Cardiac events and economic burden among patients with hypertension and treated insomnia in the USA

Hypertension (HTN) is a major risk factor for adverse cardiovascular (CV) events, including stroke, heart disease and chronic kidney disease. Insomnia disorder, defined as difficulty initiating and/or maintaining sleep with associated daytime impairment [1], is highly common among patients with HTN and is also a known risk factor for developing HTN and subsequent CV events [2–6]. Perhaps due to increased vascular endothelial function or an increase in sympathetic activity [7], insomnia is associated with, for example, coronary heart disease, incident myocardial infarction and mortality [7–9]. Where insomnia is ‘frequent, chronic and/or accompanied with short sleep duration,’ a strong association exists between insomnia and HTN/blood pressure [10]. Short sleep duration is particularly predictive of negative CV outcomes [11,12]. In a 7.5-year longitudinal study, compared with normal sleepers (≥6 h sleep), chronic insomniacs (≥1 year duration) with short sleep duration had the highest risk for incident HTN (odds ratio [OR]: = 3.8; 95% CI: 1.6–9.0) [4]. Further, the ORs of all-cause mortality associated with HTN were 1.77 (95% CI: 1.07–2.92) for individuals who slept at least 6 h, 2.78 (95% CI: 1.47–5.24) for those who slept between 5 and 6 h and 3.93 (95% CI: 2.22–6.95) for those who slept <5 h [3].

Insomnia is vastly underdiagnosed and undertreated in the USA, particularly among individuals at risk for CV conditions. In one recent study, 39% of patients at risk for CV conditions screened positive for insomnia, although only 3.8% of them had previously been diagnosed with the condition – a difference of more than 10-fold [13]. Given the impact that insomnia can have on CV outcomes, this situation underscores a clear unmet need in patients with HTN.

Medications are the most frequently used treatment for insomnia. Note, however, that the American Academy of Sleep Medicine recommends first-line use of cognitive behavioral therapy for insomnia (CBT-I) [14]. Unfortunately, access to CBT-I is limited in the USA because of system barriers to access services, clinician underutilization and patient underengagement [15]. In a large study of US veterans, 17% of insomnia patients were identified by one of the four indicators; medications were most common (15%), followed by diagnoses (6%), consults (1.5%) and CBT-I (0.6%) [16].

Broadly speaking, medication treatment of insomnia has been shown to improve sleep and quality of life in patients with comorbidities, including HTN and other CV diseases [17]. However, at the same time some commonly prescribed older-generation insomnia medications (e.g., benzodiazepines, trazodone, ‘z-drugs’ [nonbenzodiazepine sedative hypnotics, e.g., zolpidem]) have been associated with increased risk of falls [18–23] and worsened economic outcomes [24]. The argument against use of benzodiazepines and zolpidem is made explicit in the Beers Criteria, which is managed by the American Geriatrics Society. According to current Beers criteria, the use of these medications in the elderly is not recommended [25]. In addition, it recommends avoiding benzodiazepines and z-drugs for an extended period of time due to addiction potential, as well as falls and other potential negative consequences [25]. Trazodone is not US FDA-approved to treat insomnia, and the published literature documents a questionable risk-to-benefit profile in insomnia [26].

Drugs commonly used to treat insomnia – including benzodiazepines and z-drugs – have been associated with adverse effects. Benzodiazepine adverse effects may include cognitive or memory impairment; rebound insomnia; and increased risk of falls, motor vehicle accidents and dependence/addiction [27,28]. z-drugs are associated with dependence/addiction and next-day cognitive, memory and psychomotor impairments [27,28]. Clinicians often prescribe low-dose trazodone for insomnia management. Although some physicians may view trazodone as a safer alternative due to the lack of anticholinergic activity and cardiotoxicity, limited data support its efficacy or safety [14]. The most frequent adverse effects associated with trazodone use include daytime sleepiness, dizziness and hangover [29], which may heighten falls risk, particularly when an individual awakens from sleep [19].

From a clinical and economic perspective, understanding the real-world impact of insomnia treatments among patients with HTN is vital to inform clinical decision-making. From a public health perspective, such knowledge can provide evidence-based guidance to payers and policy makers responsible for managing health for the future.

To address this gap in knowledge, the objective of the current study was to evaluate the clinical and economic impact of older-generation insomnia medications among a national sample of patients with HTN and treated insomnia (H + TI). The authors hypothesized that, relative to individuals with HTN but no sleep disorders, individuals with H + TI would demonstrate increased risk of adverse clinical (CV events) and economic (healthcare resource utilization [HCRU] and costs) outcomes.

Materials & methods

Data source

This retrospective cohort study used medical and pharmacy claims from the IBM MarketScan Commercial and Medicare Supplemental Databases from January 2011 to September 2018, comprising longitudinal data on healthcare services for >41 million persons in the USA with employer-based healthcare coverage. The Commercial Database includes information from preferred provider organizations, exclusive provider organizations, point of service plans, indemnity plans, health maintenance organizations and consumer-directed health plans. The Medicare Supplemental Database includes administrative claims with diagnostic and procedural codes and associated costs for all Medicare services utilized by individuals with employer-sponsored supplemental coverage to Medicare. Because this study used fully deidentified administrative data, it did not require institutional review board review.

Study population

Patients with H + TI were required to have more than one International Classification of Diseases (9th or 10th revisions; ICD-9 or ICD-10) diagnosis code for HTN and more than one prescription for an older-generation insomnia medication during the study period. These medications were identified as the most frequently used in the USA for insomnia treatment and included zolpidem (both immediate-release [IR] and extended-release [ER] formulations) used by 1.23% of the US population, low-dose trazodone (<100 mg) used by 0.97% of the US population and benzodiazepines (as a class) used by 0.40% of the US population [30]. Because not all of these medications are FDA approved for insomnia, additional criteria were used to maximize specificity for TI; off-label benzodiazepines (clonazepam, lorazepam, alprazolam) were required to be accompanied by one or more physician-assigned ICD-9/ICD-10 diagnosis code for insomnia within 12 months before first insomnia medication claim. The study period was 1 January 2012 to 30 September 2018. The earliest insomnia medication fill date was considered the index date. Additional inclusion criteria included ≥18 years old with ≥12 months of continuous health plan enrollment before (i.e., ‘baseline’ period) and after (i.e., ‘follow-up’ period) the index date. Exclusion criteria included the presence of a prescription claim for an insomnia medication of interest during baseline; claims for a single insomnia treatment with <5 days’ supply; an index prescription for benzodiazepines and an anxiety diagnosis during the baseline period; or missing age or sex.

To quantify the impact of insomnia treated with older-generation medications on HCRU and healthcare costs, non-sleep-disordered controls with HTN but without evidence of any sleep-related diagnosis, testing, or treatment were identified and age- and sex-matched 1:1 to H + TI participants. The index date for each control was defined as the index date of the matched H + TI patient. Additional inclusion criteria included continuous health plan enrollment for ≥12 months before and after the index date. Exclusion criteria included evidence of insomnia diagnosis or prescription insomnia medication use; or evidence of any sleep-related diagnoses (i.e., ICD-9 or ICD-10 diagnostic codes for hypersomnia, sleep-related breathing disorders, circadian rhythm sleep disorders, parasomnia, sleep-related movement disorders, and drug-induced sleep disorders) or sleep-related treatment (i.e., Current Procedural Terminology codes for sleep study procedures, sleep service codes, home sleep apnea testing or Healthcare Common Procedure Coding System codes for durable medical equipment).

Study measures

Demographic characteristics included patient age as of study index date, sex, geographic region and health plan type. Comorbidity burden was assessed using the Charlson Comorbidity Index (CCI); a higher CCI score indicates greater overall comorbidity burden [31]. CV events were defined as more than one inpatient or emergency department (ED) claim for any of the following conditions during the 12-month follow-up period: angina, atrial fibrillation, coronary artery disease, myocardial infarction, ischemic stroke or heart failure. CV hospitalizations were defined as inpatient visits related to CV events, and CV-related ED visits were similarly defined. All-cause HCRU included encounters in the following points of service: inpatient, ED, outpatient, skilled nursing facility, hospice, home health agency and prescription medication use. Inpatient visits were based on inpatient admission records, and ED visits were identified as an outpatient visit with ED as the place of service. Variables pertaining to provider type, place of service and facility bill type were used to identify specific visit types. Visits attributable to skilled nursing facilities, hospices and home health agencies were not reported separately, although these costs were included in total cost calculations.

All-cause healthcare costs were computed based on all medical expenditures across each HCRU category. Within each HCRU category, costs were evaluated among patients with ≥1 HCRU claim for that specific category of HCRU (e.g., total inpatient costs represented those associated with inpatient visits among the subset of patients with one or more inpatient visit). Prescription drug costs were calculated using pharmacy claims data. All costs were estimated per patient per month. Costs were inflated to 2018 costs using the medical care component of the Consumer Price Index.

Statistical analysis

Descriptive analyses were performed for all measures, including means and standard deviations for continuous and frequency counts and percentages for categorical variables. Cardiac endpoints were described for H + TI by drug of interest (zolpidem IR, zolpidem ER, trazodone, benzodiazepines [as a class]), and for the matched control cohort during follow-up. To test our hypothesis that TI is associated with adverse outcomes among patients with HTN, logistic regression models were developed to calculate odds ratios (ORs) with 95% CIs comparing H + TI patients with matched controls (reference group). The hazard ratio (HR) and 95% CI for time to first event was estimated among the study cohorts using Cox proportional hazard models with robust variance estimation. Generalized linear models with Poisson distribution and log link were used to estimate adjusted means and estimated mean ratios (EMR) with 95% CIs for HCRU visits per month and length of stay outcomes. To compare cost between H + TI and control patients, generalized linear models with gamma distribution and log link were used to estimate mean cost per patient per month along with mean ratio and 95% CIs. For all adjusted models, p-values and 95% CI between H + TI and the matched control cohort were reported without multiplicity adjustment. In addition to age and sex, which were used in matching, other covariates for adjusted models included geographic region, baseline CCI score, and health plan type.

Sensitivity analyses

Sensitivity analyses were conducted to assess the robustness of study results based on differing operational definitions of TI. Specifically, whereas the overall sample included individuals with insomnia who could have also had other sleep disorders, the goal of sensitivity analyses was to ensure accurate attribution of costs to TI and not other sleep-related diagnoses (e.g., obstructive sleep apnea). Thus, sensitivity analyses excluded H + TI patients with any evidence of other, non-insomnia sleep-related diagnoses or procedures. Finally, analyses were rerun to compare outcomes including inpatient visits, length of inpatient stay, ED visits and total costs between patients with H + TI and matched controls.

Results

Study population attrition & characteristics

A total of 4,336,874 adults were treated with insomnia medications of interest for more than 5 days (Figure 1) during the study period. The cohort of adults with HTN but without sleep disorders included 3.3 million patients. The initial matching process included 318,298 hypertensive patients treated for insomnia matched 1:1 on age and sex to hypertensive patients with no sleep disorders; after application of continuous eligibility criteria for controls, the final matched cohort comprised 81,502 adults in each group. Of this final H + TI cohort, 46.9% received zolpidem IR, 32.3% trazodone, 18.4% benzodiazepines and 2.4% zolpidem ER.

Demographic and baseline clinical characteristics were similar for both H + TI patients and controls (Table 1). The mean (standard deviation) age was 62.1 (13.5) years, and 53% of patients were female. A higher proportion of H + TI patients had a CCI score of ≥3 (24.1 vs 14.1%), suggesting a substantial comorbidity burden among these patients.

Cardiac events & healthcare costs

Overall, patients with H + TI demonstrated worsened health and economic outcomes. Relative to controls, H + TI patients were more likely to experience a CV hospitalization or ED visit (10.8 vs 4.4%) (Table 2). Similarly, relative to controls, H + TI patients were more likely to experience at least one CV hospitalization (8.4 vs 3.2%) and at least one ED visit (4.8 vs 1.9%). Relative to controls, H + TI patients had more than twofold higher adjusted odds (OR: 2.39, 95% CI: 2.29–2.49) of CV hospitalizations or ED visits (Figure 2). The adjusted ORs and HRs were significantly elevated for each insomnia medication of interest, CV-related ED events, and CV hospitalizations. H + TI patients taking trazodone (OR: 2.37, 95% CI: 2.20–2.55), zolpidem IR (OR: 2.39, 95% CI: 2.25–2.55) or benzodiazepines (OR: 2.49, 95% CI: 2.26–2.74) had a similar likelihood of CV hospitalizations or ED visits. Relative to other medications, zolpidem ER was associated with the lowest likelihood of CV hospitalizations or ED visits (OR: 1.60, 95% CI: 1.18–2.19). Although mean number of inpatient visits were similar, relative to controls, H + TI patients had a higher mean number of ED visits (0.2 vs 0.1, p < 0.001), outpatient visits (1.9 vs 1.0, p < 0.0001) and prescriptions (1.9 vs 1.3, p < 0.001) (Figure 3).

Relative to controls, the estimated mean per patient per month costs among H + TI patients was significantly higher (US$2343.11 vs US$1013.18; estimated mean ratio, 2.31; 95% CI: 2.28–2.34; p < 0.0001). Among individuals with at least one inpatient visit and relative to controls, H + TI patients had higher estimated adjusted mean inpatient costs (US$3322.88 vs US$2794.95; EMR = 1.19, 95% CI: 1.15–1.23, p < 0.001) (Figure 4A). Among those with at least one ED visit and relative to controls, H + TI patients demonstrated higher estimated adjusted mean ED costs (US$280.35 vs US$211.28; EMR = 1.33, 95% CI: 1.2–1.36, p < 0.001) (Figure 4B). Compared with controls, estimated adjusted mean outpatient costs among those with at least one outpatient visit was higher among H + TI patients (US$853.47 vs US$392.94; EMR = 2.17, 95% CI 2.14–2.20, p < 0.001) (Figure 4C). Among individuals with prescription medication fills, relative to controls, H + TI patients demonstrated higher estimated adjusted mean pharmacy costs (US$302.22 vs US$189.00; EMR = 1.6, 95% CI: 1.58–1.62, p < 0.001) (Figure 4D).

Sensitivity analyses supported the robustness of the main study analyses. After exclusion of H + TI patients with sleep diagnoses and procedures other than insomnia (and their matched controls), key study outcomes – hospital visits, length of inpatient stay, ED visits and total costs – were consistent with the findings of the main analyses. For example, in the sensitivity analysis, estimated mean costs were $2254.00 for H + TI compared with $1014 for controls, and odds of cardiac event-related hospitalization or ED visit were 2.19 (95% CI: 2.09–2.29). Results of all the sensitivity analyses reflected similar estimates and statistical significance as the main analyses.

Discussion

In this national, real-world analysis of patients with HTN, TI was associated with increased risk for adverse clinical (CV events) outcomes and greater economic (HCRU and costs) burden. Relative to non-sleep-disordered controls, patients with H + TI demonstrated 2.4-fold higher adjusted odds of CV hospitalizations or ED visits, as well as higher HCRU and costs across nearly all points of service, during the 12 months after initiation of older-generation insomnia medications.

From a clinical perspective, an important and novel finding from the present study is the elevated risk for adverse CV events among hypertensive patients with older-generation medication-treated insomnia compared with those without sleep disorders. Prior research has suggested an association between sleep disorders and adverse CV outcomes [32], and our study expands these prior findings to include insomnia treated with commonly prescribed medications. From a public health perspective, our study also contributes the first data to demonstrate health care costs of insomnia or insomnia treatment among patients with HTN. Relative to HTN controls without sleep disorders, H + TI patients demonstrated total mean health care costs that were 19% higher. However, prior studies have shown that increased healthcare costs are associated with insomnia treated with older-generation insomnia medications [33,34]. Among Medicare beneficiaries aged >65 years, Amari et al. found that adjusted overall total per patient per month mean costs were higher for a patient cohort treated with older-generation insomnia medications compared with matched controls (US$967 vs US$455, p <0.001) [35].

It is also important to consider medication-specific implications of our findings. Patients treated with benzodiazepines, zolpidem IR or trazodone had a higher risk of CV-related hospitalization or ED admissions compared with zolpidem ER. The reasons for the relatively favorable results for zolpidem ER compared to other insomnia medications of interest is unknown, and warrants further research. Notably, relative to non-sleep-disordered controls, beneficiaries treated with trazodone demonstrated higher risk of hospitalization/ED admission, and risk of hospitalization or ED admission for trazodone-treated patients was not lower than the risk for patients treated with zolpidem IR or benzodiazepines.

Our study’s findings, particularly related to comparable outcomes observed for patients treated with low-dose trazodone relative to benzodiazepines and zolpidem IR, have important clinical implications. Low-dose trazodone is widely prescribed for off-label insomnia treatment in the USA [19] due to a misperception that it is a safer alternative to these medications, with a more favorable risk-to-benefit profile [14,26]. In fact, American Academy of Sleep Medicine Clinical Guidelines recommend that low-dose trazodone not be used for insomnia management, citing limited efficacy and safety data, and suggest that this misperception of trazodone as a ‘safer’ insomnia medication have resulted in widespread trazodone prescribing for insomnia off-label [14]. In the context of widespread use of trazodone for insomnia management, our study’s findings, which add to the literature suggesting that trazodone may not have a better risk-to-benefit profile than other older-generation insomnia medications, have important implications for population and patient health.

Strengths & limitations

Strengths of the current study include the use of real-world data from a large, nationally representative sample of commercially insured adults. We also evaluated both clinical and economic outcomes, thus proving empirical insight to both providers and payers. At the same time, several limitations warrant consideration. First, we were unable to assess clinical variables of interest, such as sleep parameters or daytime impairment because that information is unavailable in health insurance claims data. Likewise, we were unable to assess whether a patient’s HTN was controlled or uncontrolled but only whether they had the condition. Second, our study was limited to insomnia prescription medication treatments; we were unable to assess other insomnia treatments, such as over-the-counter treatments, herbal remedies, cognitive-behavioral treatments or other insomnia treatments, which have been shown to account for up 34.4% of the sleep products used by older adults [36]. Because we could not control for these other insomnia treatments, we did not include patients with untreated insomnia in the present study and were unable to assess whether the observed differences in HCRU were a result of insomnia treatment among H + TI or insomnia itself among hypertensive patients. Third, health insurance claims data do not contain potentially relevant covariates such as BMI, severity of illness, health status, physical activity, exercise and fitness [37,38]. Because we did not have information on severity of illness, we pooled all patients together regardless of their extent of medication use. Fourth, although our sample was national, MarketScan databases include individuals with employer-sponsored health insurance, and the generalizability of our findings to other populations (i.e., government-sponsored insurance or no insurance) is unknown. Fifth, we attributed outcomes to a patient’s index medication; information was unable to determine how long or frequently patients took their medication. Finally, our study, like all observational studies, only demonstrates association and precludes determination of causality.

Conclusion

Among patients with HTN and relative to non-sleep-disordered controls, individuals treated with older-generation insomnia medications were more likely to experience CV hospitalizations and ED visits. Relative to controls, H + TI patients also had greater HCRU and incurred mean health care costs that were more than twofold. Finally, it is notable that drug characteristics (e.g., therapeutic class and/or mechanism of action) were associated with outcomes because CV risk was higher for patients treated with trazodone, benzodiazepines or zolpidem IR than among those treated with zolpidem ER. Future studies should explore the impact of treated versus untreated insomnia. They should also assess the impact of alternative insomnia treatments including newer insomnia medications on CV and economic outcomes (e.g., direct and indirect costs) among patients with HTN.

New medications are being developed to provide better treatment options for insomnia patients. A new type of insomnia medication – dual orexin receptor antagonists (DORAs) – has entered the US marketplace. Current DORA options include lemborexant, suvorexant, daridorexant and possibly seltorexant in several years. DORAs offer the hope of treating sleep onset latency and wake after sleep onset issues without some of the side effects of older-generation medications, such as daytime impairment, addiction potential and falls. Further work needs to be done to evaluate the impact of these new medications on insomnia in HTN patients.