Prevalence of insomnia in a survey of 12 778 adults in France
Dr D. Leger, Centre du Sommeil, Hotel Dieu de Paris, 1 pl Parvis Notre-Dame, 75181 Paris Cedex 04, France.
This study was an epidemiological questionnaire survey of a representative sample of the French population that included 12 778 individuals and in which adapted DSM-IV criteria for the definition of insomnia were used. Our goals were not only to assess the prevalence of ‘insomnia’ using these criteria, but also to compare the results obtained with those of prior studies using different definitions of ‘insomnia’. The aim of this study was also to identify where areas of agreement and disaggreement existed, as we believe that it is important to emphasize these points because DSM-IV recommendations are supposedly reflected in clinical practice. Seventy-three per cent of the individuals surveyed complained of a nocturnal sleep problem, but only 29% reported at least one sleep problem three times per week for a month, and 19% (2428 subjects) had at least one sleep problem three times per week for a month and complained of daytime consequences (DSM-IV criteria). Only 9% had two or more nocturnal sleep problems with daytime consequences and were classified as ‘severe insomniacs’. Our study indicates that if DSM-IV criteria are used, the diagnosis of ‘insomnia’ is lower than in other epidemiological studies. The DSM criteria have an advantage in that they emphasize the daytime consequences of nocturnal sleep disturbances, which seem to be responsible for the most important socio-economic costs of the problem.
In 1979 a Gallup poll found that only 5% of the surveyed population never had a sleep problem ( Gallup Organization 1979). Despite this high level of complaints concerning sleep, the epidemiology of ‘insomnia’ is still questioned. There were several pitfalls in many of the studies performed over the past two decades, mostly in terms of sample size and population ( Karacan et al. 1976 ; Bixler et al. 1979 ; Lugaresi et al. 1987 ; Ferley et al. 1988 ), but also, when these problems were avoided, in terms of the definition of the term ‘insomnia’ ( Karacan et al. 1983 ; Partinen et al. 1983 ; Mellinger et al. 1985 ; National Sleep Foundation 1991; Quera-Salva et al. 1991 ). The DSM-III-R ( American Psychiatric Association 1987) and DSM-IV ( American Psychiatric Association 1993) definitions of ‘insomnia’ are usually used for clinical purposes, and are currently considered to be the bases of clinical treatment. However, to our knowledge, only two surveys, performed on subjects drawn from general medical practices, adopted DSM definitions for epidemiological surveys ( Hohagen et al. 1994 ; Ohayon 1996).
Our study is an epidemiological survey of a representative sample of the French population that includes 12 778 individuals and in which the DSM-IV criteria for the definition of insomnia were used. Our goals were not only to assess the prevalence of ‘insomnia’ using these criteria, but also to compare the results obtained with those of prior studies that had used different definitions of ‘insomnia’. The aim of this study was also to identify where areas of agreement and disagreement existed, because we believe that it is important to emphasize these points as DSM-IV recommendations supposedly should be reflected in clinical practice.
The survey was performed in collaboration with a national survey institute, i.e. SOFRES Inc. Fifteen thousand subjects from 8000 households regularly surveyed by this institute, and representative of the French general population for sex, age, habitat and socio-economic status, were contacted and sent a questionaire. All members of the interviewed household aged 18 y and older were asked to complete the questionaire. Completed questionaires were returned by 12 778 individuals (85.2%; 6006 men and 6772 women) from 6461 households. No incentives were given to the surveyed subjects.
These subjects were verified as being representative of the general population of France. Every one of the 15 000 subjects interviewed can be identified by an analysis database system. This system can calculate whether the responding group is also representative of the general population, based on the last census. If not, the data may be weighted based on the deviation between the group and the general population (e.g. if the general population includes 50% men and 50% women and the responding group includes 25% men and 75% women, the system will give a × 2 ponderation coefficient to the responding men and a × 2/3 ponderation coefficient to the responding women). This system is called ‘the method of quotas based on the demographic data provided by the Institut National de Statistiques et Evaluation Economique (INSEE).
Statistical methodology and analysis
All results are presented with weighted data and 95% confidence intervals. Straight and crossed tabulations were corrected for the total sample and for each investigated subgroup taking into account the demographic criteria of the sample vs. those of the general population based on the last census.
Based on the analysed data, Student’s t-test, Chi-squared tests and odds ratio were performed. Multivariable analysis with analysis of variance were applied secondarily to the analysed data, mainly to verify whether the relation between insomnia and a sociodemographic variable could be explained by the relations that maintained this variable with other demographic variables, such as age and sex. A similar analysis was performed for severe insomnia.
