Beliefs about prescription sleep medications and interest in reducing hypnotic use: an examination of middle-aged and older adults with insomnia disorder

Abstract

Study Objectives:

To examine beliefs about prescription sleep medications (hypnotics) among individuals with insomnia disorder seeking cognitive behavioral therapy for insomnia and predictors of wishing to reduce use.

Methods:

Baseline data was collected from 245 adults 50 years and older enrolled in the “RCT of the Effectiveness of Stepped-Care Sleep Therapy in General Practice” study. T-tests compared characteristics of prescription sleep medication users with those of nonusers. Linear regression assessed predictors of patients’ beliefs about sleep medication necessity and hypnotic-related concerns. Among users, we examined predictors of wishing to reduce sleep medications, including perceived hypnotic dependence, beliefs about medications, and demographic characteristics.

Results:

Users endorsed stronger beliefs about the necessity of sleep medications and less concern about potential harms than nonusers (P < .01). Stronger dysfunctional sleep-related cognitions predicted greater beliefs about necessity and concern about use (P < .01). Patients wishing to reduce sleep medications reported greater perceived hypnotic dependence than those disinterested in reduction (P < .001). Self-reported dependence severity was the strongest predictor of wishing to reduce use (P = .002).

Conclusions:

Despite expressing strong beliefs about necessity, and comparatively less concern about taking sleep medications, three-quarters of users wished to reduce prescription hypnotics. Results may not generalize to individuals with insomnia not seeking nonpharmacological treatments. Upon completion, the “RCT of the Effectiveness of Stepped-Care Sleep Therapy in General Practice” study will provide information about the extent to which therapist-led and digital cognitive behavioral therapy for insomnia contribute to prescription hypnotic reduction.

Clinical Trial Registration:

Registry: ClinicalTrials.gov; Name: The RESTING Insomnia Study: Randomized Controlled Study on Effectiveness of Stepped-Care Sleep Therapy (RESTING); URL: https://clinicaltrials.gov/ct2/show/NCT03532282; Identifier: NCT03532282.

Citation:

Tully IA, Kim JP, Simpson N, et al. Beliefs about prescription sleep medications and interest in reducing hypnotic use: an examination of middle-aged and older adults with insomnia disorder. J Clin Sleep Med. 2023;19(7):1247–1257.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Although multiple professional organizations discourage prescription of sleep medications based on incidence of adverse drug events and hypnotic dependence, many middle-aged and older adults with insomnia are treated with sleep medications. Previous studies have identified traits associated with chronic hypnotic use, including older age, female sex, and poor physical health; less is known regarding patients’ beliefs about sleep medications and predictors of patients’ interest in reducing hypnotic use.

Study Impact: This study assessed how beliefs about prescription sleep medications differ between users and nonusers seeking behavioral treatment for insomnia disorder and examined predictors underlying these beliefs. Findings increase insight into rates of hypnotic dependence among patients treated for insomnia in primary care settings and variables that predict patients’ sleep medication reduction wishes.

INTRODUCTION

Review of the literature

In primary care settings, prescription of benzodiazepines (BZDs), “z-drugs” (eg, zolpidem, eszopiclone), and other sedating medications (eg, trazodone) is the predominant treatment approach offered to older patients with insomnia.^1,2 Chronicity of prescription sleep medication use has increased over recent decades.³ The National Health and Nutrition Examination Survey collected data from over 80,000 respondents in the United States between 1999 and 2014, and analyses showed that 82% of patients prescribed sleep medications reported at least 6 months of use; 50% reported 2 or more years.⁴ The 2019 American Geriatrics Society Beers Criteria discourage any prescription of sleep medications to older adults, citing frequent issues of polypharmacy, adverse events, and potential for misuse.⁵ While incidence of BZD prescriptions seems to be decreasing, a recent cohort study found that, between 2018 and 2021, prescription of z-drugs increased in adults aged 65 and older.² Although z-drugs are often viewed as a lower-risk alternative to BZDs by patients and health providers,^6,7 the National Institute for Health and Care Excellence guidelines on treating insomnia assert there is no compelling evidence that z-drugs are superior to BZDs in performance or safety.^6,8

Previous studies have found that long-term use of hypnotic BZDs and z-drugs is associated with older age, somatic complaints, female sex, higher levels of psychopathology, concurrent prescription of psychotropic medications, and a history of substance use disorders.^9–11 Less is known regarding patients’ beliefs about using prescription sleep medications to treat insomnia and factors underlying their use. The Health Belief Model ¹² is a useful framework through which to explore how beliefs about sleep and sleep medications affect strategies utilized by patients with insomnia to improve sleep health. Developed to evaluate patients’ (1) motive to avoid or recover from illness and (2) belief in the effectiveness of a specific health behavior at preventing or facilitating recovery from illness, the Health Belief Model is comprised of four primary constructs.¹² In the context of the present study, these constructs are (1) perceived susceptibility to developing insomnia disorder, (2) perceived impact of insomnia symptoms on worsening health, (3) perceived benefit of using prescription sleep medications to reduce insomnia symptoms, and (4) perceived barriers to taking sleep medications to resolve insomnia. Two constructs later added to the Health Belief Model, cue to action and self-efficacy, are important topics for future research but are beyond the scope of the present study.¹³ Previous research has identified reliable vulnerability factors associated with developing insomnia.^9–11 However, few studies have explored the relationship between patients’ beliefs about the impacts of insomnia on health and their co-occurring beliefs about the health benefits and barriers to using prescription sleep medications.

