eAppendix. Search Strategy for Interventions to Reduce Inpatient Sedative Use
eTable 1. Medline Search String
eTable 2. Embase Search String
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Soong C, Burry L, Cho HJ, et al. An Implementation Guide to Promote Sleep and Reduce Sedative-Hypnotic Initiation for Noncritically Ill Inpatients. JAMA Intern Med. Published online June 03, 2019179(7):965–972. doi:10.1001/jamainternmed.2019.1196
Sedative-hypnotic medications are frequently prescribed for hospitalized patients with insomnia, but they can result in preventable harm such as delirium, falls, hip fractures, and increased morbidity. Furthermore, sedative-hypnotic initiation while in the hospital carries a risk of chronic use after discharge. Disrupted sleep is a major contributor to sedative-hypnotic use among patients in the hospital and other institutional settings. Numerous multicomponent studies on improving sleep quality in these settings have been described, some demonstrating an associated reduction of sedative-hypnotic prescriptions. This selected review summarizes effective interventions aimed at promoting sleep and reducing inappropriate sedative-hypnotic initiation and proposes an implementation strategy to guide quality improvement teams.
Use of sedative-hypnotic drugs for insomnia—such as benzodiazepines, nonbenzodiazepine γ-aminobutyric acid receptor agonist agents (so-called Z-drugs, including eszopiclone, zopiclone, zolpidem tartrate, and zaleplon), and some antipsychotics, antidepressants, and antihistamines—is increasing despite a clear association of these medications with adverse events.1-6Quiz Ref ID Sedative-hypnotics are categorized as high-risk medications, particularly among older adults, and are associated with preventable harm including falls, hip fractures, delirium, and death.3,7-9 Observational studies suggest that most indications for sedative-hypnotic initiation in hospitals are potentially inappropriate.10,11 As a result, multiple professional societies list sedative-hypnotics on their Choosing Wisely lists of unnecessary treatments to avoid.12-14 Reducing unnecessary sedative-hypnotic use may reduce adverse events and costs associated with medications and treatment-related harms.
Several factors contribute to incident sedative-hypnotic use among inpatients, including disruptive sleep environments, acute illness, recent surgical procedures, and inadequate awareness of the harms associated with episodic sedative-hypnotic use.10,15 Eliminating the use of unnecessary sedative-hypnotics can prove challenging as it requires multiple types of interventions. Broadly, interventions studied fall into the following categories: nonpharmacologic sleep hygiene implementation, audit and feedback, pharmacist-enabled medication reviews, computer-based interventions, and education. We provide a focused review of these intervention types and a framework for clinical teams aiming to reduce sedative-hypnotic initiation for disruptive sleep in inpatient settings for noncritically ill patients.
Sedative-hypnotic use is associated with preventable harm. Among community-dwelling older adults, an increase in the relative risk of nonvertebral fractures and dislocation was observed with an initial prescription of either zolpidem (2.55; 95% CI, 1.78-3.65; P < .001), lorazepam (1.53; 95% CI, 1.23-1.91; P < .001), or diazepam (1.97; 95% CI, 1.22-3.18; P < .001).7 In hospitalized patients, benzodiazepine administration was associated with an odds ratio of 2.3 for falls (95% CI, 1.4-3.7), with a higher proportion of those who fall receiving a benzodiazepine for the first time (20% vs 7%; P < .001). Most falls (95%) were associated with temazepam use.16 In another observational inpatient study, zolpidem use was associated with an adjusted odds ratio of 4.37 for falls (95% CI, 3.34-5.76; P < .001) even after accounting for age, sex, comorbidities, and confounders.17 Furthermore, initiation of sedative-hypnotics is associated with increased length of stay. In a retrospective review of sedative-hypnotic use in hospitalized patients, those who received sedative-hypnotics experienced a nearly 75% longer length of stay than those who did not (P < .05).18 A Japanese study of hospitalized patients found similar results.19
Another risk of sedative-hypnotic exposure in hospitals is continued use of sedative-hypnotics after discharge. Among a large cohort of sedative-hypnotic–naïve hospitalized older adults, 3.1% filled a prescription for a sedative-hypnotic within 7 days of being discharged from the hospital. Approximately half these patients became new long-term users of sedative-hypnotics.20 Another study found that initiation of sedative-hypnotic medications among older inpatients was associated with an adjusted odds ratio of 4.65 (95% CI, 1.95-11.09) for postdischarge use and at the 3-month follow-up visits.21
