1 Background

Injuries due to falls remain a concern for inpatient safety. According to the newest Cochrane database, approximately 30 % of people aged over 65 years living in the community fall each year [1]. Fall incidence in nursing homes is reported to be about three times that in the community, equating to rates of 1.5 falls per bed per year (range 0.2–3.6) [2, 3]. In hospital, on the other hand, an incidence of 3.4 falls per person-year has been reported in geriatric rehabilitation wards, and 6.2 falls per person-year in psychogeriatric wards [4, 5]. In spite of more intense risk management, the high number of accidental falls in hospital is a severe problem.

Several studies have demonstrated that some kinds of medications contribute to falls [1, 2, 4]. Benzodiazepines and hypnotics [6–9], antidepressants [10–12], anti-hypertensives and diuretics [13, 14], narcotics [8], and anti-Parkinson’s [15] drugs are reported as risk factors of falls. We began to investigate the association between accidental falls and medication in wards, and we found that only zolpidem, the ω1-selective non-benzodiazapine, showed the lowest odds ratio (OR) of falls in hypnotics tested.

Hypnotic drugs are frequently used for insomnia (with symptoms such as difficulty falling asleep or staying asleep, awaking too early in the morning, and disturbance in sleep quality), and they may cause falls because of their effect on psychomotor activity. Therefore, appropriate selection of hypnotics and assessing the related risks might be important in the prevention of accidental falls. Short-acting non-benzodiazepines are known to be relatively safe hypnotics and are widely used to treat difficulty in falling asleep. In 1999, Rudolph et al. [16] reported that the myorelaxant, motor-impairing, ethanol-potentiating, and anxiolytic-like properties of diazepam were not mediated by α1 gamma-aminobutyric acid (GABA)A receptors, but might be mediated exclusively by α2, α3, and/or α5 GABAA receptors. In 2008, Hanson et al. [17] reported that in cases of sedative/hypnotic activity of benzodiazapine receptor [BZ (ω)] agonists, determined by the ratio of selectivity in ω1/ω2 receptor subtypes, the difference in ω1/ω2 selectivity may lead to a difference in falling probability. However, the association between falling related to taking hypnotics and the ω1/ω2 selectivity of each hypnotic was not clearly established.

In this study, we assessed the falling frequency of inpatients admitted to a ward of Gunma University Hospital, to clarify the association between the risk of falling and the medication, particularly hypnotics.

2 Methods and Study Design

Gunma University Hospital is a general hospital with 725 beds in 15 medical departments. This study included all hospitalized patients; there were no exclusion criteria regarding disease or age. Medical records were obtained from 3,683 unrelated Japanese hospitalized patients (1,965 males and 1,718 females; mean age 56.5 ± 18.6 years) from October to December 2007 at Gunma University Hospital. Medical record analysis was approved by the Ethical Review Board in Gunma University Hospital. Inpatient falls are regularly registered via incident and accident reports submitted by medical staff. Falls were evaluated according to a previous report by Gibson [18], and medical charts were reviewed to obtain inpatient data. All drugs prescribed for the patients during their stay in hospital were extracted electronically from hospital charts. The frequency of falling was compared among drugs classified as hypnotics, antiepileptics, opioids, anti-Alzheimer’s, anti-Parkinson’s, antipsychotics, antidiabetics, antihypertensives, and antiarrhythmics, according to the therapeutic category of drugs defined by the Japanese Ministry of Health and Labor Welfare.

3 Statistical Methods

The medical staff in the ward (nurses, pharmacists, and medical doctors) who found the fall accident or was informed about one by a patient completed an incident sheet. The data were calculated as the number of patients who experienced one fall divided by the total number of inpatients. The Student’s t test was used for comparison of age, and a chi-square analysis was used for comparison of sex difference. The relationship between medication and the prevalence of falls was estimated by comparing the odds of exposure among patients who fell during hospitalization. A logistic regression analysis with a stepwise procedure was used to identify the independent risk factors of falling as reported by Tanaka et al. [19]. The OR and corresponding 95 % confidence interval (CI) were calculated using logistic regression in StatView (SAS, Cary, NC, USA) software. Finally, a multiple logistic regression analysis, adjusted for use of diuretics and anticoagulants was performed to evaluate the odds of falling for each drug. The full model included age and the use of zolpidem, brotizolam, zopiclone, triazolam, flunitrazepam, nitrazepam, estazolam, antiepileptics, opioids, anti-Alzheimer’s, anti-Parkinson’s, antidiabetics, antihypertensives, and antiarrhythmics. The relationship between OR and ω1/ω2 selectivity reported previously [20–42] was explored using linear regression analysis. Statistical significance was set at p < 0.05.

4 Results

4.1 Characteristics of Patients and Fall Rate

Falling accidents were reported for 116 (3.1 %) of the 3,683 inpatients during hospitalization. Mean age on admission was 56.5 ± 20.2 years. The age of inpatients experiencing a fall (64.7 ± 19.5) was significantly higher (p < 0.001, Student’s t test) than those who did not fall (56.2 ± 20.2). In male patients, the proportion experiencing a fall (67/116, 57.8 %) did not differ from those who did not fall (1,898/3,567 [53.2 %]; p = 0.33, Chi-square test).

