Abstract
Objectives:
To pilot a low-touch program for reducing benzodiazepine receptor agonist (BZRA; benzodiazepines, z-drugs) prescriptions among older veterans.
Methods:
Pilot randomized controlled trial of 2,009 veterans aged >= 65 years who received BZRA prescriptions from a Veterans Health Administration pharmacy (Colorado or Montana) during the prior 18 months. Active: Arm 1 was a mailed brochure about BZRA risks that also included information about a free, online cognitive behavioral therapy for insomnia (CBTI) program. Arm 2 was a mailed brochure (same as arm 1) and telephone reinforcement call. Control: Arm 3 was a mailed brochure without insomnia treatment information. Active BZRA prescriptions at follow-up (6, 12 months) were measured.
Results:
In logistic regression analyses, the odds of BZRA prescription at 6- and 12-month follow-up was not significantly different for arm 1 or 2 (active) versus arm 3 (control), including models adjusting for demographics and prescription characteristics (p-values >0.36).
Conclusions:
Although we observed no differences in active BZRA prescriptions, this pilot study provides guidance for conducting a future study, indicating a need for a more potent intervention. A full-scale trial testing an optimized program would provide conclusive results.
Clinical Implications:
Mailing information about BZRA risks and CBTI did not affect BZRA prescriptions.
Keywords: benzodiazepine, insomnia, aging, deprescribing, older adults, mortality
Introduction
Up to 10% of older adults in the United States (US) are prescribed a benzodiazepine or z-drug benzodiazepine receptor agonist (BZRA) (Maust et al., 2021). Nearly one-third of benzodiazepine prescriptions are for sleep problems, and the majority of z-drug prescriptions are for insomnia symptoms. Sleep problems and insomnia are reported by almost half of older adults and a majority of US veterans (Jenkins et al., 2015; Kaufmann et al., 2018; Patel et al., 2018). Clinical practice guidelines do not recommend using BZRAs in older adults due to the increased risk of adverse outcomes associated with these drugs such as falls and cognitive impairment (American Geriatrics Society Choosing Wisely Workgroup, 2013). Some studies suggest that BZRAs may even increase risk of mortality (Mathieu et al., 2021). Guidelines, instead, recommend cognitive behavioral therapy for insomnia (CBTI) as first-line treatment for chronic insomnia disorder (Qaseem et al., 2016).
A variety of programs have been tested to reduce BZRA use in older adults (Gould et al., 2014). Some interventions target prescribers (e.g., electronic health record alerts) while others target patients (e.g., direct-to-patient mailings) (Gould et al., 2014). Tannenbaum et al. demonstrated that a direct-to- patient mailed educational empowerment brochure about the risks of sleeping pills and benefits of decreased sleeping pill use resulted in fewer benzodiazepine prescription renewals compared to usual care (Tannenbaum et al., 2014). Yet BZRA use remains relatively high in older adults despite the availability of information about the potential adverse consequences of BZRA use in this patient population (Maust et al., 2021). One important consideration is that BZRAs are often prescribed for chronic insomnia disorder (Agarwal & Landon, 2019), and tapering efforts that fail to concurrently address an underlying insomnia condition may be less successful. Indeed, prior studies show that combining a supervised gradual taper of the BZRA with in-person CBTI results in fewer individuals using BZRAs at follow up (greater success in BZRA discontinuation) (Gould et al., 2014). CBTI delivered in-person by a therapist, however, is resource intensive and multiple barriers to accessing care exist,(Gould et al., 2014) limiting its application in large-scale BZRA deprescribing programs. Online and other digital self-directed CBTI programs have shown promise as an alternative scalable delivery method (Soh et al., 2020). A study conducted in Sweden found that online CBTI decreased patient-reported sleeping pill use from 50% pre-treatment to 29% at 3-years follow-up (Blom et al., 2016).
The recent launch of a free, anonymous, online, self-directed CBTI program by the Veterans Health Administration (VHA) provides a unique opportunity to study whether adding CBTI program information to a direct-to-patient mailed program is an efficacious “low-touch” and potentially scalable approach to reducing BZRA prescriptions and improving important clinical outcomes (e.g., mortality (Mathieu et al., 2021)). Furthermore, the VHA’s robust and comprehensive corporate data warehouse has pharmacy data and clinical outcomes data from patients nationwide and has been leveraged by VHA pharmacies for direct-to-patient mailings about medication risks (Mendes et al., 2018). With its data infrastructure and high prevalence of sleep disorders among the patients it serves (Folmer et al., 2020), the VHA is an ideal setting to study the impact of a low-touch program on BZRA prescriptions in older adults.
We developed and conducted a small-scale test of a program for older veterans that uses this low-touch approach. Our goal was to develop and pilot study procedures, including 1) refining data extraction methods and outcome measures from pharmacy and health data and 2) refining and deploying the intervention materials, to inform a future large-scale trial. We hypothesized that a mailed brochure about BZRA risks that also included information about first-line treatment, CBTI, and included the web address of the VHA’s free, anonymous, self-directed online CBTI program would result in fewer active BZRA prescriptions at 6 months (primary timepoint) and 12 months (secondary timepoint) follow-up compared to a similarly mailed brochure that provided general, non-directed information about sleep (control condition). We hypothesized that the addition of a brief telephone reinforcement call to the mailed program would focus participants’ attention on the mailed brochure (Humpel et al., 2004), which in turn would result in more engagement with the brochure and CBTI website and ultimately, fewer active BZRA prescriptions at follow-up. Planning ahead for a full-scale trial, we also hypothesized that the active programs would impact important health outcomes and thus would result in fewer deaths at follow up (a priori secondary outcome).