The benefit to risk ratio of using sedative-hypnotic medications for insomnia remains unclear.22-24Quiz Ref ID A large meta-analysis of 24 studies examining sedative-hypnotic use for chronic insomnia in 2417 hospitalized and community-dwelling participants found that total sleep time improved by a mean of 25.2 minutes (P < .001) and the number of nighttime awakenings decreased by 0.63 (P < .001) compared with placebo.25 However, adverse events such as cognitive disturbances (eg, confusion, memory loss, and disorientation) were 4.78 times more common with sedative-hypnotics than with placebo (95% CI, 1.47-15.47; P < .001). The authors of that study reported a number needed to treat of 13 for improved sleep quality compared with the number needed to harm of 6 for any adverse event. Similar results were found in a recent review of pharmacologic treatment of insomnia disorders as defined by the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition.26 Such modest gains in total sleep duration may not outweigh the well-documented adverse consequences of sedative-hypnotic use in the hospital.
Sleep difficulty in the hospital setting is a consequence of acute medical illness and an unfamiliar and disruptive environment.19,20 Addressing the underlying causes of poor sleep is a safer and possibly more effective alternative to sedative-hypnotic use. Causes of disrupted sleep may be categorized as either modifiable causes such as disruptive sleep environments, patient care activities, and administration of medications known to disrupt sleep (eg, corticosteroids, bronchodilators, and levodopa); or nonmodifiable causes such as unfamiliar settings causing disorientation or underlying preexisting sleep disorders.24 Targeting these root causes of poor sleep may be a safer alternative to prescribing sedative-hypnotics.
We performed a focused search of MEDLINE and Embase to identify English-language studies published in the last 15 years aimed at promoting sleep while reducing incident inpatient sedative-hypnotic use (excluding outpatient, intensive care, or psychiatric settings and populations). The search strategy is described in the eAppendix and eTables 1 and 2 in the Supplement. We identified several studies among sedative-hypnotic–naïve patients in a hospital or long-term care setting. Among studies that measured incident sedative-hypnotic prescribing, the most commonly targeted sedative-hypnotics were benzodiazepines and nonbenzodiazepine γ-aminobutyric acid receptor agonists. Other sedating drug classes, such as antihistamines, sedating antipsychotics, and antidepressants, comprised a minority of drugs studied. Considerable study heterogeneity existed in setting, methods, and outcome. However, each study described at least 1 measure of sedative-hypnotic use. Broadly, the studies can be categorized into distinct intervention types: nonpharmacologic sleep hygiene implementation, audit and feedback, pharmacist-enabled medication review, computer-based intervention, and clinician education. In this section, we describe examples in each of the categories.
Four studies described implementation of sleep hygiene protocols to reduce sedative-hypnotic use (Table).27-30 Commonly, this approach involved adopting a behavioral bundle addressing modifiable factors associated with sleep disruption, reinforcing premorbid healthy sleep habits and circadian rhythms, and promoting relaxation without pharmacologic means. Quiz Ref IDSpecific strategies include creating a sleep-friendly environment in the hospital through reduction of noise and light, modifying clinical monitoring and medication administration schedules where appropriate to maximize uninterrupted sleep time, and assisting patients with bedtime routines and relaxation techniques. In a single-center pre-post study, investigators implemented a sleep protocol that included 8 hours of quiet time with sound meters to alert staff of increasing noise levels and reduced routine medication administration, noise, lights, and vital sign monitoring.27 They observed a reduction in as-needed sedative use from 37% to 16% (P < .05). Validated sleep scores remained unchanged. In a large 2-year pre-post study based in a Korean hospital, researchers implemented an “i-Sleep” program focused on clinician education in addition to promoting nonpharmacologic sleep hygiene.28 The study observed an 18% reduction in sedative prescriptions among inpatients taking sedative-hypnotics prior to admission (relative risk, 0.82; 95% CI, 0.79-0.86).