Written informed and ethically explained consent was obtained from each participant by SOFRES, a national polling institute, and the data were immediatly transformed into anonymous files.
The files derived from the questionnaires included: age, sex, professional activity, marital status, housing conditions, and overall location and other demographic information. The questionnaire also included the following four questions on insomnia enquiring whether, in the past 4 weeks, subjects had experienced: (i) difficulty initiating sleep; (ii) nocturnal awakenings, with difficulty returning to sleep; (ii) involuntary too short a sleep period due to early awakening (6 h or less); and (iv) nonrestorative sleep with a feeling of tiredness on awakening and negative consequences in terms of daytime alertness, fatigue and irritability, as a result of these problems. Subjects were asked to respond according to the following scale: (a) yes, at least three times each week; (b) yes, but fewer than three times per week; (c) no, never.
These four questions, with a specific time limit to consider were adapted from the DSM-IV criteria for insomnia with the addition of duration derived from the criteria used in DSM-III-R, which are: (i) complaint of difficulty initiating sleep (more than 20 min to fall asleep), maintaining sleep (a minimum of two awakenings per night with difficulty getting back to sleep), or the complaint of nonrestorative sleep (i.e. sleep of poor quality with a feeling of sleep privation); (ii) the occurrence of this problem at least three times a week over a period of at least 1 month; and (iii) resultant impairment of the subject’s daytime functioning.
Before the beginning of the survey, based on reports from Schramm et al. (1993 ) and Hohagen et al. (1994 ), which indicated that patients complaining of two or three of the above symptoms were more likely to have a stable diagnosis of ‘chronic insomnia’ over a 4-month period, we defined ‘severe insomnia’. To be scored as ‘severe’, subjects must complain of at least two of the three above problems, three or more times per week during the past month, and must indicate the presence of daytime consequences of the night-time problems.
Poor sleep and insomnia
Of the total of 12 778 persons, 57% complained of difficulties initiating sleep and 21% complained of difficulties at least three times a week for the previous month. Fifty-three per cent complained of night awakenings, with 16% reporting it at least three times a week for the past month. Forty-one per cent complained of a nonrestorative sleep with 11% reporting it at least three times a week for the past month.
When all responses were added, 73% of the entire sample (i.e. 12 778 subjects considered as 100%) indicated the presence of a nocturnal sleep problem during the preceding month, but only 29% of the sample had a sleep problem at least three nights per week (with 16% reporting one only sleep problem and 13% two to three sleep problems).
Finally, approximately 19% of the total group (n=2383) had at least one sleep problem at least three times per week for a month and complained of daytime conse- quences (DSM-IV criteria for insomnia) and 9% had two or three sleep problems with impaired daytime alertness. We classified these last subjects as ‘severe insomniacs’ ( 1Table 1).
Poor sleep or insomnia among DSM-IV criteria
Impact of demographic factors on complaints
Gender and age
2Table 2 outlines the results obtained as a function of age and gender.
Insomnia by age group
In the total group, reports of sleep complaints were higher in women than in men. Seventy-eight per cent of women reported at least one sleep complaint compared with 68.1% of men (P < 10–4). With regard to insomnia, the rate was also 63% higher in women than in men (22.8% vs. 14%; P < 10–4).
The overall rate did not increase significantly with age after the age of 35. However, it increased significantly in the 25–34-year-old group, independent of gender (P < 10–4).
When we consider ‘severe insomnia’ as defined previously, the rate was almost twice as high for women as for men (12% vs. 6.3%; P < 10–4), the rate increased significantly with age in both sexes (P < 10–4), and more for women than for men (P < 10–4).
The prevalence of insomnia among socio-professional categories is presented in 3Table 3. A multivariable analysis considering profession, marital status, age and sex indicated that upper level executives and liberal professions, with a predominance of younger men, had a lower rate of insomnia (P < 10–4). Similarly, farmers, who were predominantly men, showed a trend towards a lower rate of insomnia (P < 0.02).
Professional activity and insomnia
The highest rates were seen for retired people (18.3%; P < 10–4) and white-collar workers (20.8%) who were statistically affected more by insomnia than other categories, except the unemployed (0.0004 < P < 0.03). However, these significant differences were no longer present when age and gender were taken into consideration.
4Table 4 indicates that ‘singles’ had a significantly lower rate of ‘insomnia’ than any other group (P < 10–4). This group included a higher proportion of women that were significantly younger (P < 10–4) than the overall group. Widows, which included a higher proportion of older women, complained more about sleep problems, but this was a nonsignificant trend. The most common sleep complaint was ‘difficulty initiating sleep’, which was mentioned by 34% of widows, 24% of divorcees, 21% of ‘married’ or ‘living together’ and 16% of ‘single’ individuals. When age and gender were included in the analysis, there was no difference based on marital status.
Marital status and insomnia
There were no significant regional or urban vs. rural differences (P=0.255, P=0.409, NS).
Epidemiological studies on insomnia have often made no distinction between people with a single complaint of sleep disorders and people with clinical insomnia. However, only one insomniac in four has ever complained about it to their general practitioner during a visit made for another problem, and only one in 20 attends specifically to discuss the problem of insomnia ( National Sleep Foundation 1991). Only a few patients with insomnia receive treatment for it ( Mellinger and Balter 1985; National Sleep Foundation 1991). This gap between complaining occasionally of a sleep disorder and being an insomniac has to be better understood if we want to know more accurately why insomnia seems to be so prevalent and what impact it has on society.
Comparison with previous studies
5Table 5 shows summaries from several large surveys that calculated the prevalence of insomnia, and allows us to compare some of the previous findings with our study. As can be seen, there is a wide range of estimated prevalence. This arises from the lack of comparable methodology used in these studies.
Previous epidemiological studies on insomnia
First, sample sizes vary considerably. The earliest studies were made in small community groups ( Karacan et al. 1976 ; Bixler et al. 1979 ; Lugaresi et al. 1987 ; Ferley et al. 1988 ). Epidemiologists interested in sleep disorders then tried to find larger samples more representative of the general population of several countries ( Karacan et al. 1976 ; National Sleep Foundation 1991; Quera-Salva et al. 1991 ; Schramm et al. 1993 ; Ohayon 1996; Ohayon et al. 1997 ) or else studied an extensive group of twins ( Partinen et al. 1983 ). Our study is the most extensive in a general population.
Definition of insomnia
The main difference between studies is the definition of ‘insomnia’. Since the study of sleep disorders is a rather new discipline, the definition of insomnia has changed considerably and has been improved over the years. At the beginning, insomnia was assessed though self-reported complaints or symptoms. Karacan et al. (1976 ) used six items of a 317-item general questionnaire about health disorders in Florida. The questions encompassed complaints of initiating or maintaining sleep, the use of sleeping pills and the occurrence of nightmares. Bixler et al. (1979 ) asked about the presence or absence, now and in the past, of several sleep complaints from trouble falling asleep to frightening dreams. Lugaresi et al. (1987 ) listed as insomniacs, subjects who answered ‘rarely’ or ‘never’ to the question: ‘Do you sleep well without using sleeping pills?’Quera-Salva et al. (1991 ) considered insomnia to be a treated symptom and described insomnia with regard to the use of hypnotic drugs. Partinen et al. (1983 ) labelled as ‘poor sleepers’ those who regarded their sleep as ‘fairly poor’ or ‘poor’. In many of these first prevalence studies the period of the study was either not specified ( Partinen et al. 1983 ; Lugaresi et al. 1987 ) or was vague, i.e. currently or in the past ( Karacan et al. 1976 , 1983; Bixler et al. 1979 ; Quera-Salva et al. 1991 ). Insomnia was defined as a sleep disorder rather recently with the release of classifications for sleep disorders (DSM-III-R, ICSD, DSM-IV; American Psychiatric Association 1987; American Psychiatric Association 1993; International Classification of Sleep Disorders 1990) and these criteria have been used in clinical practice ( Buysse et al. 1994 ). Hohagen et al. (1994 ) carried out the first prevalence study on insomnia using the DSM-III-R definition. Ohayon and colleagues ( Ohayon 1996; Ohayon et al. 1997 ) used these classifications to validate the definition of insomnia in epidemiological telephone studies made using an expert system (the Sleep EVAL knowledge-based system). We decided to use the DSM-IV definition, which is modified slightly compared with that used in DSM-III-R. It eliminates the minimum of at least three nights of disturbed sleep per week, but has kept the minimum of 1-month duration of complaint and the inclusion of a daytime impact of the sleep disturbance. The 1-month duration may be considered as too short to assess chronic insomnia, however, it was the duration selected by the clinical DSM classifications. Moreover 1 month is also the maximum duration admitted for the prescription of hypnotics, which was decided by French social security regulation. After 1 month the patient has to visit his doctor again to have his prescription renewed. Our data support the integration of frequency of complaint per week, at least to obtain an idea of ‘severity’, as it has important clinical correlations. We added one other temporal element based on clinical information, an ‘involuntary too short sleep period’ was defined as sleep of 6 h or less. Other reasons for using DSM-IV criteria for epidemiological studies of insomnia include: DSM-IV criteria are used in clinical settings ( International Classification of Sleep Disorders 1990; Schramm et al. 1993 ; American Psychiatric Association 1993; Buysse et al. 1994 ; Hohagen et al. 1994 ; Ohayon 1996; Ohayon et al. 1997 ) and when using these criteria, insomnia has a temporal stability in longitudinal studies ( Schramm et al. 1993 ; Hohagen et al. 1994 ).
