FormalPara Key Points

Use of benzodiazepine receptor agonists (BZRAs) was high among hospitalized older adults with chronic conditions and medications.

A quarter of BZRA users had ceased their medication 6 months after admission.

Several categories of patients require specific attention because of the higher prevalence of BZRA use (female patients and people using other psychotropic medications) or a decreased likelihood of BZRA cessation (chronic obstructive pulmonary disease co-morbidity).

1 Introduction

Benzodiazepine receptor agonists (BZRA, i.e., benzodiazepines and z-drugs) have shown an unfavorable risk (e.g., falls, hip fractures, delirium)–benefit (reduced anxiety, reduced time to fall asleep) balance. This is especially remarkable in older adults, with a number needed to treat of 13 for improved sleep quality and a number needed to harm of six for all adverse events [1]. Widely used explicit criteria on potentially inappropriate medications for older adults [2, 3] recommend avoiding BZRAs, especially in those with a history of falls or for those using three or more central nervous system (CNS)-acting drugs. Moreover, although it is widely accepted that BZRA use should be limited to a maximum of 4 weeks, long-term use is often observed [4]. Despite these recommendations, the prevalence of BZRA use in EU countries remains high, as illustrated by data from the EU-PROTECT project [5] and the OECD database (25 over 1000 adults over 65 years old have prolonged use) [6], although there are significant variations between countries. Even higher prevalence rates are observed in acute care settings (up to 49%) [7,8,9].

Deprescribing is the process of withdrawal (cessation or dose reduction) of an inappropriate medication, supervised by an appropriate healthcare professional, with the goals of managing polypharmacy and improving health outcomes [10]. Deprescribing should always be tailored to the individual patient. Evidence shows that BZRA deprescribing can reduce the risk of harms associated with chronic BZRA use, without significantly worsening sleep quality, anxiety, or depression [11]. Deprescribing interventions targeting BZRAs in older adults have achieved success rates ranging from 27 to 80%—depending on the intervention tested and the setting [12]. In their systematic review, Reeve et al. describe interventions that target the older adult, the caregiver, and/or various healthcare professionals and include pharmacological substitution, patient educational material, or psychological support [12]. Although BZRAs are considered one of the top priorities for deprescribing in older adults, we still need to identify factors associated with BZRA cessation to better frame deprescribing interventions. A recent systematic review by Evrard et al. provided useful information on barriers and enablers for BZRA deprescribing in older adults in various settings of care [13]. Many barriers remain insufficiently addressed, such as low perceived self-efficacy or older adults’ and/or caregivers’ reluctance for cessation [14].

Hospitalization in older adults may present an important opportunity for optimizing medication use during admission and after discharge, yet there are limited data on the evolution of BZRA use during and after hospital admission. A few small studies reported encouraging data showing that engaging with patients during their hospital stay may contribute to increased rates of BZRA deprescribing, but post-discharge data were reported scarce [15,16,17]. A scoping review from Neville et al. recently identified interventions implemented in the acute care setting, but not specifically for older adults [18]. While the effects of pharmacotherapy optimization approaches on overall rates of potentially inappropriate prescribing have been measured [19], little is known about how much these global approaches can modify BZRA use during hospitalization and after discharge. The recent MedSafer study, which evaluated the effect of electronic decision support for deprescribing in older hospitalized patients, reported significantly higher discontinuation rates at discharge in the intervention group [20]. Less evidence is available regarding the impact of a global medication review conducted during hospital stay on BZRA deprescribing at discharge and after discharge.

We aimed to perform a secondary analysis of the OPERAM (OPtimising thERapy to prevent Avoidable hospital admissions in the Multimorbid elderly) trial, a cluster randomized controlled trial conducted in four European countries, which evaluated the effect of a structured medication review process on drug-related admissions. Our first objective was to describe BZRA prevalence before admission and identify associated factors. Our second objective was to measure the rate of BZRA cessation 6 months after hospital admission and to identify factors associated with cessation.

2 Methods

2.1 OPERAM Trial

The OPERAM trial (ClinicalTrials.gov identifier: NCT02986425) was a cluster randomized controlled trial conducted in four European countries (Belgium, Ireland, the Netherlands, and Switzerland). Participants were aged 70 years or older and had multimorbidity (≥ 3 chronic medical conditions) and polypharmacy (≥ 5 chronic medications [21]). The intervention involved a structured history of medication, a medication review supported by an explicit criteria tool, followed by shared decision-making with both participant and attending physician, and a report to the primary care physician [22]. BZRA use for a period of > 4 weeks was one of the triggers of the clinical decision support system (CDSS) based on the STOPP criteria [2], but the attending physician was free to accept or reject the recommendation to discontinue the BZRA. No BZRA-specific cessation program (e.g., tapering schedule) was proposed. The control group received usual pharmaceutical care. Follow-up and outcome data were collected by blinded trained researchers through telephone interviews at 2, 6, and 12 months after randomization. Details of the protocol and intervention have been published previously [22, 23]. The intervention did not significantly affect drug-related admissions (the primary endpoint), but it caused no detriment to participants [24]. In the intervention arm, 86.1% of participants (n = 789) had one or more inappropriate prescription at baseline according to the STOPP criteria version 2 [2] and 62.2% (n = 491) had one or more recommendation successfully implemented at 2 months, predominantly discontinuation of potentially inappropriate drugs. The OPERAM trial received approval from ethics committees at each site. All methods were performed in accordance with the relevant guidelines and regulations.

2.2 Eligibility Criteria

We included all the participants from the OPERAM trial who were alive and had follow-up data available at 6 months (Fig. 1).

Fig. 1
figure 1

Study participant inclusion flowchart

Full size image

2.3 Outcomes

The first outcome of interest was BZRA prevalence at baseline, defined as the intake of at least one BZRA—taken daily and/or as needed—before admission. For descriptive purposes, we also recorded the prevalence of BZRA use at discharge. The second outcome of interest was BZRA cessation, defined as no intake of any BZRA at the 6-month follow-up in a patient taking at least one BZRA at baseline. Tapering and discontinuation could have happened during the hospital stay and/or after discharge. This time point was chosen because of the gradual dose tapering recommended, leading to withdrawal schedules lasting for several weeks [25]. The 2-month time point did not allow long tapering schedules to come to an end, and the 12-month time point was influenced more by intercurrent health condition changes. The following Anatomical Therapeutic Chemical (ATC) classes were considered as BZRAs: N03AE01, N05BA, N05CD, and N05CF. We also identified factors associated with BZRA prevalence at baseline and with BZRA cessation. Additional details are provided in the analysis section.

2.4 Statistical Analysis

Data are described as median (first and third quartiles [Q1–Q3]) for numerical variables and number (percentage) for categorical variables. Student’s t test was used for continuous variables, and Pearson’s Chi-squared test or Fisher’s exact test for categorical variables.

Factors potentially associated with BZRA prevalence at baseline and with BZRA cessation were selected through literature review [26,27,28,29,30] and discussion among research team members. They included demographic and administrative data, comorbidities (including mental disorders or high-risk situations such as previous falls), and medication lists—with a focus on CNS-acting medications (Online Resource Supplementary Material 1 in the electronic supplementary material). Except for falls in the previous 12 months (directly available), comorbidities were identified for each patient by the International Classification of Diseases (ICD-10 [31]) codes from discharge letters (Supplementary Material 1). The Charlson Comorbidity Index (CCI) was calculated using the ICD-10 codes [32]. Alcohol abuse was considered when consumption exceeded 7 units per week, as self-reported [33]. Functional status was assessed using the Katz index, with a lower score indicating a higher level of dependence [34]. Not living independently was defined as living in a nursing home (at least 3 months in the 6 months before the index admission) or being housebound [24]. For a quality-of-life measure, we used the item “Depression/anxiety levels” that was part of the EQ-5D Health Questionnaire [35]. We merged levels 3 (severely anxious or depressed) and 4 (extremely anxious or depressed) to avoid excessively small samples. The trial arm (intervention vs control) was also considered as a factor potentially associated with BZRA cessation because BZRA use for a period of > 4 weeks was one of the triggers of the CDSS based on the STOPP criteria. Interactions between the trial arm and the other factors were tested and found to be statistically non-significant.

Binary logistic regression models were performed to assess independent variables associated with BZRA prevalence at baseline and BZRA cessation, and adjusted odds ratios (ORs) with their 95% confidence intervals (CIs) were calculated. First, a univariate analysis was performed with all variables for each outcome. Variables with a p value < 0.20 in the univariate analysis were eligible for the multivariable analysis. Collinearity was assessed between quantitative variables (Pearson coefficient) and qualitative variables (Phi coefficient). The choice between two correlated variables of the one to be entered in the multivariable model was based on their respective clinical relevance. Multicollinearity was assessed using the variance inflation factor. Missing data were not imputed. All test results were two-sided, and a p value < 0.05 was considered statistically significant.

All analyses were performed using R software v.4.1.1 (R Foundation for Statistical Computing, Vienna, Austria).

3 Results

A total of 2008 participants were enrolled in the OPERAM trial, of which 1601 completed the 6-month follow-up [24] and were included in the present secondary analysis (Fig. 1). Baseline characteristics are provided in Table 1. The median age [Q1–Q3] was 78 [74–83] years, 714 participants (44.6%) were female, and 777 (48.5%) were randomized into the intervention arm. The median length of stay was 6 [3–10] days. The number of participants not living independently was 276 (17.3%). A total of 999 participants (62.9%) reported not feeling anxious or depressed. At baseline, 145 participants (9.1%) had a diagnosis of depression and 101 (6.3%) had a major cognitive disorder. Six hundred and nineteen participants (38.8%) had experienced at least one fall in the previous 12 months. Severe polypharmacy (10 chronic medications or more) was present in 803 participants (50.2%), and 810 (50.6%) were taking at least one CNS-acting medication at baseline.

Table 1 Substudy population characteristics
Full size table

3.1 BZRA Prevalence at Baseline and Associated Factors

At baseline, BZRA prevalence was 23.6% (378/1601), with 70.3% of cases reporting daily use (Supplementary Material 2, see the electronic supplementary material). Most frequent BZRA molecules were zolpidem and lorazepam (Supplementary Material 3). BZRA prevalence was 22.1% in the intervention arm (172/777) and 25.0% in the control arm (206/824). In Switzerland, 121 participants (17.4%) were BZRA users at baseline, a prevalence lower than in Belgium, the Netherlands, and Ireland [respectively, 32.5%, 24.8%, and 28.0%]. Among BZRA users, 35.7% were taking three or more CNS-acting medications.

In the multivariable analysis (Table 2), the following variables were significantly associated with higher BZRA prevalence at baseline: female sex (OR [95% CI] 1.52 [1.18–1.96]), trial site (Belgium OR 2.64 [1.84–3.78], being resident in the Netherlands OR 1.90 [1.32–2.75] and Ireland OR 1.88 [1.31–2.70] as compared to Switzerland), a higher depression/anxiety level (OR up to 2.45 [1.54–3.89]), a higher number of daily medications other than BZRA (for each additional medication OR 1.08 [1.05–1.12]), the use of an antidepressant (OR 1.74 [1.31–2.31]), and the use of an antiepileptic (OR 1.46 [1.02–2.07]). Diabetes (OR 0.60 [0.44–0.80]) was significantly associated with lower BZRA prevalence at baseline

Table 2 Factors associated with BZRA use at baseline
Full size table

3.2 BZRA Cessation and Associated Factors

At the 6-month follow-up, BZRA cessation had occurred in 86 participants (22.8% of BZRA users at baseline). Among these 86 participants, 42 (48.8%) had their BZRA already discontinued at discharge. Among the 67 older adults who had their BZRA discontinued at discharge, 42 (62.7%) were still BZRA-free at 6 months (Fig. 2). Regarding the trial arm, 37 participants (18.0%) from the control arm and 49 (28.5%) from the intervention arm had ceased their BZRA at the 6-month follow-up.

Figure 2
figure 2

Benzodiazepine receptor agonist (BZRA) use at discharge and 6-month follow-up

Full size image

In the multivariable analysis (Table 3), any antidepressant use at baseline (OR [95% CI] 1.74 [1.06–2.86]) and a history of falling in the previous 12 months (OR [95% CI] 1.75 [1.10–2.78]) were associated with higher BZRA cessation rates. Also, being in the intervention arm showed a similar but not statistically significant association with BZRA cessation (OR [95% CI] 1.47 [0.95–2.30]). Chronic obstructive pulmonary disease (COPD) was associated with a lower BZRA cessation rate (OR [95% CI] 0.45 [0.20–0.91]).

Table 3 Factors associated with BZRA cessation at 6-month follow-up
Full size table

4 Discussion

In this substudy including older adults with multimorbidity and polypharmacy, the prevalence of BZRA use at baseline was 23.6%. Approximately one quarter of BZRA users at baseline had discontinued their BZRA 6 months after their hospitalization, and discontinuation occurred more frequently in patients with concurrent antidepressant medication.

The observation that approximately a quarter of older adults were taking a BZRA at baseline aligns with the prevalence reported in the recent MedSafer trial involving older adults with polypharmacy hospitalized in Canada [36]. However, the prevalence of BZRA use in OPERAM significantly differed across sites. Despite the limited generalizability because data were collected from only one trial site per country, variations across countries have been previously reported [6]. Although there is no clear basis for this variation, the explanation is probably multifactorial and may include factors at the level of the healthcare system. For instance, Belgium, the country with the highest prevalence of BZRA use in OPERAM, has no specific regulation for BZRA prescription such as there is in France, where zolpidem is treated as a restricted drug [37]. More research is needed to understand the influence of patients, healthcare professionals, and system-level factors on BZRA use.

Two factors associated with BZRA use at baseline deserve specific attention. First, one third of BZRA users at baseline were using three or more CNS-acting medications and participants with a higher number of other medications were more likely to use a BZRA at baseline. More precisely, for each additional drug, there was an 8% greater risk of BZRA use. This is particularly concerning considering that both CNS-acting co-medication in particular and polypharmacy in general are associated with an increased risk of falls and fractures [38]. As falls are associated with high morbidity and mortality [39], clinicians should consider these situations as high risk and prioritize deprescribing conversations with patients. Second, consistent with published data, women were more likely to be BZRA users than men [5]. Previous evidence highlighted higher rates of depression (symptoms) and anxiety in (older) women [40], and differences in sleep behavior and sleep disorders [41]. These differences may not be driven only by biological factors but by gender differences in the way women and men report symptoms, in socioeconomic status, or in history of (gender-based) violence [42]. Further research on sex and gender differences is needed and could contribute to better address discontinuation from a gender perspective.

Discontinuation at 6 months occurred in almost one in four BZRA users. First, antidepressant co-medication was significantly associated with higher discontinuation rates. This is consistent with a recent study investigating BZRA deprescribing at discharge from Belgian specialist geriatric wards [9] and could be explained by potentially positive effects of some antidepressants on benzodiazepine withdrawal symptoms and on symptoms of anxiety after BZRA deprescribing [43]. Second, a history of falling in the previous 12 months was also associated with BZRA cessation, which could be the result of deprescribing in response to an adverse event of BZRAs, since they may cause ataxia and impaired sensorium predisposing to additional falls [3]. This so-called reactive deprescribing is indeed the most common deprescribing activity in the acute context [44]. Third, in our study, COPD was associated with a lower BZRA cessation rate, but not with a different BZRA prevalence at baseline. Although data suggest that patients with COPD are more likely to experience a higher level of anxiety or depression [45], this association is difficult to explain and should be considered with caution given the large CI.

Interestingly, we did not find differences between trial sites regarding BZRA cessation. In other words, and with the limitation of small subpopulations, BZRA cessation was similar and limited across countries, whatever the baseline prevalence was. This is an interesting finding highlighting the need for a comprehensive, multifaceted, patient-centered strategy to facilitate BZRA deprescribing.

The pharmacotherapy optimization intervention of OPERAM was not aimed specifically at reducing BZRA use, but BZRA use for a period of > 4 weeks was one of the triggers of the CDSS, generating a signal recommending deprescribing BZRA that the attending physician was free to accept or not. Data from two OPERAM secondary analyses show that this STOPP criterion was triggered in 181 of 826 participants (21.9%) in the intervention arm, and that the pharmacotherapy team accepted 64.1% of these STOPP signals, resulting in a recommendation to the attending hospital physician [46]. In addition, 55.1% of BZRA cessation recommendations were agreed with by both patient and attending physician. When disagreement occurred, the main reason was patients’ reluctance to discontinue the medication [47]. Previous publications have shown efficacy of in-hospital interventions in decreasing potentially inappropriate medication exposure [24, 48]. We could not show that the OPERAM intervention was associated with significantly higher BZRA cessation in comparison with usual care, probably in part because our sample was insufficiently powered for this purpose. In MedSafer, deprescribing (defined as discontinuation or dose decrease 30 days after discharge) occurred significantly more often in the intervention group compared to the control group (adjusted difference 22.7% [95% CI 12.0–33.5]). The MedSafer intervention included one additional component as compared to OPERAM: participants or their caregivers received an educational pamphlet on deprescribing, as well as a patient-oriented sedative-hypnotic deprescribing pamphlet from the Canadian Deprescribing Network [36]. Wilson et al. used the same sedative-hypnotic deprescribing pamphlet, with a substantial increase of BZRA deprescribing [17]. Another way of supporting deprescribing in the hospital setting would be to target practitioners’ behavior by measuring, reporting, and sharing deprescribing activity [49]. A recent scoping review on interventions implemented in the acute care setting, but not specific to older adults, reported that multifaceted interventions aimed at patients and healthcare providers that include a combination of education, sleep protocols, and deprescribing may support reductions in benzodiazepine or other sedative/hypnotic drug use [18].

Almost two thirds of older adults who had their BZRA discontinued at discharge were still BZRA-free at 6 months, which is encouraging and could have been supported by the discharge report sent to the general practitioner—a key component for sustained deprescribing [50]. Conversely, less than half of older adults with BZRA cessation at the 6-month follow-up had already ceased at discharge. This could be explained by the required slow dose tapering exceeding the length of hospital stay or the preference of the participants to either discuss discontinuation with their general practitioners and/or to initiate the process outside the acute setting.

This study has several strengths. First, it examines data from a large sample of older adults with multimorbidity and polypharmacy recruited prospectively in several countries. Second, the 6-month follow-up is a period long enough to allow for a full deprescribing scheme based on participants’ needs and preferences. This study has some limitations also. First, we did not consider specific details of dose reduction of BZRAs. This would have provided a clearer picture of BZRA deprescribing and enabled more direct comparisons with other studies. Second, we have no detailed data about possible failed attempts at BZRA cessation, although two published secondary analyses provide some highlights on the acceptance of and reasons for rejecting recommendations to discontinue medications in the OPERAM trial [46, 47]. Third, due to the broader scope of the OPERAM trial, no specific data were collected regarding the indication for BZRA use (anxiety and/or sleep disorder, other). Finally, we did not collect specific patient-reported outcomes, such as withdrawal symptoms and sleep patterns, which would have been interesting to better capture the participant experience. This information was not available since the OPERAM intervention did not focus on BZRA cessation.

5 Conclusion

BZRA use was observed in almost one quarter of hospitalized older adults with multimorbidity and polypharmacy, and BZRA cessation occurred in almost one quarter of BZRA users. To increase BZRA cessation or dose reduction, a more targeted deprescribing program is most probably needed, with components addressing barriers at patient, healthcare professional, and organization levels. Future studies also need to pay specific attention to older adults taking CNS-acting medications, females, and COPD co-morbidity.