Fall risk factors in elderly patients with cognitive impairment on rehabilitation wards


Dr Michael Vassallo FRCP PhD, Royal Bournemouth Hospital, Castle Lane East, Bournemouth BH7 7DW, UK. Email: michael.vassallo@rbch.nhs.uk


Background:  Confusion and cognitive impairment, are risk factors for falls in hospital. Evidence for reducing falls in cognitively-impaired patients is limited and to date no intervention has consistently been shown to reduce falls in this group of patients. We explored characteristics associated with falls in cognitively-impaired patients in a rehabilitation setting.

Methods:  In a prospective observational study, 825 consecutive patients were studied. Patient characteristics were assessed on admission. Factors predisposing to falls in cognitively-impaired patients were identified.

Results:  Cognitively-impaired patients were more likely to be fallers or recurrent fallers and more likely to sustain an injury than cognitively intact patients. They had a higher incidence of nursing home discharges and a significantly higher mortality. Logistic regression analysis showed that an unsafe gait (P < 0.001; 95% confidence interval, 0.13–0.57) was the only independent risk factor for falls in this group of patients. There was a cumulative higher risk of falling associated with an unsafe gait demonstrable throughout the patients' stay.

Conclusion:  Unsafe gait was the only significant independent risk factor for falls among cognitively-impaired patients in a rehabilitation environment. Interventions that improve gait patterns or that enhance safety for patients with abnormal gait are required if fall reduction in this group of patients is to be achieved.


Confusion and cognitive impairment, are well recognized risk factors for falls in both the community1 and in a hospital environment.2 Patients with dementia are as much as three times more likely to fall than non-demented patients.3,4 This might be partly due to a lack of insight into environmental dangers and a failure to comply with advice about safety or medical treatment. However, damage to central nervous system balance mechanisms often leads to specific problems with postural stability and autonomic reflexes.5 In addition, these patients often have behavioral disorders and take medications that can also predispose to falling.6,7

Evidence for reducing falls in cognitively-impaired patients is limited and often contradictory. Many studies on fall prevention measures were community based and excluded patients with cognitive impairment.8 Studies of cognitively-impaired patients alone have shown that multidisciplinary intervention was ineffective at reducing falls in a group of older patients admitted to an accident and emergency department after a fall.9 On the other hand, a very small study of focused supervision of high risk patients in a dementia unit suggested that such a strategy might lead to less falls.10 The role of medication manipulation is also unclear. While tranquillizer use is generally perceived to increase the risk of falls, Katz et al. suggested that using risperidone reduced falls in a group of patients with dementia.11

In hospital, patients with cognitive impairment contribute the majority of falls and very often are recurrent fallers.12 They are also more likely to fall within the first week of admission.8 Hospital-based studies that have included cognitively-impaired patients suggest that multidisciplinary interventions including physiotherapy had a marginal effect on reducing falls.13–15 However, it is unclear whether these interventions were effective at reducing falls in the cohort of cognitively-impaired patients alone. In order to optimize fall reduction interventions in this group, a better understanding is required of fall risk factors. We aimed to study what are the risk factors for falls in cognitively-impaired patients and how they differ from non-cognitively-impaired patients. We also aimed to explore the effect of impaired cognition on nursing home admissions and mortality.



In a prospective, open, observational study, we followed 825 patients admitted for rehabilitation in a community hospital in the UK.12 The community hospital admitted elderly patients for rehabilitation and continuing medical care after an acute medical illness or surgery including orthopedic surgery. Ethical approval was obtained from the local ethics committee. Patients were admitted to three wards without selection by administrative staff that were not involved in the study. No changes in the admission pattern were made for the purposes of the study. Referrals for admission were put on a waiting list and then allocated a ward in sequence, as beds became available. This process of admission was considered to be quasi-random. There were no exclusion criteria. We followed up 825 consecutive patients admitted to the hospital.


All wards had a similar design, equipment and safety features including the same number of bed and chair alarms, staffing levels, training and operational policies.16–18 Patient characteristics were measured routinely in all patients admitted to the wards. These included past history of falls, medication including tranquillizers, antihypertensive agents (a heterogenous group of drugs including diuretics, calcium channel blockers, angiotensin-converting enzyme inhibitors, angiotensin II receptor antagonists, centrally acting antihypertensives, beta-blockers and alpha-blockers), antiparkinsonism drugs and antidepressants. We also assessed visual and hearing impairment. Visual impairment was evaluated using a Snellen chart. Patients were deemed to be visually impaired if they had a vision of 6/60 or less. Hearing impairment was determined by the ability to follow a conversation during the research interview.19 Patients were allowed to wear glasses or hearing aids if they were normally worn. Abnormal lower limbs were defined as any abnormality of the limb judged to interfere with mobility. This included abnormalities such as hemiplegia, neuropathy or cellulitis. Abnormalities that did not interfere with safe gait such as mild arthritis were not considered as an abnormality for the purposes of this study. To assess cognitive function, an abbreviated mental test score was performed and scores of less than 7/10 were considered to constitute cognitive impairment.20 A “get up and go test” was performed to assess gait. The patient's performance was graded by a trained assessor as safe (no or low risk of falling) or unsafe (high risk of falling).21

In addition to the above assessment, staff on one of the wards developed a multidisciplinary team (MDT) to adopt a more proactive approach to fall prevention. This consisted of a physician, nurse, occupational therapist (OT), social worker and physiotherapist (PT). A fall risk assessment and case conference were conducted weekly. The aim of the team was to formulate a tailored falls prevention care plan depending on the risk factors identified. This variable practice was corrected for in the analysis section.

The outcome measures recorded were the number of fallers (i.e. patients who sustained at least one fall), number of recurrent fallers (i.e. patients who sustained more than one fall) and number of patients sustaining an injury. All falls were recorded through an incident reporting system. It was a statutory requirement to record all falls on a standard accident form. There was no evidence that fall reporting was incomplete. All patients were required to have an assessment of the circumstances of the fall and a physical examination by a doctor to record any injuries. An injury was recorded when it was at least of minor severity resulting in bruising, cuts or fractures. An investigator not involved in the intervention and blinded to the risk and intervention status of the patient recorded these outcomes. Secondary outcome measures recorded were place of discharge and mortality.


Data was collected from the time of admission to the time of discharge. The χ2-test or Fisher's exact probability test were used as appropriate to compare the number of fallers, recurrent fallers and injured patients among cognitively-impaired and non-cognitively-impaired patients. Multiple logistic regression analysis was performed to analyze the association between characteristics of cognitively-impaired patients and falls using fallers as the dependent variable and correcting for variables that differed across the groups including differing work practices. The Kaplan–Meier statistic was used to evaluate the effect of risk factors on falls.


The demographic details of the patients are shown in Table 1. Cognitively-impaired patients were older (83.1 vs 81.1 years; P < 0.001) and had a longer duration of stay in hospital (28.1 vs 23.0 days; P = 0.002). They were more likely to be taking antiparkinsonism medication and tranquillizers but less likely to be on antihypertensives, They were also more likely to have an unsafe gait, hearing impairment and a past history of falls. Fallers, recurrent fallers and fall-related injury were significantly higher among the cognitively-impaired group. They also had a higher incidence of nursing home discharges and a significantly higher rate of mortality (Table 2).

Table 1.  Patient demographics
CharacteristicCognitively impairedNon-cognitively impairedP-value
  1. SD, standard deviation.

Sex (male)1131800.62
Mean age, years (SD)83.1(6.9)81.1(7.6)<0.001
Duration in hospital, days (SD)28.1(22.9)23.0(21.3)0.002
Past history of falls (%)253(76.9)309(62.3)<0.001
Antihypertensives (%)169(51.4)309(62.3)0.002
Antidepressants (%)38(11.5)57(11.5)0.911
Antiparkinsonism (%)25(7.6)18(3.6)0.016
Tranquillizers (%)99(30.1)80(16.1)<0.001
Unsafe gait (%)258(78.4)341(68.7)0.002
Hearing (%)99(30.1)117(23.6)0.043
Abnormal limb (%)215(65.3)312(62.9)0.506
Vision (%)63(19.1)73(14.7)0.103
Table 2.  Different outcomes between cognitively-impaired and non-cognitively impaired patients (Fisher's exact probability test using categorical data)
CharacteristicNon-cognitively impaired
n = 496
Cognitively impaired
n = 329
P-valueOdds ratio95% CI
  1. CI, confidence interval.

Injurious fall1937<0.0013.181.79–5.60
Recurrent falls1137<0.0015.582.81–11.1
Nursing home admission6072<0.0012.031.39–2.96

There were no differences in the risk factors of cognitively-intact and cognitively-impaired fallers (Table 3). Cognitively-impaired patients who fell were significantly more likely to have an unsafe gait when compared to cognitively-impaired non-fallers. Mortality was significantly higher in cognitively-impaired fallers compared to non-fallers (Table 4). Multiple logistic regression analysis of risk factors for falls using a forward conditional model was used. Covariates of P ≤ 0.5 (i.e. past history of falls, P = 0.07; sex, P = 0.19; antihypertensives, P = 0.34; tranquillizers, P = 0.27; and ward, 0.21) showed that an unsafe gait alone (P = 0.0001; 95% confidence interval [CI], 0.12–0.57) was an independent risk factor for falls in cognitively-impaired patients The ward that had a multidisciplinary program for fall prevention had significantly more men than non-intervention wards (42.9% vs 31.8%; P = 0.007). We identified 102 cognitively-impaired patients in this ward and there was a trend towards less falls among these patients (24/102 [23.5%] vs 75/227 [33.0%]; P = 0.11). We performed a similar multiple logistic regression analysis on these patients separately and unsafe gait again was the only independent risk factor for falls among this group (P = 0.002; 95% CI, 0.013–0.76). A Kaplan–Meier analysis revealed that patients with an unsafe gait in addition to confusion were most likely to have a fall over a period of 50 days post-admission (Fig. 1).

Table 3.  Differing characteristics of fallers between cognitively-impaired and non-cognitively-impaired patients
CharacteristicNon-cognitively impairedCognitively impairedP-valueOdds ratio95% CI
  1. CI, confidence interval; SD, standard deviation.

Sex (male)18380.850.920.59–1.44
Mean age, years (SD)82.6 (6.9)82.6 (6.4)0.99  
Past history of falls37 (6.9)83 (83.8)0.160.850.72–1.05
Antihypertensives35 (72.5)55 (55.5)–1.50
Antidepressants9 (17.6)13 (13.1)0.621.340.62–2.93
Antiparkinsonism4 (7.8)6 (6.1)0.790.940.68–1.37
Tranquillizers11 (21.5)35 (35.3)0.120.610.33–1.09
Diuretics27 (52.9)38 (38.4)0.131.380.96–1.98
Unsafe gait46 (90.2)90 (90.1)0.880.990.88–1.11
Hearing20 (39.2)30 (30.3)0.361.290.82–2.03
Abnormal limb36 (70.5)63 (63.9)0.961.030.82–1.28
Vision10 (19.6)16 (16.2)0.761.210.59–2.48
Nursing19 (37.2)24 (24.2)–1.68
Mortality8 (15.6)30 (30.3)0.080.520.25–1.04
Table 4.  Differing characteristics of fallers and non-fallers among cognitively-impaired patients
  1. CI, confidence interval; SD, standard deviation.

Sex (male)37760.19
Mean age, years (SD)82.6 (6.4)83.1 (7.12)0.96
Past history of falls (%)83 (83.8)170 (73.9)0.07
Antihypertensives (%)55 (55.5)114 (49.5)0.34
Antidepressants (%)13 (13.1)25 (10.8)0.57
Antiparkinsonism (%)6 (6.1)19 (8.2)0.50
Tranquillizers (%)35 (35.3)64 (27.8)0.27
Diuretics (%)38 (38.4)97 (42.1)0.54
Unsafe gait (%)90 (90.1)168 (73.0)<0.001
Hearing (%)30 (30.3)69 (30.0)0.95
Abnormal limb (%)63 (63.6)147 (63.9)0.45
Vision (%)16 (16.2)47 (20.4)0.64
Nursing (%)24 (24.2)48 (20.1)0.54
Mortality (%)30 (30.3)44 (19.1)0.03
Figure 1.

Time to first fall from admission. Group 1, confusion and unsafe gait. Group 2, confusion alone. Log–rank test, P < 0.001.


Although confusion has long been recognized as a risk factor for falls, risk factors among confused patients have to our knowledge not been studied.3,4,22 While there is evidence that multidisciplinary interventions can have an effect in reducing falls in cognitively-intact patients, preventing falls among confused patients remains a challenge.23 Many studies that identify interventions successful at preventing falls often exclude cognitively-intact patients24 It is therefore important to try to separately investigate fall risk factors in confused patients to help formulate fall prevention strategies specific to confused patients.

This study confirms that cognitively-impaired patients are the most prone to have adverse outcomes. Patients with cognitive impairment had an increased risk of falls, injurious falls and recurrent falls. They also had increased nursing home discharge and higher mortality. We also found that cognitively-impaired fallers had higher mortality than cognitively-impaired non-fallers. This is not a surprising finding and is a reflection of the increased frailty and care needs required by these patients. The presence of an unsafe gait was the only independent risk factor of falls in this group of patients. Although some caution is necessary because some of the variables in the multiple logistic analysis are not independent, the result is very important clinically. This is a characteristic that is easily identified, enabling interventions that may reduce the risk of falls in this group of patients. The Kaplan–Meier statistic also demonstrated the additional cumulative effect of unsafe gait on the risk of falls in cognitively-impaired patients.

We explored whether the risk factors for falls differ between cognitively-impaired and cognitively intact patients. None of the risk factors that we studied did in fact differ. These data do not suggest that a different form of risk assessment for confused patients is warranted. A shortcoming of this study is that the number of risk factors studied was limited. More research therefore is needed and should include behavioral risk factors such as wandering or agitation.

Fall reduction has so far proved elusive in cognitively-impaired patients. Unsafe gait was the only significant independent risk factor for falls in this group of patients in a rehabilitation environment. Interventions that improve gait patterns or enhance safety for patients with abnormal gait are required if fall reduction in this group of patients is to be achieved.