8] although a less intensive intervention in the community was not [9]. Several studies included some older people with recurrent falls [7, 8, 10] but, at the time of design of this study, only one had investigated this group specifically [9]. Subsequently, a multifactorial intervention in recurrent, fallers presenting to an accident and emergency department, achieved significant reduction in total falls but not the proportion of fallers [16]. This study looks at two interventions in community dwelling older recurrent fallers, who had not attended an emergency department for their most recent fall, comparing effectiveness in preventing falls against usual care in a cluster randomised controlled trial. Methods Participants Participants were 65 and over with two or more falls in the preceding year. Community-based health professionals and social services staff identified fallers using a standard definition of a fall [7]. The most recent fall prior to recruitment was termed the index fall. Exclusions: life expectancy less than a year, plans to move from the area within 1 year, abbreviated mental test score [11] of less than 7, non-english speakers with no available interpreter and nursing home residents. The immobile were not excluded. Randomisation and baseline assessment All 19 general practices within the Mid-Hampshire Primary Care Trust area were invited to participate: one declined. Practices were stratified into urban (3) and rural (15) and randomly allocated to the three arms, in blocks of three, using a random number generator on a Hewlett Packard 21S pocket calculator. Consent was gained at the cluster level from the lead general practitioner in each practice prior to randomisation. Individual participants consented after receiving an information leaflet about the project. Local research ethics committee approval was granted. A baseline assessment (demographic information, abbreviated mental test score [11], modified Barthel index [12], timed Get up and Go test [13], medical diagnoses, drug history, details of the index fall and previous falls and risk factors for osteoporosis) was conducted by a nominated nurse for each practice. Interventions Secondary care intervention group participants attended a one-stop multi-disciplinary clinic with referral for investigations, interventions (including Homecheck) and follow-up if necessary. Primary care intervention group participants received an assessment by the designated trained nurse to identify risk factors for falls: if problems were identified, referrals to appropriate professionals in primary or secondary 2 care were made. The usual care group received a baseline assessment but were managed by their primary care team without specific guidance: referral to routine services was at the discretion of the primary care clinicians. Intervention assessments in the primary and secondary groups were standardised: further management of each participant was then individualised, with no specific protocol, and interventions were recorded (Appendix 5). Follow-up Participants were followed monthly for 12 months. If a participant had fallen, they indicated how many falls they had by selecting from the options of 1, 2, 3, 4 or >4. Concerns raised regarding accuracy of recall for higher numbers of falls meant that participants were not asked to give numbers of falls >4. If the card was not returned, the participant was contacted by telephone. An exit assessment was performed after 1 year. Blinding to the intervention group of those collecting and analysing data was impractical, but all data collected were entered without alteration. Outcomes Primary outcome was the proportion of participants in each group who had at least one fall during follow-up. Secondary outcomes were death, move to institutional care, change in Barthel score, change in timed Get up and Go score, fallrelated fractures and admissions. Sample size We assumed that 52% of controls would fall [7]. Estimating the (unknown) intra-cluster correlation coefficient to be 0.01, initial calculations estimated that 150 participants in each arm (450 total) would show a 20% reduction in the proportion of participants falling during the follow-up period using ά = 0.05 and (1 β) = 0.8. Using an interim intra-cluster correlation coefficient (0.019), final sample size was re-calculated as 172 per arm (516 total). Statistical analysis Analysis was conducted in accordance