4.2 Falling Risk of Medication

Multiple logistic regression analysis showed a significant relationship between risk of inpatient falls and several drug groups, such as hypnotics, antiepileptics, opioids, anti-Alzheimer ’s, anti-Parkinson’s, antidiabetics, antihypertensives, and antiarrhythmics (Table 1). Sex and antipsychotics were not risk factors for falling. In the analysis adjusted for use of diuretics and anticoagulants, brotizolam, zopiclone, and estazolam showed a significant increase in the risk of inpatient falls (p < 0.001, p = 0.011 and 0.013, respectively), while zolpidem, triazolam, flunitrazepam, and nitrazepam did not show any difference (p = 0.315, 0.416, 0.327, and 0.446, respectively) (Table 2).

Table 1 Relationship between fall frequency and selected variables in hospitalized patients
Full size table
Table 2 Relationship between hypnotics and selected variables of fall frequency in hospitalized patients
Full size table

5 Discussion

Risk factors for falls have been reported [43, 44] to include age, sensorial impairments, various pathologies (e.g., anemia, neoplasms, congestive heart failure, and stroke); environmental and staff conditions; gait instability; limb weakness; urinary incontinence, frequency, or need for assisted toileting; agitation; confusion; impaired judgment; and prescription of high-risk drugs, such as sedative hypnotics. Barbiturates and benzodiazepines promote sleep by binding to and allosterically modulating GABAA receptors in the central nervous system. However, these drugs have been associated with several adverse reactions, including alteration of sleep architecture, nightmares, agitation, confusion, lethargy, withdrawal, and a risk of dependence and abuse. The newest generation of sleep-aid drugs, the non-benzodiazepine hypnotics such as zolpidem, was developed to overcome some of these disadvantages [45].

In this study, only zolpidem, the most ω1/ω2-selective agent, showed an OR of <1 (Table 2). Non-benzodiazepine drugs, including zolpidem, act through a similar neural mechanism as classical benzodiazepines. They bind to the same site on the GABAA receptors but differ significantly in their chemical structure and neuropharmacological profile [46–48]. GABAA receptors have a pentameric form comprising 19 subunits (α1-6, β1-3, γ1-3, δ, ε, θ, π, and ρ1-3) [24, 49, 50]. The benzodiazepine binding site is now known to be associated with α and γ subunits. The pharmacologically defined benzodiazepine receptor subtype BZ1 (ω1) seems to correspond to the GABAA receptors containing α1 subunits, whereas the BZ2 (ω2) subtype is heterogeneous and corresponds to GABAA receptors with α2, α3, or α5 subunits [51, 52]. GABAA receptors containing different α subunits show a heterogenous distribution in the brain, and it has been suggested that different receptor subtypes may have different functional roles [53]. In case of sedative and hypnotic activity of BZ (ω) agonists, determined by the ratio of selectivity in ω1/ω2 receptor subtypes, the difference in ω1/ω2 selectivity may influence the difference in falling probability [17].

Another possible reason for the variance in the risk of falls is the difference in the pharmacokinetics of hypnotics. Zolpidem has the shortest elimination half-life and carries the lowest risk of falling. The maximum plasma concentration of zolpidem is reached 1.5 h after dosing [30]. A shorter time to reach peak concentration and a short elimination half-life may be preferable characteristics for hypnotic agents. A considerable number of accidental falls occur when a patient wakes because of a micturition urge during night. Thus, for patients with insomnia, it is important to select a hypnotic with a short half-life to avoid excessive suppression of psychomotor activity after sleeping.

Finally, low-risk drug–drug interactions could explain the low frequency of falls in patients taking zolpidem. Although the formation of alcohol derivatives of zolpidem is rate-limiting and mediated principally by cytochrome P450 (CYP)3A4 (about 60 %), the rest is metabolized by CYP1A2, CYP2C9, CYP2C19, and CYP2D6 [54, 55]. Such metabolic profiles could reduce the risk of drug–drug interaction and genetic polymorphism compared with other hypnotics, such as triazolam, which is metabolised by only CYP3A4 [56, 57]. The difference in metabolic profiles may contribute to the low risk of falling with zolpidem, even when patients are concurrently administered several drugs that inhibit the metabolic pathway of zolpidem. This is especially valid for elderly patients, most of whom receive polytherapy, which increases the risk of drug–drug interaction. Consequently, genetic analysis may be a useful tool for the prevention of falls related to medications, particularly hypnotics.

In this study, we evaluated the association of falling with medication but not the medical conditions or disease of patients. Although we clarified the difference in the risk of falling among hypnotics, in future, we should also establish the relationship between the time when falls occur, drug dosage, and medical condition or disease.

6 Conclusion

Our results show that many falls depend on the type of hypnotic agent in inpatients with insomnia. In order to clarify the correlation between each hypnotic and the risk of falling, it is still necessary to evaluate the time of taking drugs and falling accident. Falls are a common risk for all inpatients. Reduction in the number of falls and related injuries is important for maintaining patient quality of life and for reducing medical costs. However, the risk of falls is not able to be predicted from ω1/ω2 selectivity. The relationship between falling and the profiles of various hypnotics remains to be analyzed.