Method
Overview/Trial Design:
In this pilot study that used a randomized controlled trial design, we compared two active programs (mailed brochure only and mailed brochure + telephone reinforcement call) to a control program in a sample of older veterans who were identified through the VHA’s corporate data warehouse as having received a BZRA prescription within the prior 18 months. Testing mailing procedures from one central site that could be highly scaled in a future study, we prepared and mailed program materials to participants in Colorado and Montana. The Colorado site was selected for its more urban population, and the Montana site was selected for its higher rurality index. At 6- and 12-month follow-up, pharmacy and mortality outcomes were extracted from the data warehouse. The VA Greater Los Angeles Healthcare System Institutional Review Board approved the study and the trial was registered (ISRCTN11530968).
Inclusion criteria:
Participants were veterans aged ≥ 65 years who received at least one BZRA prescription within the past 18 months from one of two VHA healthcare systems (in Colorado or Montana) based on VHA pharmacy data extracted from the VHA corporate data warehouse.
Exclusion criteria:
To increase the likelihood that the BZRA prescription was for chronic insomnia disorder and would not have a contraindication to CBTI, we excluded patients with a diagnosis of seizure/epilepsy, REM sleep behavior disorder, or bipolar disorder in their problem list based on corporate data warehouse records, or who were deceased.
Recruitment:
Information letters about the study were mailed to patients who met the initial screening criteria. The letter included a return card with options to “opt-out” or “opt-in.” Patients who opted-out of receiving the brochure after receiving the information letter were excluded from randomization. Patients who returned the opt-in card were considered active opt-in participants; participants who did not return the card by the specified return date were considered passive opt-in participants and were included in randomization.
Interventions
Arm 1 (active brochure only): A previously tested direct-to-patient mailed brochure about sleeping pills (Mendes et al., 2018; Tannenbaum et al., 2014) was adapted with permission (Mendes et al., 2018; Tannenbaum et al., 2014) to include a new section about the VHA’s free, anonymous online CBTI website (https://www.veterantraining.va.gov/insomnia/). The active brochure used in the current study did not include a BZRA taper schedule, which is an approach that was used in a prior study conducted in the VHA—the removal of the taper schedule addressed potential concerns about patient safety (Mendes et al., 2018). We added a new section about the online CBTI that included a snapshot of the program webpage and the web address. Participants were also offered the option of receiving a paper CBTI workbook in lieu of completing the online CBTI course online (Ulmer et al., 2017). Requests for printed copies were filled by our research team (21 total requests).
Arm 2 (active brochure + telephone reinforcement call): The arm 1 program was combined with a telephone reinforcement call within 1–2 months of randomization. We included the telephone reinforcement arm to test whether additional procedures would be needed in a future trial to augment participants’ engagement with the mailed material. A member of the research team used a script and called the participant to inquire whether the participant received the brochure. If participants indicated they had not received the brochure, the research team member resent the brochure. Other topics covered during this 10-minute call included whether the participant was aware of the CBTI website described in the brochure. The research team member also encouraged the participant to visit the online CBTI program, by pointing out where to find the website information in the brochure as well as giving the web address to the participant over the phone.
Arm 3 (control brochure): A brochure and a website with similar appearance to the active brochure and CBTI website were developed for the study. They contained general information about the importance of sleep (non-directed information). The control brochure included information about the control website such as the web address. The brochure was mailed to participants in the control group. No telephone reinforcement call was made to participants allocated to the control group.
Randomization and Allocation Concealment
Stratified block randomization was used for group allocation (Stata uniform() function; seed 93201). We equally randomized participants to arm 1 (active brochure only) and 3 (control brochure), but not to arm 2 (active brochure + telephone reinforcement). The reason for randomizing fewer participants to arm 2 was that the purpose of arm 2 was primarily to pilot test the telephone script procedures and measure additional resources needed. After generating the group assignment for each eligible individual, the research team sent the mailings with content based on study group assignment. Stratification variables were 1) US Food and Drug Administration (FDA) BZRA approval status, 2) chronicity of prescription (≥ 90 days versus < 90 days prescription length), and 3) opt-in status (active or passive). Participants were not informed which mailing was the active versus control mailing. Randomization for the Colorado site occurred in February 2020, and due to the onset of the COVID-19 pandemic that interrupted research operations, randomization for the Montana site occurred in May 2020. The study team member who extracted the data set was blinded to group allocation.
Measures
Demographic data were extracted from the corporate data warehouse, including age, gender, race, ethnicity, and marital status. These measures were selected a priori, as these factors are associated with insomnia (e.g., race (Kaufmann et al., 2016)), medication prescribing (e.g., site, gender (Olfson et al., 2015)), use of internet (e.g., age, education (Kontos et al., 2014)), and sleep behaviors (e.g., marital status (Rogojanski et al., 2013)). We classified BZRAs as having an insomnia indication 1) if the medication was a z-drug or 2) if the benzodiazepine prescription instructions included “sleep” or “bed” or was prescribed to be taken at bedtime or nightly. We classified BZRA prescriptions as chronic if the BZRA “fill days” in the corporate data warehouse was 90 or more consecutive days.
The primary outcome measure was the presence of any active BZRA prescription in the corporate data warehouse at the 6-month time point (6 months after randomization). We defined this measure by assessing presence of any active BZRA prescription, defining “active prescription” as any prescription within 90 days of the 6-month follow-up timepoint. We selected 90 days, because VHA pharmacy does not dispense BZRAs in quantities that exceed a 90-day supply. The outcome data extracted from the corporate data warehouse were collected as part of routine clinical care and were extracted from the corporate data warehouse by our study team. A list of medications included in the corporate data warehouse query is provided in Appendix A.
Secondary outcome measures:
The same BZRA data were extracted for the time period 12 months after randomization. All-cause mortality at 6 and 12 months were extracted using the VHA’s Vital Status and corporate data warehouse files and identified all participants who were deceased during the time period 6 and 12 months after randomization.