The DSM-IV describes insomnia as ‘serious enough to induce severe fatigue or signs attributable to insomnia and remarked by nears such as irritability or disability in daytime functioning’. The impact of insomnia on daytime sleepiness is, however, generally not taken into account in epidemiological studies. In the Lugaresi study ( Lugaresi et al. 1987 ), the participants were asked about daytime somnolence, which was found to be more prevalent among insomniacs (15%) than among good sleepers (8.3%). In the 1991 Gallup poll ( National Sleep Foundation 1991) 72% of the insomniac group reported being drowsy or tired in the morning. Ohayon and colleagues ( Ohayon 1996; Ohayon et al. 1997 ) showed that daytime consequences were discriminative to separate satisfied and dissatisfied insomniacs. In our study the impact of including ‘daytime consequences’ in the definition of insomnia was also rather discriminative. Only 65% of those who have had at least one sleep problem three times a week for 1 month had daytime consequences. This criterion makes it possible to distinguish between insomniacs and short sleepers without impairment on daytime function. However, the questionnaire only asked whether symptoms of sleepiness (reduced daytime alertness, fatigue or irritability) were present and no further enquiry was made to determine whether the symptoms were mild, moderate or severe. Potentially, this could slightly inflate the prevalence of insomnia.
We did not investigate whether daytime sleepiness was due to the effects of hypnotics or other CNS drugs. In two previous studies in the French population, Quera-Salva et al. (1991 ) that found 10% of insomniacs were taking hypnotics, and Ohayon (1996) found that 9.9% were taking anxiolytics and hypnotics. Daytime sleepiness may, therefore, be related to drug intake for one insomniac in 10. It has also been widely demonstrated that insomnia is strongly linked to some psychiatric disorders such as anxiety and depression ( Bixler et al. 1979 ; Mendelson et al. 1984 ; Mellinger and Balter 1985; Ford and Kamerow 1989; Quera-Salva et al. 1991 ; Hohagen et al. 1994 ; Ohayon 1996). Patients with such disorders may take treatments with daytime consequences. It may, therefore, be hypothesized that insomniacs with daytime consequences could have more psychiatric disorders than people with sleep disorders and no consequence on alertness, however, we do not have the data to confirm this hypothesis.
Clinically, however, the impact of insomnia on daytime functioning is the subject of controversy. The consequences of transient insomnia on cognitive function are usually accepted ( Partinen et al. 1983 ; Mendelson et al. 1984 ; National Sleep Foundation 1991; Quera-Salva et al. 1991 ; Ohayon 1996; Ohayon et al. 1997 ; Roehrs et al. 1998 ), yet it is difficult to establish a direct correlation between chronic insomnia and daytime sleepiness ( Partinen et al. 1983 ; Mendelson et al. 1984 ; Schneider-Helmert 1987; National Sleep Foundation 1991; Quera-Salva et al. 1991 ; Shapiro and Dement 1993; Bonnet and Arand 1996; Ohayon 1996; Ohayon et al. 1997 ; Roehrs et al. 1998 ).
Variations between countries
Our study also has to take into account possible variability in of the prevalence of insomnia between countries. A World Health Organization international survey about mental illness (not including insomnia) in general health care carefully demonstrated that the prevalence of current psychological disorders was very different from one place to another ( Üstun and Sartorius 1995). The explanations for the observed differences in the prevalence included factors such as: concepts of illness in each country; illness behaviour of patients, which is affected by the medical system of each country; physical ill-health; and level of education. Hohagen and colleagues used the same DSM-III-R definition of insomnia as us and found a much higher prevalence of 31% in Mannheim, Germany ( Hohagen et al. 1994 ). Moreover, the prevalence of sleep disturbances may also appear different in close European countries or in two cities of the same country. In a study made in four cities in Iceland, Sweden and Belgium the prevalence of disturbances for initiating sleep varied from 6% Göteborg to 9% Uppsala ( Janson et al. 1995 ). There was no definition of insomnia in that study. In our study, however, there were no significant regional or urban vs. rural differences.