The sleep protocols described may be more effective in reducing sedative-hypnotic initiation than reducing premorbid long-term sedative-hypnotic use. A prospective pre-post study among general medical inpatients in an academic teaching hospital found that long-term sedative-hypnotic users were 1.6 times more likely to refuse sleep protocols than were nonusers (64% vs 41%; P < .05) and received sedative-hypnotics 4.5 times more often than nonusers (67% vs 15%; P = .001).29 An unblinded randomized clinical study in an academic medical center implemented nurse-driven sleep hygiene and observed a reduction in the mean number of sedative-hypnotics administered for insomnia, with 1.55 medications administered in the intervention group compared with 2.20 in the control group (P < .05).30
Another common intervention involved structured medication reviews conducted by pharmacists (Table).31,32,34,40 This approach often targets overall prescribing appropriateness, including sedative-hypnotic drug classes. A single-center, pre-post study among inpatients at risk for falls in Ireland implemented pharmacist review of medications to reduce falls.31 Treating physicians had the discretion to implement pharmacist recommendations. If no change was initiated after 1 week, study investigators sent a reminder to the physician. Of the 239 fall-risk medications that received at least 1 medication change recommendation, sedative-hypnotics comprised 13%. Physicians implemented 70% of the appropriate sedative-hypnotic medication change recommendations.
Three studies implemented various forms of pharmacist-enabled medication reviews in long-term care settings (Table).32,34,40 A multicenter retrospective cohort study conducted across 62 aged-care homes in Australia implemented a pharmacist medication review that identified sedative or anticholinergic medications and made recommendations to physicians to implement medication changes.32 After the intervention, the median drug burden index (a measure of a person’s total exposure to medications with anticholinergic and sedative properties, using the principles of dose response and maximal effect) decreased from 0.5 to 0.33 (95% CI, 0-0.67; P < .001), with a range of 0 to 3.29. A cluster-randomized clinical trial implemented interprofessional medication reviews as part of structured case conferences among long-term care residents with challenging behaviors.33 Pharmacists conducted medical record reviews and determined the Medication Appropriate Index, a validated instrument measuring medication appropriateness based on 10 criteria and associated with quality of life and medication-associated hospital admissions.40 There was an improvement in the Medication Appropriate Index for benzodiazepines among participants in the intervention group compared with controls (intervention: mean change, −0.38; 95% CI, −1.02 to 0.27; and control: mean change, 0.73; 95% CI, 0.16 to 1.30; P = .02).33 Similarly, a cluster-randomized multicenter study randomly assigned a multipronged intervention vs usual care to 12 long-term care facilities in Massachusetts.34 The intervention included academic detailing on the appropriate use of psychoactive drugs (including sedative-hypnotics) and pharmacist medication reviews targeting antipsychotics, sedative-hypnotics, and sedating antidepressants. Investigators calculated that the mean psychoactive drug use score (points assigned to type of drug used and dose, for a maximum score of 7) was reduced by 27% from 1.87 to 1.36 in experimental groups vs 1.74 to 1.60 in control groups (mean difference in risk reduction, 0.37; 95% CI, −0.08 to −0.67; P < .05).
A prospective pre-post multicenter study conducted across general-medicine and geriatric wards in Australia examined the association of providing unit-specific benzodiazepine prescribing rates to health care professionals with prescription patterns (Table).35 After a baseline audit, unit teams received their data with discussion facilitated by a geriatrician and pharmacist. Presenters also provided a brief review of benzodiazepine use among older adults. Two additional audits were conducted at 1 to 2 months and 6 months after the intervention. The primary outcome was the proportion of patients prescribed a benzodiazepine. Secondary outcomes included appropriateness of prescription and whether there was an attempt to reduce the dose. Appropriate indications included seizure, alcohol withdrawal, muscle spasm, procedural premedication, and preexisting use prior to hospitalization. Two pharmacists adjudicated the appropriateness of clinical indications, with a Cohen κ of 0.75. The proportion of patients prescribed a benzodiazepine remained unchanged (36% to 31%; P > .05); however, overall prescription appropriateness improved from 20% to 44% (P < .001).