Sample Size Calculations
Using published information (Mendes et al., 2018; Tannenbaum et al., 2014), we estimated our study’s power for detecting a difference in BZRA prescriptions for two comparisons: #1) active brochure only vs. control brochure (to detect a difference of 10% versus 5%) and #2) active brochure + telephone reinforcement call vs. active brochure only (to detect a difference of 20% versus 10%). With our original planned sample sizes of n=500 per group, the power exceeded 85% for each comparison (using alpha=0.05 and two tailed tests). We unexpectedly had fewer participants who opted out of the mailing or who had a bad mailing address, so more participants were enrolled, and the sample sizes were changed to N=943 (control brochure), n=946 (active brochure) and n=120 (active brochure plus telephone reinforcement call). Post-hoc power computations showed power=0.985 for comparison #1 and power=0.857 for comparison #2.
Data analysis
Data analysis was performed using Stata 15.1 (StataCorp LLC, College Station, Texas). Descriptive statistics were used to summarize participants’ demographic characteristics and their BZRA prescriptions (chronicity, FDA-approval status of drug) prior to randomization.
Logistic regression analyses:
We modeled presence of BZRA prescription at 6-month (model 1) and 12-month (model 2) follow-up with logistic regression models. For these models, the primary regressor was intervention arm assignment (active brochure only, active brochure + telephone reinforcement, or control brochure), and we controlled for study site (Colorado versus Montana). In additional logistic regression models (model 3: 6-month follow-up 3; model 4: 12-month follow-up), we added FDA drug approval status, indication for insomnia, chronicity of prescription at baseline (< versus ≥ 90 days), and demographics (including age, gender, race, ethnicity, marital status) to the models.
We used Fisher’s Exact tests to compare all-cause mortality (number of deaths) among groups at the 6- and 12-month timepoints.
Results
Participant screening, randomization, and follow up are summarized in Figure 1 (CONSORT diagram). A total of 437 veterans actively opted into the study by returning the opt card (i.e., active opt-in). A total of 1,572 veterans did not return the opt card (i.e., passive opt-in). The top five BZRA prescriptions among veterans invited to participate in the study and among participants at 12-month follow-up were zolpidem (most frequent/1st), clonazepam (2nd), lorazepam (3rd), alprazolam (4th), and temazepam (5th).
Figure 1. CONSORT Flow Diagram.
Table 1 presents participant characteristics. A total of 2,009 veterans participated: active brochure (n=946, 47%), active brochure + telephone reinforcement (n=120, 6%), or control brochure (n=943, 47%). Most participants received care in Colorado (n=1,774, 88%). A majority of participants were male (n=1887, 94%), between the ages of 65 and 74 (n=1,494, 74%), White (n=1,637, 81%), not Hispanic or Latino (n=1,673, 83%), and married (n=1,200, 60%). Of the 2,009 participants, 1101 participants (55%) were on a BZRA approved by the FDA for insomnia, 1,620 participants (81%) had a chronic BZRA prescription (≥ 90 days), and 1,479 participants (74%) had a BZRA prescribed with an indication for sleep.
Table 1.
Participant Demographic and Clinical Characteristics
| Characteristics | Total (n=2009) | Active brochure only (n=946) | Active brochure + Telephone reinforcement (n=120) | Control brochure (n=943) |
|---|---|---|---|---|
| Mean age (SD) | 71.9 (5.6) | 71.6 (5.6) | 72.6 (5.3) | 72.0 (5.6) |
| Age group | ||||
| 65–74 | 1494 (74%) | 710 (75%) | 80 (67%) | 704 (75%) |
| 75–84 | 426 (21%) | 195 (21%) | 36 (30%) | 195 (21%) |
| 85+ | 89 (4%) | 41 (4%) | 4 (3%) | 44 (5%) |
| Gender | ||||
| Male | 1887 (94%) | 887 (94%) | 113 (94%) | 887 (94%) |
| Female | 122 (6%) | 59 (6%) | 7 (6%) | 56 (6%) |
| Race | ||||
| Asian | 4 (<1%) | 1 (<1%) | 0 (0%) | 3 (<1%) |
| Black | 115 (6%) | 49 (5%) | 7 (6%) | 59 (6%) |
| White | 1637 (81%) | 768 (81%) | 102 (85%) | 767 (81%) |
| American Indian/Alaska Native | 31 (2%) | 15 (2%) | 1 (1%) | 15 (2%) |
| Native Hawaiian/Other Pacific Islander | 7 (<1%) | 6 (1%) | 0 (0%) | 1 (<1%) |
| Multirace | 23 (1%) | 11 (1%) | 2 (2%) | 10 (1%) |
| Unknown | 192 (10%) | 96 (10%) | 8 (7%) | 88 (10%) |
| Ethnicity | ||||
| Hispanic/Latino | 237 (12%) | 117 (12%) | 12 (10%) | 108 (11%) |
| Not Hispanic or Latino | 1673 (83%) | 777 (82%) | 103 (86%) | 793 (84%) |
| Unknown | 99 (5%) | 52 (5%) | 5 (4%) | 42 (4%) |
| Marital Status | ||||
| Married | 1200 (60%) | 551 (58%) | 76 (63%) | 573 (61%) |
| Never married | 127 (6%) | 64 (7%) | 8 (7%) | 55 (6%) |
| Divorced | 534 (27%) | 250 (26%) | 29 (24%) | 255 (27%) |
| Widowed | 96 (5%) | 52 (5%) | 5 (4%) | 39 (4%) |
| Separated | 38 (2%) | 20 (2%) | 1 (1%) | 17 (2%) |
| Unknown | 14 (1%) | 9 (1%) | 1 (1%) | 4 (<1%) |
| Site | ||||
| Colorado | 1774 (88%) | 836 (88%) | 102 (85%) | 836 (87%) |
| Montana | 235 (12%) | 110 (12%) | 18 (15%) | 107 (11%) |
| Drug | ||||
| FDA approved for insomnia | 1101 (55%) | 516 (55%) | 71 (59%) | 514 (55%) |
| Chronic prescription (≥90 days) | 1620 (81%) | 762 (81%) | 97 (81%) | 761 (81%) |
| Insomnia indication | 1479 (74%) | 696 (74%) | 90 (75%) | 693 (73%) |
SD=standard deviation
In logistic regression analyses (see Table 2), the unadjusted odds of having a BZRA prescription at 6-month follow-up (model 1) or 12-month follow-up (model 2) were not significantly different between either of the two active groups as compared to the control group (all p-values >0.39). This finding persisted at 6-month follow-up (model 3) and 12-month follow-up (model 4), with these models adjusted for demographics (age, gender, race, ethnicity, and marital status), BZRA prescription chronicity, insomnia indication, and FDA-approval status (all p-values >0.36). In the 6-month follow-up models 1 and 3, compared to participants at the Colorado site, participants at the Montana site had higher odds of a BZRA prescription at 6 months (model 1 p=0.001; model 3 p=0.01), but in the 12-month follow-up models 2 and 4, the Montana site was not statistically significant (model 2 p=0.163; model 4 p=0.07). The BZRA characteristics (chronicity, insomnia indication, and FDA approval status) were significant predictors of BZRA prescription at 6-month follow-up (chronicity OR=7.89, 95% CI [5.15, 12.08]; insomnia indication OR=17.39, 95% CI [9.54, 31.69]; FDA approval status OR=1.86, 95% CI [1.44, 2.40]; all p<0.01). At 12-month follow-up, chronicity of prescription was significant (OR=8.56, 95% CI [5.93, 12.33], p<0.01), but insomnia indication (OR=1.02, 95% CI [0.76, 1.36], p=0.12) and FDA-approval status (OR 1.24, 95% CI [0.97, 1.59], p=0.90) were not statistically significant. We also ran regression models that included opt-in status and found that active opt-in was significant in the 12-month model (OR: 1.45, 95% CI [1.08, 1.95], p=0.01; see Appendix B for model results).
Table 2.
Logistic regression models of BZRA use at follow-up timepoints (1 where 1 is a BZRA within 90 days of the 6-month timepoint or 0 is no BZRA within 90 days)
| Model 1 (6-month) | Model 2 (12-month) | Model 3 (6-month) | Model 4 (12-month) | |
|---|---|---|---|---|
| n=2,009 | n=2,009 | n=2,009 | n=2,009 | |
| Active brochure only | 0.92 [0.76, 1.12] | 0.94 [0.78, 1.14] | 0.90 [0.72, 1.13] | 0.93 [0.76, 1.13] |
| Active brochure + telephone reinforcement | 0.96 [0.64, 1.44] | 1.16 [0.79, 1.71] | 0.89 [0.56, 1.42] | 1.16 [0.77, 1.75] |
| Ref: Control brochure | 1.00 | 1.00 | 1.00 | 1.00 |
| Montana | 1.61a [1.22, 2.12] | 0.82 [0.61, 1.09] | 1.52a [1.10, 2.11] | 0.76a [0.56, 1.02] |
| Ref: Colorado | 1.00 | 1.00 | 1.00 | 1.00 |
| Age (in years) | -- | -- | 0.99 [0.97, 1.01] | 0.99 [0.97, 1.00] |
| Male | -- | -- | 0.79 [0.50, 1.25] | 0.74 [0.49, 1.12] |
| Ref: Female | -- | -- | 1.00 | 1.00 |
| Black | -- | -- | 0.88 [0.54, 1.44] | 0.79 [0.51, 1.23] |
| Asian | -- | -- | 4.89 [0.24, 99.21] | 1.12 [0.11, 10.93] |
| American Indian/Alaskan Native | -- | -- | 1.29 [0.56, 2.94] | 0.94 [0.44, 2.03] |
| Hawaiian/Pacific Islander | -- | -- | 0.75 [0.08, 7.40] | 0.76 [0.13, 4.37] |
| Multi-race | -- | -- | 0.97 [0.33, 2.88] | 1.17 [0.47, 2.91] |
| Missing | -- | -- | 1.17 [0.77, 1.79] | 0.91 [0.62, 1.32] |
| Ref: White | -- | -- | 1.00 | 1.00 |
| Hispanic/Latino | -- | -- | 0.83 [0.59, 1.19] | 0.86 [0.64, 1.17] |
| Missing | -- | -- | 0.87 [0.50, 1.54] | 0.93 [0.56, 1.55] |
| Ref: Not Hispanic/Latino | -- | -- | 1.00 | 1.00 |
| Married | -- | -- | 0.72 [0.46, 1.14] | 0.87 [0.58, 1.30] |
| Divorced | -- | -- | 0.70 [0.43, 1.13] | 0.84 [0.55, 1.28] |
| Separated | -- | -- | 0.51 [0.20, 1.26] | 0.56 [0.25, 1.24] |
| Widowed | -- | -- | 0.65 [0.33, 1.24] | 0.81 [0.45, 1.43] |
| Missing | -- | -- | 1.55 [0.37, 6.55] | 2.78 [0.81, 9.58] |
| Ref: Never Married | -- | -- | 1.00 | 1.00 |
| Insomnia indicationb | -- | -- | 17.39a [9.54, 31.69] | 1.02 [0.76, 1.36] |
| Ref: Did not have insomnia indication | -- | -- | 1.00 | 1.00 |
| Chronic prescription c (≥90 days) | -- | -- | 7.89a [5.15, 12.08] | 8.56a [5.93, 12.33] |
| Ref: Prescription <90 days | -- | -- | 1.00 | 1.00 |
| FDA approved for insomnia | -- | -- | 1.86a [1.44, 2.40] | 1.24 [0.97, 1.59] |
| Ref: Not FDA approved for insomnia | -- | -- | 1.00 | 1.00 |
p<0.05
# of participants with drug without FDA indication: model 3 (n=13), model 4 (n=169); # of participants with drug with FDA indication: model 3 (n= 632), model 4 (n=606)
# of participants without chronic prescription: model 3 (n= 26), model 4 (n=35) # of participants with chronic prescription: model 3 (n=619), model 4 (n=740)
At 6-month follow-up, total all-cause mortality was 49 (2.4%) participants: 29 (3.0%) participants in the active brochure group; 0 (0.0%) participants in active brochure plus telephone reinforcement group; and 21 (2.2%) participants in control brochure group. There were no statistically significant differences between groups (Fisher’s Exact p=0.08). Total all-cause mortality at 12-month follow-up was 104 (5.2%) participants: 51 (5.4%) participants in active brochure group, 1 (<0.1%) participant in active brochure + telephone reinforcement group; and 52 (5.5%) participants in control brochure group. There were no statistically significant differences between groups (Fisher’s Exact p=0.06).
Conclusions
In this pilot randomized controlled trial, we found that a direct-to-patient mailed educational empowerment brochure about sleeping pills that also included information about how to access a free, anonymous, self-directed, online CBTI program did not result in fewer active prescriptions for BZRAs at 6- or 12-month follow-up compared to a program that included mailed non-directed information about sleep. The addition of a single telephone reinforcement call to the mailed brochure also did not result in any statistically significant differences in active BZRA prescriptions. Mortality rates at follow-up were low and not significantly different between groups. Although no statistically significant between-group differences were detected, the primary purpose of the study was to develop the procedures, including refining data extraction methods and outcome measures, for conducting a future full-scale trial. The pilot study results suggest that a more potent intervention is needed.
The intervention was designed to be a low-touch patient empowerment program. Our findings differ from the results of other studies that have examined the effects of similar mailed educational empowerment materials on sleeping pills prescriptions. Tannenbaum et al. examined the absence of benzodiazepine prescription renewals—not including z-drugs—as their primary outcome and found that at 6-month follow-up, 27% of the intervention group had discontinued benzodiazepine use versus 5% in the usual care group in a study conducted among Canadian pharmacies (Tannenbaum et al., 2014). In a quality improvement study within the VHA that used a retrospective cohort design, Mendes et al. examined the effects of the mailed information that was used in the aforementioned Canadian study but was adapted for use within the VHA. Using propensity score matching, Mendes et al. found that the mailed information was associated with 23% of older veterans discontinuing BZRAs, but found that the risk difference for those who received the mailing versus those who did not was lower than the risk difference reported in the Canadian study (Mendes et al., 2018). Unlike Tannenbaum et al.’s study, which included in the mailed materials an example of a tapering schedule, Mendes et al. did not include the tapering schedule. Our study used the same VHA version used by Mendes et al., except that we enhanced the section on sleeping pill alternatives and added information about the online CBTI program, including the web address. A notable difference between our study and the quality improvement study is that we used a randomized trial design that included an attention control condition—a mailed brochure about sleep that was similar in appearance to the active brochure and also included a link to a website, which we carefully designed and hosted to serve as a rigorous comparator to the active group. We also included a telephone reinforcement call for one of the active arms. Another important difference when interpreting our results is that we measured and compared between arms the presence of a prescription at the follow-up timepoint using a look-back time period of 90 days, whereas the other studies focused on cessation, defined as “an absence of any benzodiazepine prescription renewal at the time of the 6-month follow-up that was sustained for 3 consecutive months or more, in the absence of substitution to another benzodiazepine”(Tannenbaum et al., 2014). We opted to focus on the presence of prescriptions, which we felt could be measured with more certainty (i.e., absence could mean the prescription is filled elsewhere or restarted at a later time). Another critical difference between our study and prior work (Tannenbaum et al., 2014) is the inclusion of z-drugs as well as benzodiazepines, given the prevalent use of z-drugs for insomnia.
We aimed to develop trial methods that could be scaled up to include veterans nationwide. However, our results suggest that the low-touch mailed approach with one telephone call may not be potent enough to reduce BZRA prescriptions. One possibility is that participants did not engage with the mailed materials. Secular trends show a decline in patients’ use of mailed information (Office, 2021). It is possible that the mailed brochures remained unopened. We previously explored differences in website use between active and control groups for this study (Mak et al., 2021). A very low number of visits to the website was reported to us in the follow-up participant questionnaire, suggesting a more robust approach to encourage use of the online CBTI program should be considered for a full-scale trial. The telephone reinforcement call was meant to increase participant engagement with the materials. One call, however, did not result in significant differences between control group in BZRA prescriptions; it is possible that additional engagement methods such as additional telephone calls, email, secure messaging, or text reminders would increase engagement.
Sleep is increasingly recognized as a health behavior (Buysse, 2014), and the application of health behavior change theories to sleep health interventions has been proposed for better prediction of sleep health outcomes (Mead & Irish, 2020). Few studies have applied health behavior change concepts to hypnotic discontinuation. In these studies, self-efficacy was related to compliance with a hypnotic taper withdrawal schedule (Belanger et al., 2005), and participants in the self-efficacy enhancement arm of a randomized trial realized increased self-efficacy and a greater percentage of hypnotic medication dosage reduction than controls (Yang, 2015). Belleville and colleagues (Belleville & Morin, 2008) incorporated Transtheoretical Model of behavioral change (TTM) concepts into secondary analyses of a hypnotic discontinuation trial. None of the TTM concepts, including stages of change, decisional balance, or self-efficacy, were related to hypnotic discontinuation. In a more recent study by Yang and colleagues (2022) (Yang et al., 2022), cognitive factors in long-term hypnotic use was examined with constructs from the Theory of Planned Behavior. Perceived behavioral control was the most significant determinant for behavioral intention of hypnotic use, which was then in turn predictive of the frequency of hypnotic use at 3 months. Although more work is needed in this area, health behavior change theory has clear application to the area of hypnotic discontinuation and should be incorporated into future studies.
In our regression models, we found that insomnia indication and chronicity of prescription were significant predictors of having a BZRA prescription at 6-month follow-up and that chronicity of the prescription but not insomnia indication was a predictor at 12-month follow-up. One explanation for the inconsistent findings across time points was small cell size for these covariates, which can affect precision of the estimates. Although the reported point estimates were statistically significant, the 95% confidence intervals (alpha=0.05) were wide (e.g., the odds ratio for insomnia indication at 6 months was 17.4 with a 95% confidence interval [9.5, 31.6]). Post-analysis investigation showed that each of these imprecise estimates were indeed associated with small cell sizes.
Our study has several strengths and weaknesses. In addition to using a randomized design with 12-month follow-up for outcome reporting and including z-drugs in the study, the construction of our outcome was a departure from prior work which had used the absence of a prescription during the follow-up time period as their outcome. We measured and compared BZRA prescription within 90 days of follow-up timepoints between the active and control groups to provide a more accurate depiction of BZRA use at the follow-up timepoints. However, doing so may have presented a more conservative estimate of the impact of the intervention, since anyone who stopped their BZRA for a short period but resumed would be counted as having an active BZRA prescription. Another strength of using a randomized design with a rigorous control condition and objective assessment of BZRA prescriptions from the VHA’s corporate data warehouse is to account for factors like BZRA prescriptions written for indications not related to sleep or the possibility that older veterans may obtain BZRAs from non-VHA pharmacies. The rigorous control condition may have inadvertently caused participants to seek information about insomnia symptoms, which may have led to improvement in sleep and discontinuation in BZRA in the control group. While the strength of a low-touch intervention is its “scalability” factor, the minimal engagement with participants is a limitation in that the research staff would not know what recipients did (or did not do) with the mailed materials. We had the foresight to include a slightly more intensive approach in the active arm (i.e., active brochure plus telephone reinforcement call) in this pilot to provide us with the opportunity to test procedures for such an approach. We conducted the study in a VHA setting, which resulted in a sample with more male patients than in the general population, limiting generalizability to female patients. Also, using ICD-10 codes from the “problem list,” we had excluded Veterans with certain disorders that are contraindications to CBTI such as bipolar disorder, or who have disorders for which benzodiazepines may be appropriate such as REM sleep behavior disorder and seizure, but may have missed other conditions or disorders. In a future full-scale study, we will consider expanding the list to include other disorders such as narcolepsy and psychosis, and we will consider adding ICD-10 code diagnoses from the billing encounters as a source of exclusion. Lastly, a majority of participants would have received the mailed brochure shortly before the COVID-19 pandemic started, which caused widespread sleep disturbance and changes in healthcare patterns of use. Individuals who would have discussed their BZRA use with their healthcare providers may have deferred clinic visits due to the pandemic. Recent studies indicate an increase in z-drug prescriptions among men and women and an increase in benzodiazepine prescriptions among women during the COVID-19 pandemic, reflecting an increase in sleep disturbance that has occurred with the pandemic (Milani et al., 2021). These unexpected changes may have resulted in an external confound to our study and may limit the generalizability of our findings.
Our pilot study informs the development of a future definitive trial. Such a study should include a confirmation that participants had received the mailing and additional steps to promote engagement with the online insomnia treatment and prescriber (e.g., second telephone call, email). A future trial might consider alternative methods of delivering the empowerment material, such as through email or patient portals (e.g., VHA’s myHealtheVet secure messaging). As more older adults use electronic forms of communication, future studies may also consider providing a direct link to the online insomnia treatment/information materials to improve rate of use of the materials. Furthermore, reducing the amount of research staff time spent on assembling paper packets in favor of sending materials electronically could enable staff to redirect time to administering activities that would help participants focus their attention on the mailed material. To inform future implementation efforts, staffing the research program with individuals with clinical training (e.g., nursing staff) who would potentially be available for reinforcement activities in routine clinical practice should be considered. A future study should attempt to include Medicare claims data so that prescriptions filled in non-VHA pharmacies will be counted, for example, through the joint Department of Veterans Affairs (VA)/Centers for Medicare and Medicaid Services (CMS) Data for Research Project.
We conducted a pilot study of a mailed patient brochure about risks of BZRA use and information about accessing a free, online, self-directed CBTI program, compared to a control condition. Although we did not observe statistically significant differences between active and control groups’ BZRA prescriptions at follow-up, our pilot was successful in informing next steps for a future full-scale trial, which could have clinical implications such as demonstrating the efficacy of a low-touch program for reducing active BZRA prescriptions and potentially, important clinical outcomes such as hip fractures and mortality.
Supplementary Material
Clinical Implications.
Clinicians and healthcare organizations interested in reducing BZRA use in older veterans should consider interventions that are more potent than a mailed brochure about the risks of BZRA and information about accessing CBTI.
The addition of a telephone reinforcement call to the mailed brochure was not sufficient to reduce BZRA prescriptions.
Biographies
Selene Mak PhD, MPH is a health services researcher at the VA Greater Los Angeles Healthcare System Center for the Study of Healthcare Innovation, Implementation and Policy (CSHIIP). She conducts research on utilization of mobile health tools and use of patient generated health data in clinical care, including physical activity and sleep.
Cathy A. Alessi MD is a Professor of Medicine at the David Geffen School of Medicine at UCLA and Director of the VA Greater Los Angeles Healthcare Center Geriatric Research, Education and Clinical Center. She is a geriatrician who conducts clinical trials examining interventions to improve sleep in older adults.
Christopher Kaufmann PhD is an Assistant Professor at the University of Florida. His research expertise is in sleep, aging, and mental health services.
Jennifer Martin PhD is a Professor of Medicine at the David Geffen School of Medicine at UCLA and Associate Director for Clinical and Health Services Research at the VA Greater Los Angeles Healthcare Center Geriatric Research, Education and Clinical Center. She is a clinical psychologist who conducts research on improving sleep in older veterans and women veterans.
Michael N. Mitchell PhD is a statistician at the VA Greater Los Angeles Healthcare Center Geriatric Research, Education and Clinical Center. He has been the statistician on multiple studies examining interventions to improve sleep in older veterans.
Christi Ulmer PhD, DBSM is an Associate Professor at Duke University School of Medicine, a clinical health psychologist, and a diplomate in Behavioral Sleep Medicine. At the Durham VA, where she is a Clinical Research Psychologist, she conducts research on adverse health consequences of untreated sleep disorders and developing strategies to increase patient access to behavioral sleep medicine.
Hillary D. Lum MD PhD is an Associate Professor at the University of Colorado, Denver – Anschutz Medical Campus. She conducts patient-centered outcomes research to improve care for older adults with serious illness. She is a geriatrician, palliative care physician.
Michaela S. McCarthy PhD RN is a researcher at the Veterans Health Administration Seattle-Denver Center of Innovation. Her research has focused on telemedicine interventions to treat insomnia in patients who have had breast cancer. She also studies links between poor sleep and suicide in veterans with post-traumatic stress disorder.
Jason P. Smith PharmD is a pharmacist at the Veterans Health Administration VISN 19 Rocky Mountain Network Pharmacy Benefits Management.
Constance H. Fung MD MSHS is an Associate Professor of Medicine at the David Geffen School of Medicine at UCLA and a Staff Physician at the VA Greater Los Angeles Healthcare System. She has clinical training in sleep medicine, geriatric medicine and internal medicine. Her research focuses on behavioral treatments for improving sleep in older adults.
Data Availability Statement:
The data that support the findings of this study are available from the corresponding author, SM, upon reasonable request.
References
- Agarwal SD, & Landon BE (2019). Patterns in Outpatient Benzodiazepine Prescribing in the United States. JAMA Netw Open, 2(1), e187399. 10.1001/jamanetworkopen.2018.7399 [DOI] [PMC free article] [PubMed] [Google Scholar]
- American Geriatrics Society Choosing Wisely Workgroup. (2013). Choosing Wisely. J Am Geriatr Soc, 61(4), 622–631. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Belanger L, Morin CM, Bastien C, & Ladouceur R (2005). Self-efficacy and compliance with benzodiazepine taper in older adults with chronic insomnia. Health Psychol, 24(3), 281–287. https://doi.org/2005-04818-007 [pii]; 10.1037/0278-6133.24.3.281 [doi] [DOI] [PubMed] [Google Scholar]
- Belleville G, & Morin CM (2008). Hypnotic discontinuation in chronic insomnia: impact of psychological distress, readiness to change, and self-efficacy. Health Psychol, 27(2), 239–248. https://doi.org/2008-03424-013 [pii]; 10.1037/0278-6133.27.2.239 [doi] [DOI] [PubMed] [Google Scholar]
- Blom K, Jernelov S, Ruck C, Lindefors N, & Kaldo V (2016). Three-Year Follow-Up of Insomnia and Hypnotics after Controlled Internet Treatment for Insomnia. Sleep, 39(6), 1267–1274. 10.5665/sleep.5850 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Buysse DJ (2014). Sleep health: can we define it? Does it matter? SLEEP, 37(1), 9–17. 10.5665/sleep.3298 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Folmer RL, Smith CJ, Boudreau EA, Hickok AW, Totten AM, Kaul B, Stepnowsky CJ, Whooley MA, & Sarmiento KF (2020). Prevalence and management of sleep disorders in the Veterans Health Administration. Sleep Med Rev, 54, 101358. 10.1016/j.smrv.2020.101358 [DOI] [PubMed] [Google Scholar]
- Gould RL, Coulson MC, Patel N, Highton-Williamson E, & Howard RJ (2014). Interventions for reducing benzodiazepine use in older people: meta-analysis of randomised controlled trials. Br. J. Psychiatry, 204(2), 98–107. https://doi.org/204/2/98 [pii]; 10.1192/bjp.bp.113.126003 [doi] [DOI] [PubMed] [Google Scholar]
- Humpel N, Marshall AL, Iverson D, Leslie E, & Owen N (2004). Trial of print and telephone delivered interventions to influence walking. Prev Med, 39(3), 635–641. 10.1016/j.ypmed.2004.02.032 [DOI] [PubMed] [Google Scholar]
- Jenkins MM, Colvonen PJ, Norman SB, Afari N, Allard CB, & Drummond SP (2015). Prevalence and Mental Health Correlates of Insomnia in First-Encounter Veterans with and without Military Sexual Trauma. SLEEP, 38(10), 1547–1554. 10.5665/sleep.5044 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kaufmann CN, Mojtabai R, Hock RS, Thorpe RJ Jr., Canham SL, Chen LY, Wennberg AM, Chen-Edinboro LP, & Spira AP (2016). Racial/Ethnic Differences in Insomnia Trajectories Among U.S. Older Adults. Am J Geriatr Psychiatry, 24(7), 575–584. 10.1016/j.jagp.2016.02.049 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kaufmann CN, Spira AP, Depp CA, & Mojtabai R (2018). Long-Term Use of Benzodiazepines and Nonbenzodiazepine Hypnotics, 1999–2014. Psychiatr Serv, 69(2), 235–238. 10.1176/appi.ps.201700095 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kontos E, Blake KD, Chou WY, & Prestin A (2014). Predictors of eHealth usage: insights on the digital divide from the Health Information National Trends Survey 2012. J Med Internet Res, 16(7), e172. 10.2196/jmir.3117 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mak S, Ulmer C, Kaufmann C, Alessi C, Martin J, Smith J, Mitchell M, Josephson K, McCarthy M, Lum H, & Fung C (2021). Direct-to-Patient Mailing about Sedative-Hypnotics and Online CBT-I: Participant Reported Use of Study Materials. SLEEP, 44, A141. [Google Scholar]
- Mathieu C, Joly P, Jacqmin-Gadda H, Wanneveich M, Begaud B, & Pariente A (2021). Patterns of Benzodiazepine Use and Excess Risk of All-Cause Mortality in the Elderly: A Nationwide Cohort Study. Drug Saf, 44(1), 53–62. 10.1007/s40264-020-00992-7 [DOI] [PubMed] [Google Scholar]
- Maust DT, Kim HM, Wiechers IR, Ignacio RV, Bohnert ASB, & Blow FC (2021). Benzodiazepine Use among Medicare, Commercially Insured, and Veteran Older Adults, 2013–2017. J Am Geriatr Soc, 69(1), 98–105. 10.1111/jgs.16825 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mead MP, & Irish LA (2020). Application of health behaviour theory to sleep health improvement. J Sleep Res, 29(5), e12950. 10.1111/jsr.12950 [DOI] [PubMed] [Google Scholar]
- Mendes MA, Smith JP, Marin JK, Bounthavong M, Lau MK, Miranda JL, Gray D, Brown M, & Hauser P (2018). Reducing Benzodiazepine Prescribing in Older Veterans: A Direct-to-Consumer Educational Brochure. Fed Pract, 35(9), 36–43. https://www.ncbi.nlm.nih.gov/pubmed/30766385 [PMC free article] [PubMed] [Google Scholar]
- Milani SA, Raji MA, Chen L, & Kuo YF (2021). Trends in the Use of Benzodiazepines, Z-Hypnotics, and Serotonergic Drugs Among US Women and Men Before and During the COVID-19 Pandemic. JAMA Netw Open, 4(10), e2131012. 10.1001/jamanetworkopen.2021.31012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Office USGA (2021, April 29, 2021). U.S. Postal Service: Volume, Performance, and Financial Changes since the Onset of the COVID-19 Pandemic. Retrieved June 14 from https://www.gao.gov/products/gao-21-261#:~:text=What%20GAO%20Found,volume%20when%20compared%20to%202019.
- Olfson M, King M, & Schoenbaum M (2015). Benzodiazepine use in the United States. JAMA Psychiatry, 72(2), 136–142. https://doi.org/2019955 [pii]; 10.1001/jamapsychiatry.2014.1763 [doi] [DOI] [PubMed] [Google Scholar]
- Patel D, Steinberg J, & Patel P (2018). Insomnia in the Elderly: A Review. J Clin Sleep Med, 14(6), 1017–1024. 10.5664/jcsm.7172 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Qaseem A, Kansagara D, Forciea MA, Cooke M, & Denberg TD (2016). Management of Chronic Insomnia Disorder in Adults: A Clinical Practice Guideline From the American College of Physicians. Ann. Intern. Med. 10.7326/M15-2175 [DOI] [PubMed] [Google Scholar]
- Rogojanski J, Carney CE, & Monson CM (2013). Interpersonal factors in insomnia: a model for integrating bed partners into cognitive behavioral therapy for insomnia. Sleep Med Rev, 17(1), 55–64. 10.1016/j.smrv.2012.02.003 [DOI] [PubMed] [Google Scholar]
- Soh HL, Ho RC, Ho CS, & Tam WW (2020). Efficacy of digital cognitive behavioural therapy for insomnia: a meta-analysis of randomised controlled trials. Sleep Med, 75, 315–325. 10.1016/j.sleep.2020.08.020 [DOI] [PubMed] [Google Scholar]
- Tannenbaum C, Martin P, Tamblyn R, Benedetti A, & Ahmed S (2014). Reduction of inappropriate benzodiazepine prescriptions among older adults through direct patient education: the EMPOWER cluster randomized trial. JAMA Intern. Med, 174(6), 890–898. https://doi.org/1860498 [pii]; 10.1001/jamainternmed.2014.949 [doi] [DOI] [PubMed] [Google Scholar]
- Ulmer C, Farrell-Carnahan L, & Hughes J (2017). Improve Your Sleep: A Self-Guided Approach for Veterans with Insomnia (Tatum J, Ed.). VA Mid Atlantic MIRECC. [Google Scholar]
- Yang CM, Lai YS, Huang YH, Huang YC, & Lee HC (2022). Predicting Hypnotic Use among Insomnia Patients with the Theory of Planned Behavior and Craving. Behav Sci (Basel), 12(7). 10.3390/bs12070209 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Yang CMT, Lai CH, Hsu YS, S.C. (2015). Self-efficacy enhancement can facilitate hypnotic tapering in patients with primary insomnia. Sleep and Biological Rhythms, 13(3). https://doi.org/ 10.1111/sbr.12111 [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the corresponding author, SM, upon reasonable request.