Little is known about factors that influence patients’ interest in or ambivalence about reducing their use of prescription sleep medications. One study offering nonpharmacological treatment for insomnia [eg, cognitive behavioral therapy for insomnia¹⁴ (CBTI)] in concert with hypnotic taper found that older age was associated with a higher rate of refusal to participate in taper¹⁵; however, factors underlying refusal were not systematically documented. Additionally, while chronic sleep medication use and risk of hypnotic-dependent insomnia increase with age, knowledge about the scope and severity of dependence in middle-aged and older adults with co-occurring insomnia and hypnotic dependence is limited.¹⁶ Such knowledge is important because, in addition to ambivalence about using sleep medications, hypnotic dependence can complicate attempts to discontinue use.¹⁷ Specifically, ambivalence may increase the likelihood of resuming use before completing a taper if a patient experiences withdrawal symptoms. This negatively reinforces reliance on sleep medications and increases psychological dependence.^18,19

Current study

The current study addresses important gaps in knowledge discussed here through two aims.

Aim 1a: To characterize demographic and clinical differences between users and nonusers of prescription sleep medications.

Aim 1b: To investigate predictors of patients’ beliefs regarding the necessity of taking sleep medications to manage the impacts of insomnia and concerns about this use.

Aim 2: To assess the frequency and severity of hypnotic dependence and identify predictors of patients’ interest in reducing their use of prescription sleep medications before beginning CBTI. Results could provide insight into patient-level barriers to pursuing hypnotic taper, despite access to affordable CBTI in the context of a clinical trial.

To address these aims, we utilized baseline data from the Randomized Controlled Study on Effectiveness of Stepped-Care Sleep Therapy (RESTING),²⁰ conducted among middle-aged and older adults with insomnia disorder.

METHODS

Study design and participants

Participants were adults age 50 years and older enrolled in the RESTING study, a two-arm randomized controlled trial with a hybrid type 1 effectiveness-implementation design testing two strategies for delivering CBTI. A detailed description of the RESTING study protocol has been published.²⁰ All RESTING study materials and methods were approved by the Stanford University Institutional Review Board, and all participants provided signed informed consent for participation. The present cross-sectional study utilized RESTING study baseline measures that captured perceptions of sleep and sleep-related cognitions, mental health, physical health, and use of sleep medications. Participants had not received any study-related intervention at the time baseline measures were collected.

Participants were recruited from primary care clinics and the Stanford Research Repository.²¹ Targeted advertisement on social media platforms was used to recruit members of the general community. In accordance with RESTING study inclusion criteria, all participants met Diagnostic and Statistical Manual of Mental Disorders (fifth edition) criteria for insomnia disorder (assessed with the Duke Structured Interview for Sleep Disorders²²), were residents of California with an in-state primary care provider, and had reliable internet access. Exclusion criteria were minimal, enhancing generalizability to real-world primary care settings while addressing (1) participant safety (eg, excluding those with high risk of falls and uncontrolled seizure disorder), (2) ability to engage in study treatments (eg, absence of thought disorders and significant cognitive impairment), and (3) confounding factors (eg, uncontrolled thyroid condition, use of substances known to disturb sleep). Data from all eligible RESTING study participants (N = 245) were included in these analyses.

Measures

Demographic information included age, sex, race, civil status, and employment status. Civil status was treated as a binary variable, with groups split between those with any current partner and those with no current partner. Insomnia severity was evaluated with the Insomnia Severity Index.²³

The Beliefs About Medicines Questionnaire-Specific scale (BMQ)²⁴ may provide insight into medication-related decision-making. Validated to assess cognitions and attitudes toward medication use in multiple populations,^25–27 BMQ-Specific subscales evaluate (1) belief in the necessity of taking [specified medication] to protect one’s health [Necessity (BMQ-N)] and (2) concern about the health consequences of taking [specified medication] [Concern (BMQ-C)]. Thus, the questionnaire can help elucidate which beliefs (eg, perception of benefit, perception of cost) most strongly contribute to observed and self-reported medication-related behaviors. Each subscale (range: 5–25) is a sum of five items rated on a 1 to 5 Likert scale. Higher scores reflect a greater belief in the necessity of sleep medications and greater concern about the consequences of use. Internal consistency is satisfactory for both subscales (BMQ-N, α = .76; BMQ-C, α = .72).²⁴ Participants can also be categorized into one of four attitudinal groups based on whether they score above or below the midpoint, a score of 15, on BMQ-N and BMQ-C subscales, respectively: (1) skeptical (low necessity, high concern), (2) indifferent (low necessity, low concern), (3) ambivalent (high necessity, high concern), and (4) accepting (high necessity, low concern). One item from the BMQ-N was adapted for this study by replacing the text, “My health at present depends on my medication” with “My ability to get enough sleep depends on my sleep medicines.”

The Dysfunctional Beliefs and Attitudes about Sleep–16 scale (DBAS; Cronbach’s α = .77–.79),²⁸ which assesses sleep-interfering cognitions, is a well-validated measure that can provide insight into patient perceptions of the extent to which poor sleep impacts their health and ability to function. Items are rated on a Likert scale of 1 to 10 and averaged to create a final score, with higher ratings reflecting more severe maladaptive beliefs about sleep. A DBAS score > 3.8 has been associated with insomnia disorder.²⁹ Four subscales addressing different domains of negative sleep-related cognitions can be calculated: (1) perceived Consequences of insomnia, (2) Worry/helplessness about insomnia, (3) sleep Expectations, and (4) Medications (bolded words indicate subscale names).²⁹ The Medications subscale examines the belief that poor sleep results from a biochemical issue, potentially necessitating a biochemical solution (eg, prescription sleep medication); as such, it can help identify common misconceptions about the etiology of insomnia disorder in middle-aged and older adults and how these beliefs relate to decisions about prescription hypnotic use.

