Abstract
Sleep problems cause a significant negative impact on mental and physical health [1–3]. Insomnia affects an estimated 10–50% of the elderly community [4–10]. This wide range in the prevalence rate of insomnia and the difficulties in comparing studies are due to the lack of standardization in the classification of insomnia. For example, some studies have defined insomnia as simply ‘sleep dissatisfaction’ [8], [9]. In contrast, other studies have defined insomnia using the presence of specific symptoms, including difficulty falling asleep, difficulty maintaining sleep, or early morning awakening [3],11– [14]. Even when using the same classification of insomnia, prevalence rates have varied between 10% and 37%, and reflect differences in the lengths of each study as well as in the frequency of symptoms required for the definition of insomnia [15].
In order to define insomnia with greater precision by taking into account the severity of the disease, frequency of symptoms and impact on social and occupational functioning, standardized international criteria for insomnia have been published in the International Classification of Sleep Disorders (ICSD-90) [16], the International Classification of Diseases (ICD-10) [17], and the third revised edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IIIR) [18]. According to DSM-IIIR criteria, insomnia required the presence of difficulty falling asleep, difficulty maintaining sleep or non-restorative sleep at least three times per week, lasting at least one month, which results in significant clinical distress in social or occupational functioning. The DSM-IV edition follows most of the DSM-IIIR criteria but removed frequency of symptoms as a part of the definition [19]. Insomnia in the elderly has been extensively studied in Western nations [12], [20]. To date, only one study has been reported in a Chinese population in Hong Kong [21]. Despite high utilization of hypnotic use in Taiwan [22], knoweldge of sleep disturbance in the elderly remains clouded. Our aims are to determine the prevalence of insomnia using relatively strict criteria and to identify modifiable risk factors. The identified subjects will be suitably managed.
Method
Participants
The study population included people aged 65 years or older residing in Taipei's Shi-Pai community from June 1999 to November 2000. Subjects were recruited by door-to-door visits. Of the 4750 individuals eligible (in accordance with the district's resident lists) for the survey, 650 people were not at home, 48 had died and 298 were institutionalized, disabled or were unable to be interviewed; 1709 refused to participate. The final number available for analysis was 2045; the response rate was 54.6%. Shi-Pai community located at the northern part of Taipei and was surrounded by six large hospitals. The elderly population in this area was of middle socio-economic class. While most of them lived with their family of three generations, some with diseases lived in nursing home.
General assessment
All participants completed a comprehensive questionnaire administered by research assistants with medical backgrounds. The initial section surveyed the individual's demographic characteristics (age, gender, marital status, education level), living arrangements, selfreported history of medical conditions (e.g. hypertension, diabetes mellitus, cardiovascular disease, stroke, gout, pulmonary disease and mental disorders) and status of cigarette smoking and alcohol consumption (current, former, or abstinent).
Assessment of cognitive and mood status
As cognitive deterioration and depressed mood may impact quality of sleep, participants underwent cognitive assessment as well as screening for mood disorders. Cognitive impairment was defined as: (i) individuals with more than 6 years of education scoring less than 24/30 on the Mini-Mental State Examination (MMSE); or (ii) individuals with 6 years or less of education scoring less than 15/30 on the MMSE [23]. Screening for mood disorders was done using the Geriatric Depression Scale – (Short Form) (GDS-SF), in which subjects scoring more than five points out of 15 questions are considered to have depression [24].
Assessment of insomnia
The third section of the questionnaire evaluated the individual's perception of sleep. Disruptions in sleep patterns were defined in the following way by using criteria of DSM-IV [19] and Pittsburgh Sleep Quality Index [25]:
Cluster 1
(a) Difficulty falling asleep was defined as failing to fall asleep after at least 30 min after going to bed or;
(b) Difficulty maintaining sleep (interruption of sleep at least three times during the night) or early morning awakening (wake up two hours earlier than usual).
Each symptom must occur at least three times a week to qualify for Cluster 1.
Cluster 2
(a) Poor, subjective sleep quality (scoring 3 or 4 among 4 severity scales) or;
(b) Moderate degree or above of daytime dysfunction in the aspects of mood, work efficiency and daily activities (scoring 3 or 4 among 4 severity scales).
To meet the criteria for 1-month insomnia, subjects must have experienced both of the symptom clusters over the past month. People suffering from jet-lag secondary to frequent flying or work shifts were excluded from the study sample.
Subjects were also questioned about several other components of sleep: (i) hypnotic usage, quantified as the number of nights over the past month a hypnotic medication was used; (ii) nocturnal micturation, sleep that is disturbed by nightly urination at least three times per week; (iii) excessive snoring loud enough to interfere with a partner's sleep; and (iv) excessive daytime sleepiness resulting in the subject's desire to sleep during work and meals.
Statistical analysis
The prevalence rate of one-month insomnia was calculated by dividing the number of subjects who met the criteria for one-month insomnia by the total number of surveyed people. Calculation of the lifetime prevalence rate was similar.
All statistical analyses were conducted using SPSS V.8.0 statistical software. The χ2 test was utilized in calculating the univariate association of one-month insomnia as a dichotomized variable in terms of demographic variables, physical and mental health status, social activity, cigarette smoking, alcohol consumption, nocturnal micturation, snoring, excessive daytime sleepiness, use of sleeping pills, and GDS-15 scores.
In order to seek risk factors, odds ratios were calculated using multivariate logistic regression models, with one-month insomnia being the dependent variable for all independent factors. Three separate analyses for both the univariate and multivariate logistic regressions were performed for the entire study sample, male and female.
Results
Demographic data
The study sample consisted of 1156 men (54.9%) and 889 women (43.5%) (Table 1). More men (66.5%) were currently married, which reflected the higher prevalence of elderly widowed women. Table 1 also demonstrates that men had a higher level of education than women and 94% of the study population lived with family members or others.
Demographic characteristics by gender
Prevalence of insomnia and regular hypnotic use by age and gender
Six percent (n = 123) of the elderly population met the criteria for one-month insomnia and 17.7% (n = 361) met the criteria for lifetime insomnia. Although the clinical course of the disorder fluctuates, 80% (n = 99) of one-month insomniacs also met the criteria for lifetime insomnia. Thus, few one-month insomniacs were of recent onset. Increasing age did not correlate with a higher prevalence of either one-month or lifetime insomnia (Table 2). The female elderly had greater insomnia rates than men. In addition, 8.4% of the elderly population were regular hypnotics users over the past month; in particular, female users outnumbered male users by a factor of 1.6 (p < 0.01). Among these hypnotic users, 21% still experienced symptoms of insomnia despite the use of hypnotics, while 79% were not symptomatic.
Prevalence of insomnia, hypnotic use, and combined rate of both by age and gender
Gender, prevalence of sleep disturbance symptoms and their relationship with hypnotics
Certain symptoms of insomnia are more prevalent in women than in men (Table 3). These include difficulty falling asleep, difficulty maintaining sleep or early morning wakening, poor sleep quality and daytime dysfunction (1.7–2.1-fold, p < 0.001). In contrast, men were more likely than women to report snoring (2.1-fold, p < 0.001). The only symptom of sleep disturbance that was equally prevalent in men and women was excessive daytime sleepiness (p = NS).
Prevalence of sleep disturbances by gender
Subjectively poor sleep quality was the most prevalent insomnia complaint in the elderly, followed by difficulty falling asleep, difficulty maintaining sleep or early morning awakening, and emotional distress, work and social dysfunction (Table 3). When compared to the rate of difficulty falling asleep (47%) with difficulty maintaining sleep or early morning awakening (25%), the former sleep symptom was greater than the latter, suggesting that elderly people experienced more initial sleep disturbance. Among the insomniacs with poor sleep quality alone (n = 88), only 25% used sleep medications in contrast to 43% of the subjects with other sleep symptoms who took hypnotics. Significant differences in the rate of hypnotic use between these two groups (χ2 = 7.8, p < 0.01) illustrates that the majority of subjects reporting poor sleep quality alone did not seek medical attention.
Are symptoms of excessive daytime sleepiness caused by the previous night-time hypnotic use? First, we selected a group of subjects (n = 248) who reported excessive daytime sleepiness but did not have insomnia. In this group, only 11.7% took hypnotics, strongly arguing the association of occurrence of excessive daytime sleepiness symptoms with use of hypnotic medications.
Prevalence of cognitive impairment (MMSE) and depression (GDS-15)
Two-thirds of the study sample (n = 1358) took the MMSE and only 2.7% (n = 36) and 0.3% (n = 4) had an MMSE score less than 24 and 15, respectively. All the subjects with the educational level above 6 years had a score greater than 26. This suggests that the majority of the study population did not suffer from significant cognitive impairment. In addition, 181 subjects (8.9%) were found to have depression.
Relationship between insomnia, depression and use of hypnotics
Insomniacs had a three-fold higher prevalence of depression than non-insomniacs (25.2% vs. 7.8%, p < 0.001). Conversely, subjects who were depressed were also more likely to have insomnia than nondepressed subjects (17.1% vs. 4.9%, p < 0.001). Among 123 subjects with insomnia, the prevalence of depression was the same in those who used hypnotics (n = 36) compared to those who did not use hypnotics (n = 87) (25% vs. 25%, respectively, p = NS), indicating that occurrence of depression is related to insomnia regardless of hypnotic use. However, when multivariate regression analysis was performed using ‘use of hypnotics’ as an outcome measure and insomnia, depression and interaction of insomnia and depression as predicting factors, it was found that the odds ratio for insomnia and depression versus their counterparts were 4.5-fold (95% CI = 2.9–6.9) and 3.2-fold (95% CI = 2.9–6.9) and interaction of both had an even higher odds ratio: 4.7-fold (95% CI = 2.1–10.3). These findings suggest that relation of depression with insomnia is tight, but with hypnotics, is unclear.
Risk factors for insomnia
Using univariate analysis of demographic data, habits, and health status with one-month insomnia, certain risk factors were strongly predictive of insomnia (Table 4). Some of these associated factors regarless of gender were: nocturnal micturation, use of hypnotics, selfrated poor health, body pain and depression symptoms (p < 0.001). However, several risk factors were gender-specific. Single marital status, a history of stroke (p < 0.03), excessive daytime sleepiness and history of mental disorders (p < 0.005) were strong predictors of insomnia in men, whereas cardiovascular diseases (p < 0.005) and less than 6 years of education (p < 0.1) were predictors of insomnia in women.
Univariate analysis of risk factors in insomnia
A multivariate logistic regression model was used to calculate the independent association of all the univariate factors with insomnia (Table 5). Significant risk factors associated with insomnia in both genders were nocturnal micturation (OR = 20.6), lack of education, body pain, depressive symptoms and regular use of hypnotics (OR = 1.7–3.2). Among these risk factors, nocturnal micturation and regular use of hypnotics increased the risk of insomnia. In particular, nocturnal micturation and the regular use of hypnotics doubled the risk of insomnia in men compared to women (OR = 39.1 vs. 14.5, and 5.8 vs. 2.6, respectively). In addition, gender-specific risk factors were identified. Single status, excessive daytime sleepiness and history of mental disorders or pulmonary diseases were found to be highly associated with insomnia in men, whereas low education, depression and pain symptoms were risk factors for insomnia in women.
Multivariate regression analysis of risk factor in insomnia
Female gender has been a well-established risk factor for sleep disturbances. A similar result was elicited in our univariate analysis. However, after low education level, regular use of hypnotics and depression were controlled, female gender was no longer an independent risk factor (95% CI = 0.96–2.33). These results imply that the female insomniacs of this elderly population have low education, depression and use of hypnotics.
Discussion
The present study is the first large-scale communitybased survey of sleep disturbances in the elderly in Taiwan. This elderly urban population had a low prevalence rate of both one-month (6%). Regular hypnotic use over the previous month was 8.4%. Age over 65 years did not correlate with a higher rate of insomnia, but women were more affected by insomnia, sleep disturbance symptoms in general and were more likely than men to use hypnotics. Excessive daytime sleepiness was much more common in male insomniacs and this was independent of hypnotic use. In addition, the present study identified a group of the subjects who reported poor sleep quality, but did not meet the criteria for insomnia, requiring close longitudinal follow-up. This study also elicited risk factors in general and specifically the role of gender for the occurrence of insomnia.
Prior studies have reported varied prevalence rates. The variation appears to be the result of the differences in the criteria used to define insomnia. Studies that use a stricter definition of insomnia, that is, using symptoms and frequency, tended to report lower prevalence rates ranging from 12% to 15.8% [5], [26], [27], Studies reporting higher rates (29.0–43.2%) [3], [28], tend to use broader definitions, which required only one sleep symptom (difficulty falling asleep, difficulty maintaining sleep or early morning wakening) without considering the frequency. When frequency of sleep disturbance was added as part of the criteria, the insomnia rate was reduced. One prior study of an elderly population in Hong Kong demonstrated that the insomnia rate was 38.2%, which reduced to 13.7% for frequent insomnia [21].
The present study revealed a low insomnia rate (6%) over the past month relative to other studies but was modestly higher than the rate (4%) reported by Hajak [29]. That study used stricter criteria; two or more symptoms (including difficulty falling asleep, difficulty maintaining sleep or insufficient hours of sleep) were required for the definition of insomnia. In contrast, only one sleep symptom was required for our study.
We attribute the low prevalence rate of insomnia found in the present study to several factors. First, strict criteria, specifically cluster 1 and cluster 2 symptom groups, were used to define insomnia. Furthermore, certain specific symptoms such as difficulty falling asleep and difficulty maintaining sleep/early morning wakening were less prevalent in this population than in Western studies [3],[11–13], thus lowering the overall prevalence of insomnia.
Second, the study design likely excluded groups such as the institutionalized population and those individuals with poor health or cognitive problems. Prior studies have reported higher rates of insomnia among institutionalized individuals. A study of 205 institutionalized elderly individuals in Taipei revealed that 69.3% experienced poor sleep quality [30]. This result was consistent with the 1990 National Institute of Helath Consensus Development Conference, which estimated that sleep disturbances affect more than 50% of persons over age 65 living at home and approximately 75% of those living in long-term care facilities [31].
Finally, a high rate of hypnotic use in the population may have contributed to a low insomnia rate. Our study showed that 8.4% of subjects regularly used hypnotics and of these 172 subjects, 79% did not suffer from insomnia. This suggests that the insomnia rate in the present study may be underestimated.
Sleep problems are generally believed to increase with increasing age, but epidemiological evidence to support this contention is mixed. Some studies demonstrated that the prevalence of insomnia increases after the age of 20 years [4], [31], [32]. They found differences in the prevalence of insomnia between a younger population (20–40 years) and an elderly population (≥ 65 years). However, when specific age groups within an elderly population (65 to over 85 years) were studied, the prevalence of insomnia did not increase significantly with age [5], [15], [26]. Our results are consistent with these findings. In contrast, Roberts [33] found a tendency for sleep complaints to increase with age. However, after controlling for physical disease and mental disorders, increasing age was no longer a risk factor for insomnia [33], [34].
Multiple studies [5], [12], [26], [33], [34] have found that women are at greater risk of developing insomnia, even after controlling for other risk factors [33], [35]. Our data however, do not confirm female gender alone as a risk factor for insomnia after controlling for education levels, depression, body pain and use of hypnotics. Illiteracy as a risk factor for insomnia (OR = 1.74) is also supported by Schechtman's report [24].
Another interesting issue is the relative predominance of difficulty falling asleep or difficulty maintaining sleep/early morning wakening in older insomniacs. The majority of investigations from Western [11], [12], [14] and Japanese populations [27], [28] have suggested that the rate of difficulty falling asleep in the elderly is less than that of difficulty maintaining sleep/EMA. In contrast, our study, consistent with the studies of Althius and Ganguli [3], [13], demonstrated that the prevalence initial sleep disturbance is higher that slepp disturbance occurring later at night. Poor sleep quality was the most frequent complaint; however, subjects who reported only this symptom were less likely to use hypnotics and had a lower rate of lifetime insomnia than those subjects who complained of multiple symptoms. This finding suggests that patients with complaints of poor sleep quality alone may only need education regarding sleep hygiene and behavioural modification rather than hypnotics and may avoid unnecessary benzodiazepine use.
In addition to gender and education, nocturnal micturation is a risk factor for insomnia (OR = 20.6). The rate of nocturnal micturation is approximately 47 times higher in male insomniacs (OR = 39.1) and 16 times higher in female insomniacs (OR = 14.5) compared to non-insomniacs of each gender. The higher prevalence of nocturnal micturation among men may be secondary to prostate enlargement in older men.
Our data, consistent with previous reports [12], [14], [35–37], also shows that the presence of pulmonary diseases is a key risk factor for insomnia in elderly men (OR = 2.7). Second, excessive daytime sleepiness is closely associated with insomnia. Third, excessive daytime sleepiness, heavy snoring and pulmonary disease are reported to often coexist with obstructive sleep apnoea syndrome, an important cause of insomnia in the elderly[38]. Thus, there might be some elderly men in our sample with undiagnosed sleep apnoea syndrome. On the other hand, our data demonstrates that excessive daytime sleepiness was not caused by hypnotic use but by insomnia. To differentiate whether excessive daytime sleepiness is induced by insomnia or by sleep apnoea syndrome, further epidemiological data and polysomnographic studies are required.
Regular use of hypnotics is associated with insomnia (OR = 3.2). Many community studies have shown a high prevalence of hypnotic usage (10–27%) in elderly populations [8],[36–39] and a sudden increase of use in elderly men [6]. In agreement with these studies, we found that 8.4% of the elderly subjects used hypnotics regularly within a one-month period and 75% of these subjects used hypnotics longer than 6 months (data not shown). Regression analysis revealed that insomnia persisted in twice as many male (OR = 5.78) than female (OR = 2.57) hypnotic users (Table 5) despite the higher rate of hypnotic use in women (10.6%) than men (6.7%) (Table 2). This finding may suggest that elderly men rather than women are more likely to suffer from sleep disturbances when sleep-inducing medicines were given frequently or that long-term usage of hypnotics leads to the eventual loss of therapeutic effects and the development of drug tolerance and dependence [4], [12], [14], [34], [36]. The reason for vulnerability to addiction in elderly men remains unclear. Overall, the use of benzodiazepines is very popular in Taiwan [22] and has become a crucial epidemiological issue.
Consistent with other studies, a close relationship of insomnia with depression [12], [27], [33], [34], [40], [41] and pain syndrome [14], [26], [34] were also observed in this elderly female population.
Depression has a particularly complex relationship with insomnia. To investigate the relationship between depression and insomnia, three prospective studies [5], [42], [43] reported that sleep disturbances often precede depression in two different situations. First, sleep disturbance may be an epiphenomenona of depression. The improvement of insomnia would therefore not affect the clinical course of depression. Second, sleep disturbance may be an early symptom of major depression and thus the treatment of insomnia might preclude the development of depression [5]. The present study was a cross-sectional investigation and therefore did not address the relationship between insomnia and depression. However, it is clear from our study that if insomnia coexists with depression, both symptoms may be highly associated with constant use of hypnotics.
The strength of this study is that the data are reliable in terms of what the researcher can control; it was collated from one-on-one interviews rather than telephone questioning in a community survey with relatively low-educated elderly. As Hajak contends, the use of strict diagnostic criteria better ensures a study sample that ‘undoubtedly (requires) treatment’ [29]; we too have set our criteria high in this investigation. Nonetheless, there are some caveats. First, use of cross-sectional data prohibits any assessment of the cause-and-effect relationship between sleep disturbances and physical and psychiatric disorders. Second, all information was selfreported and might have been influenced by recall bias. Third, the response rate was around 50% and thus the study may be biased against those who were frail, ill or institutionalized.
Footnotes
Acknowledgements
This study was supported by the grant of Shi-Pai Community Medical Service program (89-404-1) from Taipei Veterans General Hospital. We thank John Chang and Irene Su for their comments.