with the principles laid out in ICH E9 guidelines on an intention to treat basis [14]. The small number of allocation units in this trial made it unlikely that randomisation alone would control entirely for confounding. Dichotomous outcomes were analysed using logistic regression analysis in Stata 9.0, adjusted for clustering by practice and for baseline variables of age, modified Barthel index, previous number of falls, consumption of four or more drugs, calcium or calcium and vitamin D use and joint disease. Adjustment for clustering in Stata specifies that observations are independent across groups (clusters), but not necessarily within groups. To cope with this a Huber/White/sandwich robust estimator of variance is used in place of the traditional calculation to allow observations that are not independent within cluster (although they must be independent between clusters) [17 19]. The primary outcome was additionally analysed using Cox regression, adjusted for the same potential

3 confounders, using the month of first fall during follow-up as a proxy for time to first fall. Stata uses a similar method to allow for clustering in Cox regression to that used in logistic regression. One outcome (change in Barthel score) was a continuous variable and was analysed in a multi-level model in SPSS 12.0, adjusting for the same potential confounders [20]. Separate analyses were performed for primary and secondary intervention groups against usual care as a control. The intention to treat analysis included all eligible randomised patients and assumed those lost to follow-up had either undergone a negative event (fall/death) or had not undergone a positive change (modified Barthel index). A further analysis was undertaken using observed data alone. Results Participants Recruitment occurred between June 2000 and June 2003 with follow-up completion in June Seven hundred and twenty-eight people were referred and 516 recruited [93/212 (44%) of those who were not recruited declined to participate]. The average age of those declining participation or meeting exclusion criteria was 85 years and 28% (59/212) were male. Eleven participants were found to be ineligible because they had attended an emergency department with their index fall. Flow of participants is given in Figure 1 with cluster details in Appendix 1. Baseline characteristics are in Table 1: groups were remarkably similar, but there were a greater number of participants recruited to the secondary care arm due to differences in the underlying demography of participating practices. Appendix 2 gives details of reasons for non-completion. Appendix 3 shows the identified risk factors and triggered referrals in the primary care arm. Appendix 4 gives details of risk factors identified in secondary care arm participants. The assessment process and risk factor identification are not directly comparable between the two groups because of the design of the interventions. A quarter (47/192, 25%) of the index falls in the secondary care arm participants were multifactorial. (The Appendices are available at Age and Ageing online.) Follow-up and falls diary completion Completion rate (those who completed the 12th month of data were considered to have completed follow-up) was high and consistent (Appendix 5). Six participants did not return data for 1 month each (two in primary care arm, four in secondary care arm) and two participants did not return data for 2 months each (both in secondary care arm). Intention to treat analysis Table 2 shows an intention to treat analysis which assumes that all lost to follow-up fell or suffered an adverse event. Eighty-four per cent (133/159) of the usual care group fell again compared to 75% (158/210) in the secondary care The Winchester falls project group and 87% (118/136) in the primary care group. Secondary care intervention was more effective than usual care [adjusted OR 0.52 (95% CI ) P = 0.002] but the primary care intervention was not [adjusted OR 1.17 (95% CI ) P = 0.673]. There were no statistically significant differences between the two intervention groups and the control group for fractures, hospital admission or moves to institutional care. The secondary care group had a significantly more positive Barthel index at exit than the usual care group, but there was no significant difference for the primary care group. A Cox regression analysis for time to first fall showed a hazard