Age, sex and other factors
As in previous studies, women reported suffering from insomnia more than men ( Karacan et al. 1976 ; Bixler et al. 1979 ; Gallup Organization 1979; Lugaresi et al. 1987 ; Ferley et al. 1988 ; Weyerer and Dilling 1991; Hohagen et al. 1994 ) independent of age category. The difference between men and women is more obvious when ‘severe insomnia’ is considered. The more restrictive the criteria of insomnia, the more dramatic the difference between the sexes.
We found an increase in insomnia for both sexes in 25–34-year-olds. With one exception ( Karacan et al. 1983 ), this asymmetric increase in the rate of insomnia was not found in previous studies ( Bixler et al. 1979 ; Mellinger and Balter 1985; Lugaresi et al. 1987 ; Ferley et al. 1988 ). Insomnia was thought to increase progressively with a little peak in the fourth decade ( Hohagen et al. 1994 ). In our group, the prevalence of insomnia did increase gradually beyond the age of 35 for women, however, it did not for men. Men presented the highest score for insomnia in the 35–49-year-old group (16.3%). We did not ask participants the cause of their insomnia, however, we may postulate that the increase of insomnia was due to professional duties. The 35–50 age group is usually considered as one of the worse in terms of work load and it may have consequences on sleep, i.e. Lavie demonstrated a clear link between poor job satisfaction and insomnia ( Lavie 1981). In all groups of severe insomniacs, the prevalence increased gradually with age, as in other studies.
Multivariable analysis of our results demonstrated that the differences observed concerning professional, marital and matrimonial status were mostly related to age and sex, this is in opposition to other studies with no multivariable analysis ( Ferley et al. 1988 ; Partinen et al. 1983 ; Quera-Salva et al. 1991 ).
Establishing appropriate criteria for the definition of severe insomnia is a central point of our discussion. Using our criteria for ‘severe’ insomnia, we graded the severity of insomnia based upon the assessment made by patients and mentioned in the reports of the previous surveys. The 1991 Gallup survey ( National Sleep Foundation 1991) showed that approximately 27% of all American adults experience either transient (lasting only a few nights, sometimes episodic) or short-term (lasting no more than 3 weeks) insomnia, but only 9% claimed that insomnia occurred on a chronic basis. We found the same percentage of severe insomniacs, which also corresponds to the average percentage of sleeping pill users, reported in several studies (around 10%).
It makes sense that severe insomniacs are responsible for much of the burden that insomnia places on the healthcare system (e.g. the use of primary care physicians and psychiatric services, and the psychopharmacological treatment of insomnia). In the 1991 Gallup survey ( National Sleep Foundation 1991), chronic insomniacs were two-and-a-half times as likely as noninsomniacs to report accidents in which fatigue was a factor.
Concerning indirect costs, poor sleep and insomnia can result in fatigue, impaired alertness, impaired cognitive ability, reduced productivity on the job and increased opportunity for human error and fatigue-related accidents ( Balter and Uhlenhuth 1992; Dement and Mitler 1993; Shapiro and Dement 1993; Léger 1994). Moreover, some studies have considered chronic insomnia to be a major factor in the development of depression and alcoholism ( Ford and Kamerow 1989), or a powerful predictor of death, even after controlling for other factors, such as physical health ( Wingard et al. 1982 ). Stoller (1994) estimated the cost of insomnia in the United States as being between $92.5 and $107 billion for 1988. Severe insomniacs, as defined by our study, are not the only ones responsible for this burden, but they are probably more responsible for these socio-economic consequences than are less severe insomniacs.
The fact that most studies ignore the impact of insomnia on daytime functioning undermines the value of these studies in determining the socio-economic costs of insomnia, and the absence of standardized criteria for defining insomnia diminishes the usefulness of many of the studies performed. From our survey, we recommend that DSM-IV criteria should be applied in epidemiological surveys on the prevalence of insomnia, just as they are in clinical studies. Because these criteria may better reflect the symptoms and the consequences of this common disorder, they should be employed consistently in epidemiological studies to define both insomnia and severe insomnia. If this is done, our ability to understand the epidemiology of insomnia and the socio-economic consequences of chronic insomnia will be greatly enhanced.