A single-center study conducted on 4 medical inpatient wards in an Indian hospital followed the pattern of drug prescribing using Modified Beers Criteria and included benzodiazepines (Table).36 Investigators conducted random audits and pharmacists provided the information back to the prescribing clinician. Overall, the study found a change in therapy in 30.7% of prescriptions.
A single-center prospective pre-post study among older adults implemented a computer-based reminder in addition to education to reduce prescriptions of 4 sedative-hypnotic agents: diphenhydramine hydrochloride, diazepam, lorazepam, and trazodone hydrochloride in older hospitalized patients (Table).37 The hospital formulary excluded other common sedative-hypnotics (zolpidem, triazolam, and zaleplon). The intervention included 3 computer-based real-time alerts: the first alert was triggered after an order for diphenhydramine or diazepam to ascertain the indication for the order. If the clinician selected sleep as an indication, a second educational reminder appeared, providing information on adverse effects associated with the drug and alternative options that included nonpharmacologic interventions or lorazepam or trazodone (medications that were thought to be less harmful than study drugs at the time of the study). Finally, if clinicians selected pharmacologic interventions, a third screen defaulted to the lowest possible dose for each drug. The authors observed an 18% relative risk reduction in prescriptions for all sedative-hypnotics after implementation of the intervention (odds ratio, 0.82; 95% CI, 0.76-0.87; P < .001), with a significant decline in lorazepam prescribing comprising most of the reduction observed.
A before-and-after cross-sectional survey study conducted in a tertiary Australian hospital examined the association of an educational campaign on sedative-hypnotic use with knowledge and prescription rates (Table).38 During a 12-week period, study investigators, pharmacists, and drug and alcohol specialists conducted educational sessions in the form of medical grand rounds, seminars, workshops, and ward meetings with distributed booklet and videos. Education targeted health care professionals as well as patients. Content focused on sleep physiology, the effect of the sedative-hypnotic on sleep patterns, and associated adverse effects. Data collection occurred through patient interviews and mailed surveys. After the intervention, there was an increase in knowledge of associated harm of sedative-hypnotics (58.4% vs 42.1% before the intervention reported knowing that sedative-hypnotics were associated with harmful adverse effects; P < .005) and a decrease in patient-reported sedative-hypnotic use among patients 3 months after discharge from 39.4% to 27.3% (P < .005).
A similar educational intervention was implemented in a multicenter controlled cohort study in Alberta, Canada.41 Investigators provided education (30- to 45-minute sessions) to physicians, nurses, pharmacists, and families on the indications and harms of neuroleptic drugs (including benzodiazepines) that stressed a nonpharmacologic approach to agitation and insomnia. The main outcome measure was prescription rates of antipsychotics. However, the study found no difference between the intervention and control groups in benzodiazepine prescriptions.
The RedUSe (Reduce Use of Sedatives) study implemented a multifaceted intervention to reduce antipsychotic and benzodiazepine use in long-term care facilities in Tasmania (Table).39 This controlled pre-post study examined the association of a combination of education, training, nonpharmacologic approaches to behavioral and psychological symptoms of dementia and insomnia, academic detailing, and pharmacist-enabled sedatives review with sedative prescriptions in 13 intervention centers. The study found that the benzodiazepine prescription prevalence decreased from 31.8% (baseline) to 26.9% (postintervention) (P < .005) without a change in controls. A follow-up study found sustained reduction in benzodiazepine prescriptions 12 months after implementation of the intervention, from 31.8% to 28.5% (P < .05).42
A recent study took a policy approach to appropriate medication prescribing. This 6.5-year observational study examined the association of restricted pharmaceutical detailing (ie, restricting visits from pharmaceutical sales representatives to physicians) with prescription patterns across 8 drug classes among 19 academic medical centers and private clinics in 5 US states.43 The study found a mean decrease in market share of all detailed drugs of 1.67 percentage points (95% CI, −2.18 to −1.18 percentage points; P < .001). Among sleep aids, restrictive pharmaceutical detailing policies were associated with significant decreases in market share for these drugs. Although this study did not implement interventions at a local level, the findings support the restriction of targeted pharmaceutical detailing. This finding may be of interest and value to hospital administrators and policymakers.