Participant physical health was approximated with the Patient-Reported Outcomes Measurement Information System Global-10 Physical subscale (PROMIS-P; Cronbach’s α = .81³⁰).³¹ This subscale is composed of three questions rated on a Likert scale of 1 to 5. One additional item, “In the past 7 days, how would you rate your pain on average?” is asked on a scale of 1 to 10 and recoded to fit the 5-point scale when calculating the sum total. PROMIS-P total scores range from 4 to 20. A standardized T-score can also be calculated (M = 50, SD = 10), with higher values indicating better physical health.

Psychiatric well-being was measured with the PROMIS Global-10 Mental Health subscale (PROMIS-MH; Cronbach’s α = .86³⁰).³¹ The PROMIS-MH is scored as a sum of four items on a 1 to 5 scale that can be converted to a T-score (M = 50, SD = 10); higher scores represent better psychiatric health.

Participants using prescription sleep medications also completed the five-item Severity of Dependence Scale (SDS; Cronbach’s α = .81),³² which instructs individuals to consider the past 4 weeks when responding. Items (Likert scale: 1–4) are summed, with higher scores reflecting more severe substance dependence. Studies that have compared BZD user SDS scores to an interview-based diagnostic classification of users as “dependent” or “not dependent” identified SDS scores of 5.5, 6, or 7 as optimal for distinguishing patients who met interview criteria for a BZD use disorder and those who did not.^16,32,33

Participants using prescription sleep medications were asked about their treatment goals via the following question: “Which of the following outcomes would make this an effective treatment?” Response options included, “Reducing how much prescription sleep medications I take” and/or “Discontinuing all prescription sleep medications that I take.” Those who selected one or more of these response options were classified as users with a wish/interest in reducing prescription sleep medications. A list of all potential treatment goals is located in Table S1 ^{(455.3KB, pdf)} in the supplemental material.

Statistical analyses

All statistical methods were performed using IBM SPSS Statistics software (version 28). Prior to conducting analyses, survey responses were checked for missing data; a total score was considered missing if < 75% of items were answered. For each measure, we classified a Z-score ≥ 3.0 as a potential outlier to be examined. We conducted sensitivity analysis excluding potential outliers that were identified.

Full sample analyses

A significance level of α = .01 was set for correlation and t test analyses. Bivariate correlations were calculated for all nondemographic, continuous baseline variables that were completed by the full sample. Participants were classified as prescription sleep medication users or nonusers based upon self-reported current use of any hypnotic BZDs, z-drugs, or non-GABA_A receptor agonists prescribed specifically for sleep. A series of Welch’s independent sample t tests were conducted to compare demographic and clinical characteristics of users and nonusers. We used stepwise multivariable linear regression to evaluate predictors of BMQ-N and BMQ-C scores; independent variables were termed significant if P < .05 and were automatically removed from the model if P ≥ .20. For the final models, we examined goodness of fit assumptions (Hosmer-Lemeshow goodness-of-fit test) and collinearity. To compare the prevalence of different BMQ attitudinal profiles between medication users and nonusers, we utilized a Pearson chi-square test. To account for uneven group sizes and a low number of participants in some categories, we also ran a Fisher-Freeman-Halton exact test.

Sleep medication user analyses

Bivariate correlations were computed for all nondemographic continuous baseline variables, including the SDS. To compare patients who reported a wish to reduce and/or eliminate their use of sleep medications with those endorsing no reduction wish, we conducted Welch’s independent samples T-tests. A significance level of α = .01 was set for the aforementioned analyses. Stepwise binary logistic multivariable regression was used to examine predictors of participants’ “wish to decrease hypnotics: yes/no.” The entry criterion was set at P < .05 and the removal criterion at P ≥ .20. Additionally, exploratory chi-square and Fisher-Freeman-Halton exact tests were used to examine BMQ attitudinal differences between patients with and without an interest in sleep medication reduction.

Regression model predictors

Given the paucity of research examining patient beliefs about prescription sleep medications, the demographic (sex, work status, and civil status) and clinical (DBAS, PROMIS-P, and PROMIS-MH) covariates were selected for inclusion in the linear and logistic regression models based upon observed bivariate correlations, an examination of data plots, and relevance to existing insomnia literature. Use of prescription sleep medications (yes/no) was also included in the linear models as a covariate of BMQ-N and BMQ-C scores, given that the BMQ literature suggests facets of medication use (eg, frequency, adherence, chronicity) are associated with patients’ beliefs about medications.^7,34 Additional medication-specific covariates tested in the logistic regression model were BMQ-N, BMQ-C, and SDS scores, as well as the current number of different prescription sleep medications used by a participant. Based upon data plot observations, an interaction term measuring the relationship between civil status and global mental health was added to the logistic model.

RESULTS

We examined skewness and kurtosis statistics for all continuous variables, both within the full sample and within the subsample of medication users. All variables were within bounds of a normal distribution for both samples. Given that we reported Welch’s independent samples t test, which does not assume homogeneity of variance, for all variables, it is not expected that results would be impacted by violated homogeneity of variance.

Our examination of potential outliers revealed that only two participants, both sleep medication users, had a Z-score > 3 on one measure (PROMIS-P: Z = 3.22; BMQ-C: Z = 3.32). We subsequently examined each case and determined that neither was a true outlier. Thus, both cases were included in the final analyses. Nonetheless, we conducted sensitivity analysis for aims 1 and 2, excluding potential outliers, and results did not change.

Sample characteristics

Participants were 245 middle-aged and older adults [M_AGE = 63.09 ± 8.15 (range: 50–87); female = 74.3%] enrolled in the RESTING trial. The majority of participants were White (72.2%) or Asian (17.6%), married or with a common law partner (63.3%), and retired or unemployed (52.7%). We found no statistically significant demographic differences between participants who endorsed taking sleep medications at baseline (n = 89) and those who did not (n = 156). Insomnia severity was also similar between groups. See Table 1 for detailed participant characteristics by medication use status and for the total sample.

Table 1.

Participant characteristics.

Characteristics	Nonusers (n = 156)	Users (n = 89)	Full Sample (N = 245)
Age	63.10 (8.65), 50–87	63.07 (7.24), 50–84	63.09 (8.15), 50–87
Sex
Female	118 (75.6%)	64 (71.9%)	182 (74.3%)
Male	38 (24.4%)	25 (28.1%)	63 (25.7%)
Race
American Indian/Alaska Native	1 (0.6%)	0	1 (0.4%)
Asian	33 (21.2%)	10 (11.2%)	43 (17.6%)
Native Hawaiian/Other Pacific Islander	1 (0.6%)	0	1 (0.4%)
Black or African American	2 (1.3%)	2 (2.2%)	4 (1.6%)
White	105 (67.3%)	72 (80.9%)	177 (72.2%)
More than one race	7 (4.5%)	2 (2.2%)	9 (3.7%)
Unknown/not reported	7 (4.5%)	3 (3.4%)	10 (4.1%)
Civil status
Single	29 (18.6%)	14 (15.7%)	43 (17.5%)
Married or common law partner	97 (62.2%)	58 (65.2%)	155 (63.3%)
Separated, divorced, or widowed	30 (19.2%)	17 (19.1%)	47 (19.2%)
Employment status
Currently working	71 (45.5%)	45 (50.6%)	116 (47.3%)
Unemployed or retired	85 (54.5%)	44 (49.4%)	129 (52.7%)
Prevalence of medical comorbidities	107 (68.6%)	58 (65.2%)	165 (67.4%)
Prevalence of psychiatric comorbidities	20 (12.8%)	23 (25.8%)	43 (17.6%)
Insomnia Severity Index	15.76 (4.32), 5.0–25.0	15.47 (4.04), 7.0–25.0	15.65 (4.21), 5.0–25.0
Beliefs About Medicines: Necessity*	9.89 (3.74), 5.0–25.0	15.30 (4.44), 6.0–24.0	11.87 (4.78), 5.0–25.0
Beliefs About Medicines: Concern*	18.25 (3.65), 8.0–25.0	16.78 (3.98), 5.0–25.0	17.71 (3.83), 5.0–25.0
Dysfunctional Beliefs and Attitudes About Sleep*	6.18 (1.41), 2.88–9.81	6.73 (1.36), 3.63–10.0	6.39 (1.41), 2.88–10.0
Scored > 3.8	149 (96.1%)	87 (97.8%)	236 (96.7%)
Scored ≤ 3.8	6 (3.8%)	6 (2.2%)	8 (3.3%)
PROMIS Physical Health (T-score)	47.64 (6.59), 32.40–67.70	46.78 (7.30), 26.70–61.90	47.33 (6.85), 26.70–67.70
Scored > 60	2 (1.3%)	1 (1.1%)	3 (1.2%)
40 ≤ scored ≤ 60	126 (80.8%)	70 (78.7%)	196 (80.0%)
Scored < 40	28 (17.9%)	18 (20.2%)	65 (18.8%)
PROMIS Mental Health (T-score)	45.19 (7.41), 25.10–67.60	43.79 (7.73), 25.10–62.50	44.68 (7.54), 25.10–67.60
Scored > 60	5 (3.2%)	2 (2.2%)	7 (2.9%)
40 ≤ scored ≤ 60	115 (73.7%)	58 (65.2%)	173 (70.6%)
Scored < 40	36 (23.1%)	29 (32.6%)	65 (26.5%)

Data presented as n (%) or mean (standard deviation), range. *P < .01. PROMIS = Patient-Reported Outcomes Measurement Information System; user = prescription sleep medication user.

Sample results

Correlation analysis revealed that greater belief in the necessity of taking prescription sleep medications to preserve health was significantly associated with higher levels of dysfunctional sleep-related cognitions (DBAS; r = .34, P < .001) and worse mental health (PROMIS-MH; r = –.21, P = .001). The BMQ-C was not significantly associated with any baseline sleep or health characteristics, but greater concern about the negative consequences of using sleep medications was related to lower necessity beliefs (r = –.20, P = .002). Table 2 contains correlation analyses for DBAS subscales.

Table 2.

Correlation matrix for Dysfunctional Beliefs and Attitudes About Sleep subscales.

	Consequences	Worry	Expectations	Medications
Full sample
BMQ-N	.18*	.23*	.03	.61*
BMQ-C	.16	.28*	.05	−.19*
Sleep medication users
BMQ-N	.20	.14	.14	.38*
BMQ-C	.15	.39*	−.02	−.13

Full sample N = 244. Sleep medication users n = 89. *P < .01. BMQ-C = Beliefs About Medicines Concern subscale; BMQ-N = Beliefs About Medicines Necessity subscale; Consequences, Worry, Expectations, and Medications = DBAS subscales.

For group comparisons, sleep medication users reported significantly stronger belief in the necessity of taking sleep medications (BMQ-N; M = 15.30, SD = 4.44) than nonusers (M = 9.89, SD = 3.74), t(159.45) = –9.70, P < .001, d = 1.45, and less concern about the consequences of this use (BMQ-C; users: M = 16.78, SD = 3.98; nonusers: M = 18.25, SD = 3.65), t(170.82) = 2.86, P = .005, d = 0.40. Sleep medication users also scored significantly higher on the DBAS (M = 6.73, SD = 1.36) than nonusers (M = 6.19, SD = 1.41), t(188.87) = –2.96, P = .004, d = 0.39. Of the DBAS subscales, only the Medications subscale differed between users and nonusers of sleep medications at a statistically significant level (users: M = 5.65, SD = 1.76; nonusers: M = 4.07, SD = 1.83), t(189.42) = –6.66, d = 0.86. Physical and mental health did not differ significantly between these groups.

Multivariable linear regression models were used to test if participant baseline characteristics significantly predicted BMQ-N and BMQ-C scores. The final models met multivariable normality and multicollinearity assumptions. Homoskedacity was identified for the BMQ-N model only and was addressed by calculating and comparing the robust standard errors of regression coefficients from a generalized linear model that utilized a sandwich estimator for the estimation of the variance-covariance matrix; differences were inconsequential. Table 3 contains statistical details for all predictors included in the models.

Table 3.

Linear regression models predicting sleep medication necessity and concern beliefs from demographic and clinical covariates.

Predictors	Necessity				Concern
	Model 1		Model 2		Model 1		Model 2
	B	P	B	P	B	P	B	P
Sex	−0.28	.62	–	–	0.25	.65	–	–
Work status	0.73	.15	0.82	.10	−0.37	.44	–	–
Civil status	−0.24	.64	–	–	0.88	.08	0.83	.10
Sleep medication use	5.01	<.001	5.0	<.001	−1.83	<.001	−1.82	<.001
DBAS	0.74	<.001	0.84	<.001	0.62	.001	0.53	.002
PROMIS-P	−0.01	.93	–	–	0.02	.87	–	–
PROMIS-MH	−0.12	.23	–	–	0.11	.27	–	–
Constant	9.98	<.001	10.02	<.001	18.33	<.001	18.26	<.001

DBAS = Dysfunctional Beliefs and Attitudes About Sleep scale; PROMIS-MH = Patient-Reported Outcomes Measurement Information System Mental Health subscale; PROMIS-P = Patient-Reported Outcomes Measurement Information System Physical Health subscale.

For BMQ-N scores, the fitted regression equation was as follows: BMQ-N = 10.02 + 5.0*(sleep medication use) + 0.84*(DBAS) + 0.82*(work status). The overall model was statistically significant, such that R² = .36, F(3, 240) = 45.83, P < .001. We found that use of sleep medications significantly predicted higher BMQ-N scores (β = 5.0, P < .001), as did dysfunctional beliefs about sleep (β = 0.84, P < .001). Work status did not meet the statistical significance threshold as an independent predictor.

For BMQ-C scores, the fitted regression equation was as follows: BMQ-C = 18.26 + –1.82*(sleep medication use) + 0.53*(DBAS) + 0.83*(civil status). The model was statistically significant, such that R² = .08, F(3, 240) = 6.95, P < .001. Use of sleep medications independently predicted lower BMQ-C scores (β = –1.82, P < .001), as did dysfunctional beliefs about sleep (β = 0.53, P = .002). Civil status did not reach the significance threshold.

Figure 1 presents additional details regarding the frequency of BMQ attitudinal profiles within the full sample. Results from the chi-square test comparing attitudinal profiles between prescription sleep medication users and nonusers were statistically significant X²(3, n = 244) = 55.58, P < .001, as were the Fisher-Freeman-Halton exact test findings (P < .001). Of nonusers, 74.2% endorsed a skeptical attitude toward sleep medications as opposed to only 37.1% of users. The remaining nonusers reported ambivalent and indifferent attitudes at similar frequencies, with no non-medication users falling within the acceptance profile. Ambivalent attitudes toward sleep medications were reported by 40.5% of hypnotic users. Remaining users were distributed somewhat evenly between indifference and acceptance profiles.

Figure 1. Attitudes toward prescription sleep medications, as defined by BMQ Necessity and Concern subscale scores, in users and nonusers.

BMQ = Beliefs About Medicines Questionnaire; high concern = score ≥ 15 on BMQ-Concern; high necessity = score ≥ 15 on BMQ-Necessity; low concern = score < 15 on BMQ-Concern; low necessity = score < 15 on BMQ-Necessity; users = prescription sleep medication users.

Sleep medication user results

Correlation analysis revealed that, among participants using sleep medications, the strongest relationships observed were between the SDS and the BMQ-N (r = .49, P < .001) and BMQ-C (r = .43, P < .001). As with the full sample, stronger belief in the necessity of sleep medications was significantly associated with more severe dysfunctional sleep-related cognitions (DBAS; r = .28, P = .009). However, neither necessity nor concern beliefs were significantly related to physical or mental well-being, nor was there a significant association between the BMQ-N and BMQ-C subscales.

Table 4 presents differences between users with a wish to reduce and/or eliminate prescription sleep medications and those with no reduction wish. Severity of hypnotic dependence (SDS) was the only variable significantly differentiating these groups, with participants wishing to reduce sleep medications reporting greater hypnotic dependence (M = 8.63, SD = 3.20) than those disinterested in reduction (M = 5.50, SD = 2.62), t(33.20) = –4.24, P < .001, d = 1.20.

Table 4.

Sleep medication user clinical characteristics.

Characteristics	Goal to Reduce Use (n = 67)	No Reduction Goals (n = 22)	P
Z-drugs	38 (56.7%)	11 (50.0%)	.58
BZDs	20 (29.9%)	7 (31.8%)	.86
Other prescription used for sleep	29 (43.3%)	9 (40.9%)	.85
Trazodone	15	3
Gabapentin	6	3
Doxepin	2	2
Belsomra	2	1
Seroquel	2	0
Sertraline	1	1
Amitriptyline	1	0
Cyclobenzaprine	0	1
Naltrexone	1	0
Tramadol	1	0
Number of sleep medications used	1.36 (0.64), 1.0–4.0	1.36 (0.66), 1.0–3.0	.97
One	48 (71.6%)	16 (72.7%)
Two	15 (22.4%)	4 (18.2%)
Three	3 (4.5%)	2 (9.1%)
Four	1 (1.5%)	0 (0.0%)
Multiple drug classes	17 (25.4%)	5 (22.7%)	.79
Beliefs About Medicines Questionnaire: Necessity	15.81 (3.99), 7.0–24.0	13.77 (5.40), 6.0–24.0	.08
Beliefs About Medicines Questionnaire: Concern	17.25 (3.65), 5.0–23.0	15.32 (4.65), 8.0–25.0	.01
Dysfunctional Beliefs and Attitudes About Sleep	6.74 (1.32), 3.75–10.0	6.68 (1.51), 3.63–9.06	.86
Scored > 3.8	66 (98.5%)	21 (95.5%)
Scored ≤ 3.8	1 (1.5%)	1 (4.5%)
DBAS Subscales
Consequences	7.10 (1.72), 2.20–11.0	6.79 (2.27), 2.20–10.20	.57
Worry	7.61 (1.61), 3.17–11.0	6.68 (1.62), 3.17–8.83	.24
Expectations	6.36 (2.81), 1.0–11.0	7.61 (2.60), 2.50–11.0	.06
Medications	5.58 (1.74), 1.33–10.0	5.86 (1.85), 1.67–8.67	.53
PROMIS Physical Health (T-score)	46.64 (7.23), 26.70–61.90	47.21 (7.66), 34.90–57.70	.76
Scored > 60	1 (1.5%)	0 (0.0%)
40 ≤ Scored ≤ 60	52 (77.6%)	18 (81.8%)
Scored < 40	14 (20.9%)	4 (18.2%)
PROMIS Mental Health (T-score)	43.95 (7.61), 28.40–62.50	43.96 (7.61), 28.40–62.50	.73
Scored > 60	2 (3.0%)	0 (0.0%)
40 ≤ Scored ≤ 60	43 (64.2%)	15 (68.2%)
Scored < 40	22 (32.8%)	7 (31.8%)
Severity of Dependence Scale (n = 80)*	8.63 (3.20), 3.0–15.0	5.50 (2.62), 2.0–9.0	<.001
Scored ≥ 7	44 (70.97%)	7 (38.89%)
Scored < 7	18 (29.03%)	11 (61.11%)

n (%) or mean (standard deviation), range. *Only completed by participants using sleep medications. BZDs = benzodiazepines.

Logistic regression results are detailed in Table 5. Multicollinearity and linearity of the independent variables and log odds (Box-Tidwell test) tests were run and met assumptions. The regression coefficients of the final logistic model were statistically significant, X²(6, n = 89) = 26.63, P < .001, accounting for 47.20% (Nagelkerke R²) of the variance in the decision to decrease sleep medications and correctly classifying 85.0% of cases (decrease yes: 95.20% correct; decrease no: 50.0% correct). Increasing severity of hypnotic dependence was associated with a higher likelihood of wishing to decrease sleep medications (odds ratio = 1.73, 95% confidence interval 1.22–2.47, P = .002), as was better mental health (odds ratio = 2.32, 95% confidence interval 1.14–4.72, P = .02). Sex, work status, and civil status were not statistically significant as independent predictors of this decision. The interaction between civil status and mental health was significantly associated with patient medication goals (odds ratio = 0.34, 95% confidence interval 0.16–0.75, P = .002), such that individuals without a current partner (ie, single, divorced, widowed) who endorsed the wish to reduce sleep medications tended to report better mental health than those without this wish. Mental health did not distinguish individuals in current relationships with different medication use wishes.

Table 5.

Logistic regression models predicting intent to reduce sedative-hypnotic use from demographic and clinical covariates.

Factor	Model 1		Model 2
	OR (95% CI)	P	B	Wald	OR (95% CI)	P
Sex	0.30 (0.07–1.33)	.11	−1.17	1.88	0.31 (0.06–1.65)	.17
Work status	0.51 (0.12–2.160	.36	−1.37	3.21	0.25 (0.06–1.14)	.07
Civil status	0.93 (0.19–4.44)	.93	−2.01	2.41	0.13 (0.01–1.70)	.12
Number of prescription medications for sleep	0.90 (0.18–4.48)	.90	–	–	–	–
BMQ-N	1.03 (0.84–1.25)	.81	–	–	–	–
BMQ-C	1.11 (0.92–1.34)	.28	–	–	–	–
SDS	1.52 (1.10–2.10)	.01	0.55	9.34	1.73 (1.22–2.47)	.002
DBAS	0.90 (0.50–1.62)	.72	–	–	–	–
PROMIS-P	0.94 (0.69–1.29)	.71	–	–	–	–
PROMIS-MH	0.97 (0.71–1.34)	.87	0.84	5.43	2.32 (1.14–4.72)	.02
Civil status by PROMIS-MH	–	–	−1.07	7.30	0.34 (0.16–0.75)	.007
Constant	12.03	.03	−5.37	2.44	0.01	.12

BMQ-N = Beliefs About Medicines Necessity Subscale; BMQ-C = Beliefs About Medicines Concern Subscale; CI, confidence interval; DBAS = Dysfunctional Beliefs and Attitudes About Sleep Scale; OR = odds ratio; PROMIS-P = Patient-Reported Outcomes Measurement Information System Physical Health subscale; PROMIS-MH = Patient-Reported Outcomes Measurement Information System Mental Health subscale; SDS = Severity of Dependence Scale.

A chi-square analysis comparing BMQ attitudinal profiles between sleep medication users with a wish to decrease their use and those disinterested in reduction was statistically significant, X²(3, n = 89) = 12.45, P = .005. The Fisher-Freeman-Halton exact test was also significant (P = .004). Participants that expressed a wish to decrease hypnotics tended to report high concern (85.1%, n = 67), with ambivalence being the most common attitudinal profile (reported by 49.2%) followed by skepticism (reported by 35.8%). Low cell counts for individuals disinterested in reducing or eliminating hypnotics (n = 22) limit the informational value of BMQ attitudinal findings for this subgroup.

DISCUSSION

In alignment with previous publications, the current study found a high rate of prescription sleep medication use among middle-aged and older individuals seeking nonpharmacological treatment for insomnia disorder.^1,2 Z-drugs were the most commonly prescribed hypnotics, with over 50% of users in this sample taking at least one z-drug. In our study, contrary to existing literature, group comparisons identified no significant differences in age, sex, or civil status, nor in global physical or mental health (examined by t tests comparing means and proportions) between patients using vs not using sleep medications.^9–11,35 However, given that this sample has several specific characteristics that are not representative of the general population [restricted age range, largely female, high rates of medical comorbidities (67.4%), and low rates of psychiatric illness (17.6%)], these findings may be due to sample characteristics. Group comparison for the DBAS cut-off score indicative of insomnia disorder (3.8)²⁹ yielded no significant results, as would be expected given that all participants in the RESTING study met full diagnostic criteria for insomnia disorder.

Our findings suggest that users have stronger beliefs about the necessity of taking sleep medications and less concern about the consequences of this use than nonusers. Thus, in order to minimize use of prescription hypnotics among patients with insomnia disorder,^5,8 informing them about the risks of long-term use and effective nonpharmacological treatment alternatives will be important. Tannenbaum and Martin found that directly providing patients with educational materials detailing the harms of BZDs and a stepwise approach to taper altered the medication-related beliefs of adults 65 and older participating in the Eliminating Medications Through Patient Ownership of End Results cluster randomized trial; this remained true among a subsample of individuals with mild cognitive impairment.^36,37 An informational brochure, such as the one used in the Eliminating Medications Through Patient Ownership of End Results trial, could be adapted to address z-drugs and other sedating medications, as well as enhanced to include information about CBTI as an effective alternative to sleep medications. This could help patients make informed decisions about treatment options for insomnia from the outset, potentially reducing first-time prescription of these medications in addition to chronic use.

Enhancement of patient-facing interventions targeting hypnotic taper can be informed by results from the present study that indicate stronger dysfunctional sleep-related cognitions (DBAS) are predictive of stronger beliefs about the necessity (BMQ-N) of sleep medications and less concern (BMQ-C) about the consequences of this use. The association between the DBAS and BMQ-N was driven strongly by the DBAS Medication subscale. This suggests that, alongside information about prescription hypnotics and taper strategies, providing patients with psychoeducational materials that address common concerns about a biochemical etiology of insomnia disorder may increase some patients’ willingness to initiate hypnotic taper. However, associations between DBAS Consequences and Worry subscales and, respectively, the BMQ-N suggest that broader patterns of negative cognitions and perceived impacts of insomnia on daytime functioning and health may also contribute to beliefs about the necessity of hypnotic medications. For example, the association between DBAS and BMQ-N totals may indicate that cognitive rigidity and negative expectations contribute to a contextual lack of patient self-efficacy regarding the ability to control medication use (in sleep medication users) or medication-related consequences to health (in nonusers). Successful reduction/elimination of sleep medications may require changing unhelpful expectations prior to attempting a taper, especially as withdrawal side effects, including transient worsening of sleep, may function to confirm beliefs about the necessity of hypnotic use.³⁸ Patients reluctant to attempt sleep medication reduction may benefit from receiving CBTI before or concurrently with a taper, as it could help shift beliefs about sleep-related self-efficacy and increase willingness to attempt taper.

Although concern beliefs were generally greater in nonusers, findings revealed that most users (77.6%) identified barriers to taking sleep medications to resolve symptoms of insomnia and reported high concern about their use. The majority of participants characterized by an ambivalent attitude on the BMQ (ie, high necessity and high concern) reported using sleep medications, which indicates that they may weigh the consequences of sleep disruption and/or benefits of medication-related improvements to sleep more heavily than concern about harms related to medication use. Future research exploring the relationship between hypnotic self-efficacy, sleep self-efficacy, and belief in the necessity of sleep medications can help clarify factors underlying patients’ use of hypnotics even when they are ambivalent about resulting costs and benefits.

Notably, over one-third of users reported skeptical attitudes (ie, low necessity and high concern), suggesting that factors other than necessity and concern substantively contribute to their decision to take prescription sleep medications. Potentially contributing factors include a lack of knowledge about alternative nonpharmacological treatments, one’s degree of focus on obtaining immediate relief from symptoms, sleep and hypnotic self-efficacy,³⁹ prescriber attitudes about the use of sleep medications, and barriers to nonpharmacological treatment, such as insurance coverage and long waitlists.^6,35,40,41 Better understanding these and other potential factors underlying the use of prescription sleep medications in those with skeptical attitudes could inform whether efforts to decrease hypnotic use need to focus on providers, patients, or both.

Approximately two-thirds of middle-aged and older adults using sleep medications in this sample reported a clinically significant level of hypnotic dependence (SDS score ≥ 7).^16,32,33 This is consistent with concerns that long-term use of prescription sleep medications is associated with a high risk of developing dependence, one of the main factors underlying professional organizations’ guidance discouraging use.^5,8

Two notable findings emerged when we examined predictors of expressing a wish to reduce or eliminate sleep medications. First, a high proportion (86%) of users categorized as hypnotic-dependent (SDS ≥ 7) reported a wish to decrease their use. We believe this rate is likely higher than that in the overall population of hypnotic users given that RESTING study participants were seeking to receive CBTI. Even so, this suggests that hypnotic dependence is, in and of itself, not a barrier to users’ interest in reducing sleep medications. Second, users interested in reduction most commonly held an ambivalent attitude toward hypnotic use. Specifically, 49.2% of these users reported ambivalence toward sleep medications, while skepticism, acceptance, and indifference were less prevalent, reported by 35.8%, 9.0%, and 6.0% of users, respectively. This is consistent with research that suggests intensifying ambivalence can be an important precipitant to transitioning from a phase of precontemplation about substance reduction to active contemplation and action taking.^42,43

Participants in this study were actively seeking CBTI, meaning it is important to consider selection bias as a limitation when interpreting the results. It is likely that hypnotic users in this sample tended to experience less satisfaction and greater concern about sleep medications than the broader population of individuals taking hypnotics. All RESTING study participants also met Diagnostic and Statistical Manual of Mental Disorders (fifth edition) criteria for insomnia disorder, meaning beliefs about medications and severity of dependence cannot be generalized to patients who experience remission from insomnia with pharmacological treatment. Additionally, although the present study did not detect any significant demographic differences between sleep medication users and nonusers, this cannot be interpreted as absence of an effect because of uneven distribution and, in some cases, under-representation of certain demographic groups. Low cell counts for BMQ attitudinal profiles among sleep medication users, particularly those without a wish to decrease use, also prevent substantive interpretation of chi-square test results. Future research should examine a more diverse sample of sleep medication users and explore patients’ interest in reducing prescription hypnotics when fast, affordable access to CBTI is not guaranteed.

Future research grounded by the Health Belief Model can help identify internal cues (eg, daytime sleepiness, depressed mood) and external cues (eg, advice from doctors, information online) that predict patients’ engagement in hypnotic taper. Examining the relationship between hypnotic self-efficacy, sleep self-efficacy, and taper outcomes could improve knowledge about when it is most effective to address beliefs about sleep medications (ie, before, after, or concurrently with dysfunctional sleep-related cognitions) and would contribute to ongoing research evaluating how to better harness sleep-related beliefs as clinical tools.³⁹

CONCLUSIONS

We found that, as would be expected, prescription sleep medication users endorsed stronger beliefs about the necessity of these medications and less concern about the consequences of use than nonusers. Additionally, we found that the majority of users scored above threshold for hypnotic dependence on the SDS and that, nonetheless, most of these patients expressed a wish to reduce or discontinue use. Results show that the BMQ attitudinal profile of ambivalence was predominant among users who expressed a wish to reduce/discontinue sleep medications. It is unknown whether and how the different attitudinal profiles about use of sleep medications impact successful hypnotic reduction among those seeking/undergoing CBTI. Upon its completion, the RESTING study could provide insight into this issue. Also unknown, and a direction for future research, is whether strengthening ambivalence about sleep medication use in patients who do not report a wish to reduce hypnotics may support future consideration of taper.

DISCLOSURE STATEMENT

All authors have seen and approved this manuscript. Work for this study was performed at Stanford University School of Medicine. The Stanford RESTING Insomnia Study is supported by Award Number R01AG057500-03 from the National Institute of Aging. The REDCap platform services at Stanford are supported by Stanford School of Medicine Research Office and by the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through Grant UL1 TR001085. The cohort identification is performed by the Research Informatics Center at Stanford University and is supported, in part, by the Stanford CTSA Award Number UL1TR003142. Ms. Isabelle Tully, Dr. Norah Simpson, Dr. Latha Palaniappan, and Dr. Joshua Tutek have not received financial support and have no conflicts of interest to report. Dr. Jane Kim has received National Institutes of Health research funding and reports no conflicts of interest. Dr. Nicole Gumport has received National Institutes of Health research funding and reports no conflicts of interest. Dr. Jessica Dietch has received financial support from the National Heart, Lung and Blood Institute and reports no conflicts of interest. Dr. Rachel Manber has received National Institutes of Health financial support and reports no conflicts of interest.

ABBREVIATIONS

BMQ

Beliefs About Medicines Questionnaire (Specific Version)

BMQ-C

Beliefs About Medicines Questionnaire (Specific Version) Concern subscale

BMQ-N

Beliefs About Medicines Questionnaire (Specific Version) Necessity subscale

BZD

benzodiazepine

CBTI

cognitive behavioral therapy for insomnia

DBAS

Dysfunctional Beliefs and Attitudes About Sleep Scale

PROMIS-MH

Patient-Reported Outcomes Measurement Information System (Global 10) Mental Health subscale

PROMIS-P

Patient-Reported Outcomes Measurement Information System (Global 10) Physical Health subscale

RESTING

RCT of the effectiveness of stepped-care sleep therapy in general practice

SDS

Severity of Dependence Scale