ratio of in the secondary care arm (95% CI ; P = 0.012) compared to the usual care arm, but a hazard ratio of (95% CI ; P = 0.427) in the primary care arm. The mean time to fall in the secondary care group was 5.2 months (95% CI ) and in the primary care group was 3.8 months (95% CI ). Analysis based upon observed events Appendix 6 shows analysis based upon observed events. Significant reductions are shown in the secondary care group in proportions of people falling, sustaining fractures, being admitted to hospital and dying in 12 months following intervention. There is also a significant improvement in Barthel score in the secondary care group. One outcome is significantly improved in the primary care group in this analysis, but not in the secondary care group: the proportion with a timed Get up and Go score of <30 s is 82% in the primary care group compared to 72% in the usual care group [OR = 2.01 (95% CI ) P = 0.045]. Interventions in primary and secondary care groups Interventions in the usual care group were not monitored, to avoid a Hawthorne effect. The mean duration of the primary care intervention was 71 min (including risk factor review and subsequent GP/ therapist assessments). A risk factor assessment alone took an average of 21 min. The majority in the primary group required referral to one or more disciplines (87%, 118/136) with most (64%, 76/118) requiring two or more referrals. Overall 66% (90/136) were referred to the general practitioner, 39% (53/136) to an occupational therapist and 58% (79/136) to a physiotherapist. The mean duration of the secondary care intervention was 121 min for assessment by doctor, nurse, physiotherapist and occupational therapist. Interventions in the primary and secondary groups are given in Appendix 7. The most striking difference between the groups was in the proportion that had medication changes [22/135 (16%) in the primary care group versus 99/192 (52%) in the secondary care group]. A small number of participants in the primary care groups (4%, 6/135) and control group (3%, 5/159) were referred for evaluation by a geriatrician or falls clinic. 3

4 C. L. Spice et al. Assessed for eligibility 728 Excluded 212 Not meeting criteria 102 Declined 110 Usual care Primary care Secondary care Allocated 162 Allocated 141 Allocated 213 Completed study 131* Exit assessment months f-up 131 Included in ITT analysis months data 131 Fall before lost to follow-up 18 Died 17 Withdrew 10 Ineligible 3 Received intervention 140 Completed study 114 Exit assessment months f-up 114 Included in ITT analysis months data 114 Fall before lost to follow-up 16 Died 1 Died 11 Withdrew 10 Ineligible 5 Received intervention 195 Completed study 176 Exit assessment months f-up 176 Included in ITT analysis months data 176 Fall before lost to follow-up 11 Died 1 Withdrew 17 Died 10 Withdrew 6 Ineligible 3 Fall assumed 10 Fall assumed 6 Fall assumed 23 Figure 1. Trial profile for participants. 1 participant who did not withdraw and did an exit assessment but did not return the 12 th month of data is missing so is classified as not completing the study. Discussion A multi-disciplinary day hospital-based assessment reduced the absolute risk of a further fall in older recurrent fallers living in the community by 9% and reduced the time to first fall, but a nurse-led risk factor assessment with targeted referral was not effective. We found no significant differences between the groups for death, fracture, admission to hospital or institutional care on an intention to treat analysis, but using observed data alone we also found significant reductions in these outcomes in the secondary care intervention arm. The difference may be accounted for by 18 people in the secondary care intervention arm who withdrew from the trial prior to intervention. Of these, one died and three withdrew 4