Quiz Ref IDDespite an inclusive review, there are several important limitations to consider. First, many of the studies reviewed are of low methodological rigor, lack a randomized design or control group, and may be subject to the Hawthorne effect. Second, while the interventions described in this guide have shown efficacy in certain settings, we acknowledge the association of context with observed outcomes, which may limit reproducibility as no intervention has been tested across broad settings and patient populations. Third, the inpatient settings differ in many important ways. It is possible that an intervention studied in the long-term care setting may lack efficacy in an acute-care hospital because of contextual factors. However, learnings from interventions may be applicable to both hospital and long-term care settings where drivers of inappropriate sedative-hypnotic use are similar (eg, disruptive sleep environment). In addition, team composition and models of care are comparable, which would enable a team-based approach in a supervised setting. Fourth, studies included a broad range of at-risk medications, which may affect generalizability across drug classes. However, focusing on the indication of insomnia helps to narrow the scope. Fifth, studies did not consistently describe incident prescriptions only and may have included patients with long-term sedative-hypnotic prescriptions. Finally, there were no adverse effects reported with reducing sedative-hypnotics, but it is possible that adverse effects were not systematically measured.
Despite the limitations of the available evidence, we found some evidence that promoting sleep while restricting sedative-hypnotics for insomnia is feasible. In addition, it is also important to note the widespread adoption of sedative-hypnotics for insomnia despite the low level of quality evidence supporting their use as effective safe sleep aids. In this context, we frame this implementation strategy as an approach to de-adopt a harmful non–evidence-based practice. Teams should adapt and modify interventions to best suit the needs and resources specific to the institution while carefully considering the effect of context. It is likely that a combination of interventions may better address different contributors to the problem specific to the local setting, particularly given the rather modest improvements observed across individual interventions. The following is a list of intervention options to guide quality improvement initiatives to actively de-adopt sedative-hypnotic use for sleep use among inpatients.
A major extrinsic contributor to insomnia and resulting sedative-hypnotic prescriptions among inpatients is the disruptive sleep environment.10 Creating a sleep-promoting hospital environment can mitigate the traumatizing effects of hospitalization and should be a necessary component of any effort to reduce sedative-hypnotic use for hospital-acquired insomnia.34,40 Teams, stakeholders, and hospital administrators can implement a nonpharmacologic sleep enhancement bundle. This plan may include policies and procedures to achieve a sleep-friendly hospital setting such as designated overnight periods of reduced noise, lighting, and interruptions for unnecessary clinical monitoring or nonessential medication administration. Noise, disruptions, and requests for sleep aids can be monitored as process measures. Physicians can enhance sleep by optimizing pain and symptom management and avoiding requesting nonurgent investigations that may disrupt sleep (such as routine laboratory testing, for which blood is frequently drawn during early morning hours). Clinical teams can work together at the unit level to create a culture of sleep promotion and emphasize a nonpharmacologic approach as a first-line management of insomnia (ie, use of ear plugs, eye masks, warm noncaffeinated beverages, soothing music, and toileting).27,28,35,36
Quiz Ref IDDepending on local context and resources available, several clinician-focused interventions can further reduce sedative-hypnotic initiation for insomnia. These include pharmacist-enabled medication reviews to flag new prescriptions, computer-based interventions, and education. These strategies can be implemented as distinct interventions or form a sedative-hypnotic reduction bundle of interventions.
Pharmacists represent key stakeholders positioned to identify potentially inappropriate sedative-hypnotic initiation and/or maintenance. Structured medication reviews among hospitalized patients is already an established organizational practice in many settings (eg, hospitals, postacute care, and long-term care) and can be leveraged to enhance appropriate sedative-hypnotic prescribing. This review can be performed at multiple stages along an inpatient’s continuum of care.
Nurses and nurse practitioners are also ideally positioned to promote sleep and prevent sedative-hypnotic use at the point of care. When a patient reports difficulty sleeping, the nurse on duty can perform an environmental assessment and introduce appropriate sleep hygiene practices in lieu of dispensing a sedative-hypnotic. Nurses can ensure that sedative-hypnotic prescriptions are removed from standard postoperative sleep or transfer protocols when the patient arrives on the ward. When filling out a fall report, the nurse could indicate on the form whether a sedative-hypnotic had been taken by the patient during the night. Sedative-hypnotic–induced falls are an excellent prompt to cancel as-needed sedative-hypnotic prescriptions.
Performance data feedback to clinicians is an effective means of influencing behavior change.44 Regular data feedback can provide clinicians with a target of desired outcomes as well as serve as reminders of project goals. Outcome measures should be predefined and may include overall sedative-hypnotic prescribing rates, or more precisely, dose reduction and deprescribing (for patients already prescribed sedative-hypnotics); reduced rates of new sedative-hypnotic initiation after admission; and uptake of nonpharmacolgic interventions. There was no benefit of audit and feedback on benzodiazepine reduction in the study that was reviewed, so further research is needed before this strategy is widely adopted to reduce sedative-hypnotic prescriptions.
Computer-based interventions such as restrictive ordering or indication-based nudges are commonly described effective methods to reduce low-value care.44,45 They have been shown to be effective and may further enhance other strategies as part of a multifaceted bundle aimed at reducing sedative-hypnotic use.45
Educational strategies provide awareness of the project rationale and evidence of preventable harm. They can be delivered via a number of methods targeting both clinicians and patients and may be a part of a larger academic detailing strategy. Recurrent educational materials and sessions may also serve as a reminder of the initiative and aid in sustainability efforts. In the studies reviewed, the strength of the overall evidence for education was weak and this strategy alone is unlikely to produce behavior change. Instead, education may be better used as a component of a multifaceted intervention.45
Although we describe several studies reporting successful reductions in inappropriate sedative-hypnotic prescribing, rigorous evaluation of successful interventions to reduce prescribing will improve our understanding of effective strategies and their outcome. We propose iterative testing and refinement of the interventions described followed by an evaluation of clinical and patient-reported outcomes such as falls, delirium, and sleep quality in multicenter implementation studies. Based on the local context and contributors of inappropriate sedative-hypnotic prescribing, some or all of these targeted interventions can be implemented in addition to further study and evaluation.
Promoting sleep while reducing sedative-hypnotic initiation for insomnia among inpatient settings can benefit patients through avoidable harm reduction. Several strategies have demonstrated efficacy in reducing sedative-hypnotic use and intervention selection should be based on local drivers of inappropriate prescribing. Creating a sleep-friendly inpatient environment is an important enabler of sedative-hypnotic reduction efforts. This guide can aid quality improvement teams, administrators, and clinicians toward achieving a safer hospital environment for inpatients.
Accepted for Publication: March 14, 2019.
Corresponding Author: Christine Soong, MD, MSc, Division of General Internal Medicine, Mount Sinai Hospital, 433-600 University Ave, Toronto, Ontario M5G 1X5, Canada (email@example.com).
Published Online: June 3, 2019. doi:10.1001/jamainternmed.2019.1196
Author Contributions: Dr Soong had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Soong, Burry, Cho, Gathecha, Kisuule, Tannenbaum, Morgenthaler.
Acquisition, analysis, or interpretation of data: Soong, Burry, Cho, Gathecha, Tannenbaum, Vijenthira, Morgenthaler.
Drafting of the manuscript: Soong, Burry, Morgenthaler.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Soong.
Administrative, technical, or material support: Soong.
Supervision: Soong, Tannenbaum.
Conflict of Interest Disclosures: None reported.
Additional Contributions: Pamela Johnson, MD, Department of Radiology, Johns Hopkins University, and Justin Turner, PhD, Faculty of Pharmacy, Université de Montréal, provided review and helpful suggestions in the development of this guide. They were not compensated for their contributions.
Additional Information: Drs Soong, Burry, Cho, Gathecha, Kisuule, Vijenthira, and Morgenthaler are members of the High Value Practice Academic Alliance. The High Value Practice Academic Alliance is a consortium of academic medical centers in the United States and Canada working to advance high-value health care through collaborative quality improvement, research and education. For additional information, visit http://www.hvpaa.org.
Create a personal account or sign in to: