Insomnia Subtypes and the Subsequent Risks of Stroke
Report From a Nationally Representative Cohort
Background and Purpose—The studies assessing the impact of insomnia on stroke are still lacking. We aim to investigate insomnia in relation to subsequent stroke during the 4-year follow-up.
Methods—Data from the Taiwan National Health Insurance Research Database were used. Enrollees with International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis code for insomnia were compared with randomly selected, age- and sex-matched noninsomnia enrollees with subsequent hospitalization for stroke during the 4-year follow-up. All enrollees, insomniacs and noninsomniacs, did not have previous diagnosis of stroke, sleep apnea, and insomnia. Individuals with insomnia were further categorized into different subgroups based on their insomnia patterns to explore whether the risk of stroke varies by subtype. The risk of outcomes was assessed with Kaplan–Meier curves and the impact of insomnia was estimated using Poisson regression analysis and Cox proportional hazards models.
Results—The study included 21 438 (mean age, 52±16 years) insomniacs and 64 314 matched noninsomniacs (mean age, 51±16 years). Compared with noninsomniacs, insomniacs had 54% higher risk of developing stroke (adjusted hazard ratio, 1.54; 95% confidence interval, 1.38–1.72). When breaking down into insomnia subgroups, the persistent insomniacs had a higher 3-year cumulative incidence rate of stroke than those in the remission group (P=0.024). The insomniacs-to-noninsomniacs incidence rate ratio for stroke was highest among those aged 18 to 34 years (incidence rate ratio, 8.06).
Conclusions—Insomnia predisposes individuals to increased risk of stroke and this association is profound among young adults. Our results underscore the clinical importance of identifying and treating insomnia. A novel behavioral intervention targeting insomnia that may prevent stroke should be explored.
Sleep problems and cardiovascular-related diseases often coexist. More recent reports1,2 have focused on habitual sleep duration (short or long), influences of sleep apnea on cardiovascular disease outcomes, and cardiovascular disease–related mortality. Nevertheless, there has been limited examination of the influences of insomnia on cardiovascular disease–related diseases, especially stroke. Research on sleep and stroke has focused on the relationships among sleep apnea,3–5 sleep-disordered breathing,6 sleep duration,2 and stroke; much less attention has been paid to insomnia. The few studies examining the associations between insomnia and stroke were limited by focusing on insomnia among patients with stroke.7,8 In a group of 277 ischemic patients, more than half of them reported insomnia complaints and ≈38% had insomnia symptoms before stroke.7 However, without the use of longitudinal study designs, it is unclear whether the presence of insomnia may be antecedent to the development of stroke.
An additional, although largely unexplored, issue is the course of insomnia over time9 and subsequent risks of stroke. Previous research has indicated that insomnia is often chronic and could be a persistent condition; the persistence rates ranged from 32% to 69% during the observational period of 1 to 20 years.10–13 The remission rate of insomnia varied from 54% to 56% during the follow-up of 1 to 3 years.14 Taking into consideration of potential dynamic changes in insomnia status, it is possible that different patterns of insomnia may predispose individuals to different magnitudes of risk for the same disease.
Therefore, we conducted a longitudinal analysis of nationwide data to assess the influence of insomnia on the risk of subsequent hospitalization for stroke across the 4-year follow-up.
Taiwan launched a single-payer National Health Insurance (NHI) program in 1995 to finance healthcare. The NHI is a mandatory system, and all outpatient and inpatient medical benefit claims are included in the NHI Research Database. Data used in this study came from the Longitudinal Health Insurance Database 2000, which contains all claims data (from 1996 to 2009) of 1 million beneficiaries who were randomly selected from the NHI Research Database in 2000, representing ≈5% of all enrollees in Taiwan. Confidentiality assurances were addressed by abiding by data regulations of the Bureau of NHI, and a formal written wavier for ethical approval was obtained from the Chi-Mei Medical Center Institutional Review Board.
We conducted a retrospective cohort study, which included a case cohort and a comparison cohort based on outpatient and inpatient claims in the Longitudinal Health Insurance Database 2000. The study cohort consisted of all patients who met the following criteria between January 1, 2002, and December 31, 2005: (1) ≥1 single hospitalization with a diagnosis code for insomnia or (2) ≥3 outpatient visits with a diagnosis code for insomnia within the same year. From subjects without insomnia and stroke in the Longitudinal Health Insurance Database 2000, a comparison group was assembled by matching an insomniac with 3 noninsomnia individuals on index enrollment date, sex, and age. To avoid potential confounding effects, we excluded those with sleep apnea (the International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes: 327.2, 780.51, 780.53, and 780.57) and those with previous diagnosis of stroke (ICD-9-CM codes 430–438) and insomnia before the index enrollment date in both case cohort and comparison cohort. To investigate the risk of developing stroke during the follow-up period, each individual in the study and comparison cohorts was tracked for 4 years from their index enrollment date until death or the end of the 4-year follow-up.
Definition of Insomnia
An individual with any of the following insomnia-related ICD-9-CM codes was classified as an insomniac: 780.52 (insomnia, unspecified), 307.41 (transient disorder of initiating or maintaining sleep), and 307.42 (persistent disorder of initiating or maintaining sleep). For individuals meeting the inclusion criteria, we further categorized them into 1 of the 3 insomnia subtypes11 based on their incidence of insomnia episodes during the 4-year follow-up period. Since the index date of enrollment, a period of 180 days was used as the time period for assessment to determine individuals’ status of insomnia (yes versus no) based on their outpatient and inpatient claims at each time period throughout the duration of follow-up. The cutoffs of 180 days was derived empirically from a report by the National Institutes of Health,15 which stated that chronic insomnia could last from 30 days to as long as 6 months. Individuals with diagnosed chronic insomnia within each 180-day period from their index date of enrollment across the 4-year follow-up period were considered persistent insomniacs. Relapse of insomnia was defined as a return of insomnia after being diagnosed free of the disease for >180 days at any assessment time point. Remission of insomnia was defined as a change from a diagnosis of insomnia at one time point to a noninsomnia diagnosis at the subsequent time point and stayed in the noninsomnia status throughout the remaining study period.
Definition of Stroke
The main outcome of interest was an end point of the first hospitalization for acute stroke (ICD-9-CM codes 430–431 for hemorrhagic stroke, 433–434 for ischemic stroke, 435 for transient ischemic attack, or 436 for acute but ill-defined cerebrovascular disease [apoplexy, apoplectic: not other specified, attack, cerebral seizure, and cerebral seizure]).16 Individuals were followed up to the first occurrence in the end point of stroke, death on discharge, withdrawal from the NHI program, or for a full 4-year term after the index date of enrollment.
A proxy measure of socioeconomic status in this study was obtained from the monthly insurable wage in the Longitudinal Health Insurance Database 2000 database. Individuals with a well-defined monthly wage (measured in New Taiwanese dollars (NTD; USD 1≈NTD 30) were classified into 1 of 3 groups: ≥NTD 25 001, NTD 15 841 to 25 000, and <NTD 15 840.17
Age cutoffs were used to categorize individuals into 4 groups to assess the odds of stroke by different stages of life: 18 to 34, 35 to 49, 50 to 64, and ≥65 years.
Differences in demographics and the selected comorbidities between the insomnia and noninsomnia groups were evaluated using χ2 tests. We designated a set of comorbidities, including hypertension (ICD-9 codes 401–405), diabetes mellitus (250), hyperlipidemia (272), depression (296.2, 296.3, 300.4, and 311), anxiety (300.0–300.3 and 300.5–300.9), and atrial fibrillation (427.31) each as a dichotomous variable. The incidence rate was calculated as the number of stroke per total person-years of follow-up. We used log-linear Poisson regression to calculate incidence rate ratios (IRRs) of stroke, hereby comparing the rate in insomniacs with that in noninsomniacs. Kaplan–Meier curves were used to plot and compare the cumulative risk of hospitalization for stroke between the insomnia and noninsomnia groups with the log-rank test. Interim mortality was censored at the date of death. Cox proportional hazards regressions were applied to estimate the effect of insomnia on the risk of stroke. Covariates for the multivariate analysis included age, sex, comorbidities, socioeconomic status, and geographic region. Analyses were performed using SAS version 9.3 (SAS Institute, Cary, NC). The significance was accepted at P<0.05 level.
The present study included 21 438 subjects with insomnia and 64 314 matched subjects without insomnia (Table 1). The mean age of the insomnia population was 52.1±16.0 years and that of the control group was 51.2±16.0 years. Prevalence of each kind of comorbidity was higher in insomniacs than in noninsomniacs (P<0.0001). Table 2 shows the risks of hospitalization for stroke in the insomnia and noninsomnia groups. During the 4-year follow-up period, there were a total of 583 stroke admissions among insomniacs and 962 among noninsomniacs. The incidence rate of stroke was significantly higher (IRR, 1.85; 95% confidence interval [CI], 1.67–2.05) in insomniacs than in noninsomniacs. The distributions of stroke subtypes differ in the insomnia group, with higher IRR in the insomniacs as compared with the noninsomnics: ischemia (IRR, 1.79; 95% CI, 1.56–2.06), transient ischemic attack (IRR, 2.84; 95% CI, 2.23–3.61), hemorrhage (IRR, 1.32; 95% CI, 1.03–1.68), and unspecified stroke (IRR, 2.07; 95% CI, 1.41–3.03). Across all age groups and sex groups, individuals with insomnias had higher incidence rate of stroke than noninsomniacs. Insomniacs-to-noninsomniacs IRR for stroke decreased as age advanced; the largest IRR for stroke was observed in those aged 18 to 34 years (IRR, 8.06) and it continually decreased thereafter. As for comorbidities, a significant difference was only found for diabetes mellitus. Among individuals with diabetes mellitus, the insomniacs had a greater risk of stroke than noninsomniacs (P=0.0442).
Table 3 shows the results of the Cox regression analyses. Compared with the noninsomnia individuals, those with insomnia were at increased risk of stroke by a magnitude of 54% (adjusted hazard ratio [HR], 1.54; 95% CI, 1.38–1.72) after adjusting for all other covariates. When breaking down insomniacs into 3 subtypes, we found that, compared with the noninsomniacs, persistent insomniacs had higher risk of stroke (crude HR, 2.04; 95% CI, 1.78–2.34), followed by those with a relapse of insomnia (crude HR, 1.76; 95% CI, 1.53–2.02) then remission of insomnia (crude HR, 1.55; 95% CI, 1.20–2.01) in unadjusted analysis. Nevertheless, these HRs were attenuated and became similar among the 3 insomnia subtypes after adjusting for other relevant covariates. Older age (adjusted HR, 3.38–19.89; 95% CI, 2.16–30.56), diabetes mellitus (adjusted HR, 1.93; 95% CI, 1.70–2.20), hypertension (adjusted HR, 1.94; 95% CI, 1.73–2.17), atrial fibrillation (adjusted HR, 2.26; 95% CI, 1.52–3.36), and lower socioeconomic status (adjusted HR, 1.52–1.68; 95% CI, 1.22–2.07) were significantly associated with increased risk of stroke. Women were 28% less likely to develop stroke than men (adjusted HR, 0.72; 95% CI, 0.65–0.79).
According to the Figure, the 4 survival curves were significantly different (P value for log-rank test<0.0001). When comparing the 4-year cumulative incidence among groups, the significant differences in survival curves were observed when comparing the persistence (P value for log-rank test<0.0001), relapse (P value for log-rank test<0.0001), remission groups (P value for log-rank test=0.0008) with the control group. As we used healthcare-seeking behavior of the insomniacs for each 180-day time period for 4 years to categorize insomniacs into subgroups, healthcare-seeking behavior may not vary among the 3 insomnia subgroups during the first year. Thus, we further assessed whether there were group differences in cumulative incidence rate of stroke after the first year. There was no significant difference between the remission and persistent groups during the first year; however, persistent insomniacs had a higher 3-year cumulative incidence rate of stroke than those in the remission group (P=0.0241).
Our study shows that the presence of insomnia raises the likelihood of subsequent hospitalization for stroke for 4 years. To our knowledge, this is the first study to indicate that individuals with remission insomnia, relapse insomnia, and persistent insomnia subtypes have greater risks of developing stroke as compared with the control group. When breaking down to types of stroke, the insomniacs:noninsomniacs ratio of incidence of stroke was the highest for transient ischemic attack, followed by unspecified stroke, ischemia, and hemorrhage.
Evidence on influence of insomnia on stroke is still lacking. With regard to the relation between sleep duration and stroke, findings have been inconclusive. Prolonged sleep duration (≥9 hours/night; relative risk, 1.72) was related to a greater risk of stroke and short sleep (≤6 hours/night; relative risk, 1.22) increased risk for stroke in postmenopausal women without clinically overt cardiovascular diseases.1 Although a U-shaped association between sleep duration and stroke has been reported, Westerlund et al18 pointed out that sleep duration was unrelated to risk of stroke. Our finding linking insomnia to greater risk of stroke demonstrates the need to consider patterns of sleep jointly, such as interruption of nocturnal sleep, in addition to sleep quantity. Our results reinforce the need to screen and treat insomnia, which can be early observed, to lower future risk of stroke.
A major novelty of the current study is to separate subtypes of insomniacs based on longitudinal patterns of insomnia and to assess the risk of stroke separately. Two studies that have investigated pattern-oriented insomnia subtypes using multiple time points focused on either the natural history of insomnia11 or the course of insomnia comorbid with cancer.19 Morin et al11 assessed the course of insomnia subtypes for 3 years in a sample of 388 adults and found that insomnia is often an enduring condition, especially among those with more severe insomnia at baseline. Savard et al19 conducted a study in 962 patients undergoing treatment for nonmetastatic cancer and stated that 37.6% of the patients with an insomnia syndrome at baseline remained at the same status for an 18-month period. Both studies revealed that pattern-oriented insomnia subtypes play a role in the progress of insomnia severity; however, it remains unclear whether the insomnia subtype could affect disease risks. Based on our results, it seems that the risks of stroke during the 4-year follow-up increased with the severity of insomnia (risk of stroke: persistent subtype>relapse subtype>remission subtype) in the unadjusted analysis. Moreover, persistent insomniacs had a higher 3-year cumulative incidence rate of stroke than those with the remission subtype. Altogether, our data highlight the clinical relevance of categorizing insomnia subtypes based on incidence. Future research is warranted to investigate whether the associations vary between insomnia subtype and disease risk according to subtype over time.
Potential pathway of insomnia to stroke is not yet clear. Some of the proposed pathophysiological mechanisms linking sleep duration and cardiovascular system may shed the light on how the presence of insomnia could predispose an individual to stroke. Sleep deprivation and insomnia symptoms may alter cardiovascular health through inflammation,20,21 endocrine or metabolic dysregulation,18,22 and increased sympathetic nervous activity.23 Clinical studies suggested that sleep deprivation may affect levels of cytokines known to play a role in regulating systematic inflammation,20 such as elevations in C-reactive protein,20,21 tumor necrosis factor-α,24 interleukin-1,25 interleukin-6,26 and interleukin-17.26 Future research is needed to disentangle the complex of associations between insomnia and the development of stroke. Further investigations should be conducted to explore whether short sleep and prolonged sleep may trigger different biological mechanisms contributing to the cardiometabolic deterioration.
It is worth mentioning that although comparing the incidence rates of stroke between insomniacs and noninsomniacs by age group, IRR was highest in the youngest age group (18–35 years) and decreased as age advanced. The impact of insomnia on stroke was reduced with age may be explained by higher incidences of stroke among elders in general. Nevertheless, the incidence rate of stroke was ≈8× higher in insomniacs than in noninsomniacs among younger adults and it is particularly striking. This observation points toward a need for screening for insomnia to seek early treatment in younger age as a feasible strategy to prevent the development of stroke and stroke-related morbidity.
Strengths and Limitations
There are several limitations in this study. First, sampling bias may exist. It is important to note that the observed associations reflect the healthcare-seeking behavior of individuals with insomnia, stroke, and comorbidity; the real incidence may be underestimated when exclusively using outpatient and inpatient benefit claims for coding diagnosis. There are chances that those in the control group may have insomnia but not seeking medical care. However, if the association between insomnia and stroke still could be observed under this circumstance, the true effect might be even greater when the pure control and case groups could be identified. Second, accuracy of insomnia diagnoses in the database, as coding errors may have occurred in various local clinics and hospitals of different levels. Despite these limitations, the study has several strengths. This nationwide population-based data set provided statistical power to detect the impact of insomnia on stroke among insomnia subtypes after adjusting for confounding variables. Although selection and nonresponse biases could not be ruled out, those biases may be diluted by the large sample size. The 4-year observation period allows us to evaluate causal relationships and assess the potential public health relevance. Finally, information on subtypes of ischemic stroke (eg, Trial of Org 10172 in Acute Stroke Treatment [TOAST] criteria) and established vascular risk factors, such as tobacco usage, work stress, and physical inactivity, were not available in the NHI Research Database. Future research with comprehensive database on stroke classification is needed to elucidate the pathogenic mechanisms of how insomnia may promote the development of stroke.
Our study provides population-based evidence that people with insomnia have higher incidences and risks of hospitalization for stroke as compared with noninsomniacs. The results of this study suggest that intervention to improve insomnia is needed and should be examined whether it could be a strategy to improve cardiovascular health. To reduce the risk of stroke effectively, health education should be promoted to get individuals to be aware of insomnia symptoms at young age, to treat it, and to track its patterns, especially for younger adults. More prospective studies with representative populations and longer follow-up are needed to identify the long-term risks of developing stroke by insomnia subtypes. In addition, the pathophysiological pathways connecting insomnia and stroke need to be investigated.
This study was based in part on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health and managed by National Health Research Institutes in Taiwan.
Sources of Funding
This research was supported by the Chi-Mei Medical Center, Tainan, Taiwan (CMFHR10216; CHN Grant No. 102092).
- Received September 26, 2013.
- Revision received February 4, 2014.
- Accepted February 21, 2014.
- © 2014 American Heart Association, Inc.
- Chen JC,
- Brunner RL,
- Ren H,
- Wassertheil-Smoller S,
- Larson JC,
- Levine DW,
- et al
- Wilt TJ,
- Carlson KF,
- Goldish GD,
- MacDonald R,
- Niewoehner C,
- Rutks I,
- et al
- Savard J,
- Ivers H,
- Villa J,
- Caplette-Gingras A,
- Morin CM
- Ferrie JE,
- Kivimaki M,
- Akbaraly TN,
- Singh-Manoux A,
- Miller MA,
- Gimeno D,
- et al
- Dettoni JL,
- Consolim-Colombo FM,
- Drager LF,
- Rubira MC,
- Souza SB,
- Irigoyen MC,
- et al