5 The Winchester falls project Table 1. Baseline characteristics of participants Characteristic Usual care (n = 159) Primary care (n = 136) Secondary care (n = 210)... Mean age, years (SD) 83 (6.6) 83 (6.7) 81 (6.6) Range Male, n (%) 38 (23.9) 35 (25.7) 60 (28.6) Accommodation, n (%) Independent 131 (82.4) 111 (81.6) 180 (85.7) Warden controlled 25 (15.7) 22 (16.2) 25 (11.9) Residential home 3 (1.9) 3 (2.2) 5 (2.4) Abbreviated mental test score Median Mean Median Barthel index (range) 18 (11 20) 18 (5 20) 19 (6 20) Timed Get up and Go test >30 s, n (%) 43 (27) 35 (25.7) 51 (24.3) Able to go out alone, n (%) 86 (54.1) 64 (47.1) 124 (59) Medical history, n (%) Heart disease 54 (34) 49 (36) 90 (42.9) Stroke/TIA 32 (20.1) 30 (22.1) 51 (24.3) Hypertension 80 (50.3) 58 (42.6) 109 (51.9) Respiratory disease 30 (18.9) 19 (14) 39 (18.6) Diabetes 16 (10.1) 18 (13.2) 29 (13.8) Epilepsy 6 (3.8) 5 (3.7) 2 (1) Parkinson s disease 3 (1.9) 7 (5.1) 7 (3.3) Joint disease 99 (62.3) 66 (48.5) 144 (68.6) Neurological disease 5 (3.1) 8 (5.9) 14 (6.7) Drug history Mean total drugs (range) 4 (0 11) 5 (0 11) 5 (0 16) Four or more drugs, n (%) 95 (59.7) 90 (66.2) 144 (68.6) Diuretic, n (%) 62 (39) 57 (41.9) 100 (47.6) Hypnotic/sedative, n (%) 21 (13.2) 23 (16.9) 34 (16.2) Calcium, n (%) 4 (2.5) 11 (8.1) 8 (3.8) Calcium and vitamin D, n (%) 8 (5) 0 (0) 11 (5.2) Bisphosphonate, n (%) 7 (1.4) 11 (8.1) 8 (3.8) Number of previous falls, n (%) 2 38 (23.9) 23 (16.9) 29 (13.8) (59.7) 85 (62.5) 129 (61.4) More than 6 26 (16.4) 28 (20.6) 52 (24.8) Injury from previous falls, n (%) Fracture 29 (18.2) 15 (11) 42 (20) Laceration 33 (20.8) 33 (24.3) 55 (26.2) Bruising 138 (86.8) 110 (80.9) 164 (78.1) Index fall characteristics, n (%) Got up without help 85 (53.5) 67 (49.3) 99 (47.1) Laceration 41 (25.8) 34 (25) 51 (24.3) Bruising 135 (84.9) 110 (80.9) 159 (75.7) Possible fracture 10 (6.3) 5 (3.7) 14 (6.7) Median time on the floor (min) Range of time Interquartile range of time for medical reasons: the other 14 gave no reason, but it is possible that the requirement to travel to a hospital for intervention played a part in their decision. It would have been unethical to follow them up but it is difficult to believe that they all died within 12 months of withdrawing, or that they all suffered fractures. By assuming that they did, the intention to treat analysis may have concealed some real effects of intervention. The reduction in falls in the secondary care group is consistent with other findings and demonstrates that falls are preventable, with appropriately skilled intervention, in a high-risk group [7, 8, 15]. Our findings are supported by a further study of recurrent fallers, published since ours was conducted, achieving a reduction of 36% in the number of falls, although not of fallers, with specialist multi-disciplinary assessment [16]. At the time of the trial design, individuals presenting in primary care were not routinely referred for further assessment. Subsequent national and international guidelines recommend a specialist review for recurrent fallers, wherever they present [21 23]. A systematic review and a meta-analysis of falls prevention in older people have been published since the start of our study, and both conclude that the multifactorial assessment is beneficial in reducing falls [24, 25]. It is unclear which 5

6 C. L. Spice et al. Table 2. Conservative intention to treat analysis of outcomes assuming all missing cases have undergone a negative event (e.g. fall or death) or have not undergone a positive change (e.g. Barthel score) Usual care group, Primary care group, Secondary care group, Outcome N = 159 N = 136 N = Completed 12 months follow-up (%) 131/159 a (82) 114/136 (83) 176/210 (84) Number (%) completing 12 months follow-up or falling before exiting 149/159 (94) 130/136 (96) 186/210 (89) follow-up Number (%) falling during follow-up b 133/159 (84) 118/136 (87) 158/210 (75) Adjusted odds ratio (95% CI) OR = 1.17 ( ) OR = 0.52 ( ) Statistical significance P = P = Number (%) with fracture during follow-up 35/159 (22.0) 29/136 (21.3) 40/210 (19.0) Adjusted odds ratio (95% CI) OR = 0.85 ( ) OR = 0.90 ( ) Statistical significance P = P = Number (%) with a fall-related hospital admission during follow-up 27/159 (17.0) 17/136 (12.5) 39/210 (18.6) Adjusted odds ratio (95% CI) OR = 0.55 ( ) OR = 1.17 ( ) Statistical significance P = P = Number (%) moving to residential home during follow-up 33/159 (20.8) 26/136 (19.1) 39/210 (18.6) Adjusted odds ratio (95%CI) OR = 0.78 ( ) OR = 0.95 ( ) Statistical significance P = P = Number (%) moving to nursing home during follow-up 32/159 (20.1) 27/136 (19.1) 37/210 (17.6) Adjusted odds ratio (95% CI) OR = 0.79 ( ) OR = 0.89 ( ) statistical significance P = P = Difference from the usual care group in the mean Barthel index at 0.07 ( ) 0.63 ( ) 12 months (95% CI) Significance of difference estimated using the MIXED procedure in P = P = SPSS c Number (%) with the final Get up and Go test of 0 30 s at 12 months 113/159 (71.1) 109/136 (80.1) 158/210 (75.2) adjusted for baseline Adjusted odds ratio (95% CI) OR = 1.79 ( ) OR = 1.41 ( ) Statistical significance P = P = Baseline 116/159 (73.0) 101/136 (74.3) 159/210 (75.7) Number (%) dying during follow-up 29/159 (18.2) 23/136 (16.9) 34/210 (16.2) Adjusted odds ratio (95% CI) OR = 0.70 ( ) OR = 0.92 ( ) Statistical significance P = P = a 131 because data were missing for the 12th month for one participant who did return an exit assessment; b participants who completed 1 year of follow-up or who sustained an event before they were lost to follow-up are included; c trial arm fitted as fixed variable, others as random variables and adjusted for clustering by practice and with covariates of age, sex, baseline Barthel score, whether on four or more drugs, joint disease, calcium, calcium and vitamin D and number of previous falls. are the most effective components of multi-disciplinary assessments and what format they should take. We attempted to clarify this by directly comparing two approaches with usual care. Simple risk factor identification and subsequent appropriate referral are an attractive option as it is less time consuming and can be applied within many settings, including primary care. Two other studies of interventions similar to those in our primary care intervention arm have been unsuccessful in reducing falls, and in one of these, this may have been due to difficulties in ensuring that suggested interventions took place [9, 10]. Tinetti et al. [8] demonstrated a reduction in falls of 12% in a high-risk group but the intervention was bi-disciplinary and much more intensive and specific than the primary care intervention in this study. This study showed a difference in the identification of risk factors, for example postural hypotension, environmental hazards and visual problems. In our study, the major difference apparent between the interventions in the two groups was in review of medication and recommendation for change: this occurred in 52% in the secondary care group compared to only 16% in the primary care group. At the exit assessment, hypnotic/sedative prescription in particular was actually higher in the usual care and primary care groups than at baseline but lower in the secondary care group. A medication review was an important component of most multifactorial assessments examined by Chang and colleagues [24], and planned withdrawal of hypnotic/sedative drugs has also been found to reduce falls [26]. It is noteworthy that the percentage identified as being at risk from medication in the primary care arm (85%) and the number referred for the GP assessment (66%) were both much higher than that in whom medication change occurred (16%). What does this mean? Possibly, that reviewing and changing medication prescribed by someone else is easier than reviewing and changing medication one has prescribed oneself. This suggests that a role may be found for an independent medication review in primary care. However, although not specifically undertaken with falls prevention in mind, trials of medication review in primary care have sometimes produced results contrary to the expected [27]. 6

7 Conclusion A structured specialist multi-disciplinary assessment of recurrent fallers is effective in reducing subsequent falls but a nurse-led risk factor assessment and intervention based in the community was not. These results extend the findings of previous studies and concur with current guidelines; they indicate that recurrent fallers require assessment and management by those with appropriate skills and expertise. Key points Falls in older people often have multifactorial causes. Recurrent fallers are at an increased risk of functional decline and instistutionalisation. Multifactorial interventions can reduce falls but the benefit of different ways of delivering such interventions is not clear. This study demonstrates that some recurrent fallers can be prevented from falling again. A nurse-led risk factor assessment with targeted referrals did not reduce falls. A structured multi-disciplinary assessment and intervention can reduce falls. Acknowledgements We thank the participants and all involved staff. Conflicts of interest The authors declare there are no conflicts of interest to report. Funding Grants were received from Winchester Health Promotion Service, Shire Pharmaceuticals and Proctor and Gamble, with later funding