Multifactorial intervention after a fall in older people with cognitive impairment and dementia presenting to the accident and emergency department: randomised controlled trial

Design

We conducted a prospective single centre randomised controlled trial of multifactorial assessment and intervention after a fall compared with assessment followed by conventional care. We recruited patients from two inner city accident and emergency departments, which were screened for 52 weeks. Our study was approved by Newcastle and North Tyneside Health Authority ethics committee.

At baseline interviews we recorded personal details, previous falls, current medical diagnoses, and research criteria for dementia according to the ICD-10 (international classification of diseases, 10th revision).7 We sought consent or permission for entry into the study from three people: patient, immediate carer, and next of kin, following ethical principles.8 The patients underwent multifactorial clinical assessment (medical, physiotherapy, occupational therapy, and cardiovascular) at baseline. Those randomised to the intervention group received intervention for all identified risk factors for falls.

We collected data on falls, injuries, attendance at accident and emergency department, hospital admission, and mortality prospectively for 1 year. At 3 months we repeated physiotherapy and occupational therapy assessments and cardiovascular tests if they were abnormal. We also assessed compliance with the multifactorial intervention at 3 months.

Outcome measures

The primary outcome measure was number of participants who fell at least once in the year after intervention. Secondary outcome measures were number of falls (corrected for diary returns), time to first fall, injury rates, fall related attendance at accident and emergency department, fall related hospital admissions, and mortality.

We recorded data on falls prospectively by asking informants to complete a weekly diary in the form of a postcard. If we received no postcard for two consecutive weeks, we contacted the informants by telephone. We collected additional data on injuries from accident and emergency department records. We obtained data on attendance at accident and emergency department and hospital admission at monthly intervals from computerised records.

Clinical assessment and intervention protocol

Table 1 outlines the multifactorial assessment and intervention protocol. Participants in the control group received conventional care from all health professionals who were or became involved in their management during the year's follow up.

Cooperation with each component of assessment was judged clinically. Compliance with intervention was assessed at 3 months by direct questioning of the informant, inspection of drugs, review of physiotherapy documentation, and observation of the environment and patient. The same criteria were used to assess treatment received by controls.

Statistical analysis

Our sample size was based on the primary outcome measure of the number of patients who fell during the year's follow up. We judged a 30% reduction in the proportion of patients who fell over 1 year as clinically significant and similar to that of cognitively normal older people. 14 15 We estimated that at the 5% level we would require 90 patients in each group to give an 80% power of detecting a reduction of 30% (from an estimated 66%1 to 46%) in the proportion of patients who had at least one fall. The actual proportion of the control group that fell was 80%. The sample size of 90 participants per group therefore gave a 90% power at the 5% level of detecting a 30% reduction (from 80% to 56%) in the proportion of patients who fell.

We used SPSS statistical software for our analysis. We calculated the weekly rate of falls for each patient by dividing the number of recorded falls by diary returns and used this when comparing differences in number of falls between control and intervention groups. We compared categorical variables between groups with Fisher's exact test, and we calculated a relative risk ratio. For variables that were at least ordinal, we used the Mann-Whitney U test to compare independent groups, and we calculated an estimated mean difference. For normally distributed data we used the independent samples t test. We used the log rank test to compare the two groups for time to first fall. We analysed data on an intention to treat basis.

Group assignment and blinding

We randomised patients by block randomisation using computer generated random numbers either to assessment plus targeted multifactorial intervention (intervention group) or to assessment plus conventional care (control group). Randomisation was stratified by mini-mental state examination score at study entry: 20 to 23 (mild cognitive impairment); 12 to 19 (moderate cognitive impairment); 4 to 11 (severe cognitive impairment); 0 to 3 (very severe cognitive impairment). Group allocation was performed by a researcher who was independent of the recruitment process and blind to baseline interview data. Data from the postcards (primary outcome) were processed and coded off site by a researcher who was blind to group allocation and otherwise unconnected with the study. Data on secondary outcomes, compliance with intervention, treatment received by control group, and objective effects of intervention, were by necessity recorded and coded by members of the study team, who were not blind to randomisation.

Baseline characteristics and identified risk factors for falls

Table 2 shows the baseline characteristics of the patients. Overall, 73% (n=199) of participants cooperated with multifactorial assessment (medical 86% (236), cardiovascular 73% (201), physiotherapy 78% (214), occupational therapy 79% (216)). We identified 1011 risk factors for falls (485 in intervention group, 526 in control group); for both groups, median number 4 (interquartile range 3-5; table 3). We found no significant differences in baseline characteristics or risk factors for falls between intervention and control groups.

Effect of intervention

The intervention group had 652 falls and the control group 728 falls during the year's follow up. Table 4 shows the results of the intention to treat analysis. We found no significant differences between the intervention and control groups for any study outcomes.

Implementation of intervention

Objective effects of intervention were shown for the risk factors of gait, environmental fall hazards, and cardioinhibitory carotid sinus hypersensitivity (table 5). In patients who had orthostatic hypotension on active standing at baseline, at 3 months' follow up those in the intervention group showed a smaller fall in blood pressure on active standing than did controls, but the difference was not significant. Table 6 details the proportion of the intervention group complying with key components of the multifactorial intervention strategy at 3 months' follow up (when all interventions were completed) and the proportion of control patients receiving treatment meeting criteria for compliance with these interventions.

Strengths and limitations

It is unlikely that contamination of the control group was a major factor in the negative outcome. The control and intervention patients lived in the same residential or nursing homes, and all received a comprehensive assessment of risk factors at baseline. However, although 14% of control patients had changes in drugs meeting criteria for compliance with intervention, for all other key risk factors less than 10% of control patients received intervention outwith the study. Fall rates in the control group were equivalent to the highest reported rates of nursing home populations including residents with dementia. 2 17 Poor compliance with intervention was anticipated, yet compliance with key interventions at 3 months was similar to that reported in studies where multifactorial intervention had been successful in preventing falls. 15 16

Our study is limited by lack of generalisability—there was relative under-recruitment of participants from the community, and recruitment was from a specific population in a single centre. A further limitation is that effective single blinding was feasible for only the primary outcome measure (number of participants who fell) and the secondary outcome measures of number of falls and time to first fall.

Neurodegenerative process

The neurodegenerative process is a probable source of much of the increased risk of falls in patients with cognitive impairment and dementia. Studies of postural instability have found particular impairments in patients with Alzheimer's disease when central processing of conflicting information is required, and neuropathological studies show degeneration in parts of the brain controlling cardiovascular (autonomic) reflexes. 18 19 A high prevalence of autonomic dysfunction, manifest as orthostatic hypotension and cardioinhibitory carotid sinus hypersensitivity has been shown in older people with dementia.20 It may be that for effective prevention of falls in patients with cognitive impairment and dementia, different strategies are required to those used in cognitively normal older people. For example, physiotherapy may need to be delivered for longer, and intervention to modify cardiovascular risk factors may be relatively more important.

Practical implications and conclusion

The recommendations for fall prevention within the national service framework for older people are broadly similar to a guideline issued in 2001 by a joint panel of the American Geriatrics Society, British Geriatrics Society, and American Academy of Orthopaedic Surgeons.21 Implementation of these recommendations will have considerable resource implications for the NHS in the United Kingdom. It seems appropriate, particularly in areas where effective services for falls do not yet exist, to target scarce resources where benefit is proved. Research data support multifactorial intervention in cognitively normal older people living in the community with risk factors for falls and those who present to the accident and emergency department after a fall. 14 15 Several studies also support the use of exercise as a single intervention in cognitively normal older people living in the community. 22 23 Although one randomised controlled trial of multifactorial intervention showed a reduction in recurrent falls in residents of nursing homes, most studies in this setting have been unable to prevent falls. 6 16 24

We did not show a significant reduction in the proportion of patients who fell or in number of falls by using a multifactorial assessment and intervention strategy after a fall in patients with cognitive impairment and dementia presenting to the accident and emergency department. This suggests multifactorial intervention after a fall is less effective in these patients than in cognitively normal older people. Limited resources may be used more effectively if targeted towards cognitively normal older people who fall. However, as older people with cognitive impairment and dementia are at particularly high risk of falls and their associated morbidity, it is important that prevention of falls remains a research priority in this patient group. Further work is required in patients with cognitive impairment and dementia who fall to determine optimal delivery of interventions and to identify the most important modifiable risk factors.