from Mid-Hampshire Primary Care Trust. The pharmaceutical companies had no role in design, implementation, analysis, or reporting. Supplementary data Supplementary data are available at Age and Ageing online. References 1. Donald IP, Bulpitt CJ. The prognosis of falls in elderly people living at home. Age Ageing 1999; 28: Sattin R W, Lambert Huber DA, DeVito CA et al. The incidence of fall injury events among the elderly in a defined population. Am J Epidemiol 1990; 131: Vellas BJ, Wayne SJ, Romero LJ, Baumgartner RN, Garry PJ. Fear of falling and restriction of mobility in elderly fallers. Age Ageing 1997; 26: The Winchester falls project 4. Salkfeld G, Cameron ID, Cuming RG et al. Quality of life related to fear of falling and hip fracture in older women: a time trade off study. BMJ 2000; 320: Tinetti ME, Williams CS. The effect of falls and fall injuries on functioning in community-dwelling older persons. J Gerontol: Med Sci 1998; 53A: M Swift C. Falls in late life and their consequences implementing effective services. BMJ 2001: 322; Close J, Ellis M, Hooper R, Glucksman E, Jackson S, Swift C. Prevention of falls in the elderly trial (PROFET): a randomized controlled trial. Lancet 1999; 353: Tinetti ME,Baker DI,McAvay G et al. A multifactorial intervention to reduce the risk of falling among elderly people living in the community. N Engl J Med 1994; 331: Van Haastregt JC, Diederiks, JP, vanrossum E, dewitte LP, Voorhoeve PM, Crebolder HF. Effects of a programme of multifactorial home visits on falls and mobility impairments in elderly people at risk: randomized controlled trial. BMJ 2000; 321: Lightbody E, Watkins C, Leathley M, Sharma A, Lye M. Evaluation of a nurse-led falls prevention programme versus usual care: a randomized controlled trial. Age Ageing 2002; 31: Hodkinson HM. Evaluation of a mental test score for assessment of mental impairment in the elderly. Age Ageing 1972; 1: Mahoney FI, Barthel DW. Functional evaluation: the Barthel index. Md State Med J 1965; 14: Mathias S, Nayak USL, Isaacs B. Balance in elderly patients: the Get up and Go test. Arch Phys Med Rehabil 1986; 67: ICH. International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use Harmonised Tripartite Guidelines: Statistical Principles for Clinical Trials Recommended for Adoption at Step 4 of the ICH Process on 5 February 1998 by the ICH Steering Committee, Day L, Fildes B, Gordon I, Fitzharris M, Flamer, Lord S. Randomised factorial trial of falls prevention among older people living in their own homes. BMJ 2002; 325: Davison J, Bond J, Dawson P, Steen IN, Kenny RA. Patients with recurrent falls attending Accident & Emergency benefit from multifactorial intervention a randomised controlled trial. Age Ageing 2005; 34: Huber PJ. The behavior of maximum likelihood estimates under nonstandard conditions. In: Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Vol. 1. Berkeley, CA: University of California Press, 1967; White H. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 1980; 48: Rogers, WH. Regression standard errors in clustered samples. Stata Tech Bull 1993: 13: Peugh JL, Enders CK. Using the SPSS Mixed procedure to fit cross-sectional and longitudinal multilevel models. Educ Psychol Meas 2005; 65: American Geriatrics Society, British Geriatrics Society and American Academy of Orthopaedic Surgeons panel on Falls Prevention. Guideline for the prevention of falls in older persons. J Am Geriatr Soc 2001; 49:

8 C. L. Spice et al. 22. Department of Health. National Service Framework for Older People. London: DoH, National Institute for Clinical Excellence. Falls: the assessment and prevention of falls in older people. Clinical Guideline 21, Chang JT, Morton SC, Rubenstein LZ et al. Interventions for the prevention of falls in older adults: systematic review and meta-analysis of randomised clinical trials. BMJ 2004; 328: Gillespie LD, Gillespie WJ, Robertson MC, Lamb SE, Cumming RG, Rowe BH. Interventions for preventing falls in elderly people. Cochrane Database Syst Rev 2003; John Wiley & Sons Ltd. DOI: / CD Campbell AJ, Robertson C, Gardner MM, Norton RN, Buchner DM. Psychotropic medication withdrawal and a home-based exercise program to prevent falls: a randomized, controlled trial. J Am Geriatr Soc 1999; 47: Holland R, Lenaghan E, Harvey I et al. Does home based medication review keep older people out of hospital? The HOMER randomised controlled trial. BMJ 2005; 330: