The joint contribution of insomnia and obstructive sleep apnoea on sickness absence


  • Børge Sivertsen,

    Corresponding author
    1. Department of Clinical Psychology, University of Bergen, Bergen, Norway
    2. Division of Psychiatry, Helse Fonna HF, Haugesund, Norway
    • Department of Public Mental Health, Division of Mental Health, Norwegian Institute of Public Health, Bergen, Norway
    Search for more papers by this author
  • Erla Björnsdóttir,

    1. Faculty of Medicine, University of Iceland, Reykjavik, Iceland
    Search for more papers by this author
  • Simon Øverland,

    1. Department of Public Mental Health, Division of Mental Health, Norwegian Institute of Public Health, Bergen, Norway
    2. Research Centre for Health Promotion, University of Bergen, Bergen, Norway
    Search for more papers by this author
  • Bjørn Bjorvatn,

    1. Department of Public Health and Primary Health Care, University of Bergen, Bergen, Norway
    2. Norwegian Competence Center for Sleep Disorders, Haukeland University Hospital, Bergen, Norway
    Search for more papers by this author
  • Paula Salo

    1. Finnish Institute of Occupational Health, Helsinki, Finland
    2. Department of Psychology, University of Turku, Turku, Finland
    Search for more papers by this author


Professor Børge Sivertsen PhD, Department of Public Mental Health, Division of Mental Health, Norwegian Institute of Public Health, Bergen, Norway, Kalfarveien 31, 5018 Bergen, Norway.

Tel.: +47 55 58 88 76;

fax: +47 55 58 98 77;



Several studies have indicated a high degree of overlap between insomnia and obstructive sleep apnoea, but little is known regarding how the overlap may affect adverse outcomes associated with each of the disorders. The aim of the current study was to examine the separate and combined effects of symptoms of insomnia and obstructive sleep apnoea on long-term sick leave. We used an historical cohort design with 4 years follow-up. Information on sick leave was obtained from Norwegian official registry data, and merged with health information from the Hordaland Health Study in western Norway, 1997–99. A total of 6892 participants aged 40–45 years were assessed for self-reported symptoms of insomnia and obstructive sleep apnoea (snoring and breathing cessations), as well as confounding factors. The level of overlap between insomnia and obstructive sleep apnoea was low (7–12%). Both insomnia and obstructive sleep apnoea alone were significant risk factors for subsequent sick leave after adjusting for confounding factors (odds ratios ranging from 1.4 to 2.3). Having comorbid insomnia and obstructive sleep apnoea increased the risk significantly. There was an additive interaction effect between the two conditions in the unadjusted analyses, but this was reduced to a non-significant level when adjusting for confounders. This study is the first to report the separate and combined effects of insomnia and obstructive sleep apnoea on any adverse outcome. Having both insomnia and obstructive sleep apnoea increased the risk of later sick leave, but there was no evidence of an independent synergy effect of the two conditions.


Sleep problems are some of the most common complaints in the general population, and are associated with a range of individual consequences. Sleep problems have also been estimated to represent a significant economic burden, with insomnia and obstructive sleep apnoea (OSA) causing the largest costs (Hillman et al., 2006; Leger et al., 2012). Several studies have examined the association between insomnia and OSA and found a high degree of overlap between the conditions (Beneto et al., 2009; Björnsdóttir et al., 2012). Co-existing insomnia and OSA may exacerbate the severity of each condition, as well as increase the cumulative medical and psychiatric morbidity (Krakow et al., 2001; Luyster et al., 2010; Smith et al., 2004), and may reduce patients' quality of life (Björnsdóttir et al., 2012).

Accumulated evidence has demonstrated that sleep problems are associated with subsequent sick leave and work disability. By linking large epidemiological health surveys [including the current Hordaland Health Study 1997–99 (HUSK)] with official registries on work status, studies from several countries have shown that insomnia represents a strong and independent risk factor for participants' inability to stay within the work force (Eriksen et al., 2003; Lallukka et al., 2011; Overland et al., 2008; Salo et al., 2010; Sivertsen et al., 2006, 2009a,b). Similar adverse consequences have been found for OSA; both self-reported and physician-certified OSA has been shown to almost double the risk for subsequent sick leave and work disability, even after adjusting for potential confounding factors (Sivertsen et al., 2008; Sjosten et al., 2009).

Given two common conditions, both linked to adverse outcomes and with a high degree of overlap, a prospective exploration of their separate and joint effects on objectively defined occupational outcomes seems warranted. Therefore, the aims of the current study were to: (1) examine the overlap between self-reported DSM-IV insomnia and symptoms of OSA; (2) investigate the separate and combined contribution of insomnia and OSA on subsequent sick leave; and (3) explore if any association is explained by confounding factors.


Population and data material

The HUSK was a joint epidemiological research project carried out by the Norwegian Health Screening Service in collaboration with the University of Bergen. The base population for the study reported herein included all 29 400 individuals in Hordaland County, western Norway born 1953–57, aged 40–45 years at the time of data collection. Data were collected by questionnaires and clinical examinations. A total of 18 581 (8598 men and 9983 women) answered both the basic questionnaire and came to the clinical examinations, yielding a participation rate of 63% (57% for men and 70% for women).

After the clinical examinations, a second questionnaire including questions on sleep was distributed to a subgroup comprising 8896 individuals, of whom 7888 (89%) provided valid responses. Participants who were receiving disability pension benefits at baseline or were awarded disability pension within 12 months after baseline were excluded (n = 266). By excluding all disability pensions from 0 to 12 months after participation in HUSK, we aimed to exclude subjects who were in the process of applying for disability pension while they attended HUSK. Participants who were on sick leave or otherwise reporting not working at the time of HUSK were also excluded (n = 730). Thus, the final population comprised 6892 individuals (59% females).

Using official data from Statistics Norway and the National Insurance Administration, a follow-up study of non-responders was conducted on different demographic variables (Husk Research Center, 2004). In sum, it was found that non-responders were more likely to be male and of older age. Adjusting for these factors, non-responders had significantly less education and annual income compared to responders. Also, more non-responders received disability benefits, including sick leave compensation and disability pension awards. No data were available on clinical variables for non-responders.


Dependent variables

The National Insurance Administration records all periods of sick leave beyond 14 days, as well as all disability pension awards. In Norway this is, in essence, a public responsibility, and because correct registration is a prerequisite for transfers of payments, the records are highly accurate. In the present study, we used two durations of accumulated sick leave, 4 and 8 weeks, within 48 months following the HUSK baseline.

Independent variables

Symptoms of insomnia were assessed using the Karolinska Sleep Questionnaire (Kecklund and Åkerstedt, 1992). The DSM-IV criteria for insomnia include difficulty falling asleep, difficulty maintaining sleep or experiencing non-restorative sleep for a period of no less than 1 month. In addition, it is a prerequisite that the sleep disturbance significantly impairs daily functioning. In this study, the measurement of insomnia diagnosis was based upon four items, each rated along a five-point scale [‘never’, ‘rarely (a few times per year)’, ‘sometimes (a few times per month)’, ‘mostly (several times a week)’ or ‘always’]. Subjects were categorized as having insomnia if they reported sleep onset insomnia, maintenance insomnia or early morning awakening insomnia (or a combination) ‘several times a week’ or ‘always’ during the past 3 months, in addition to reporting impaired work performance during the preceding year caused by the sleep problems. No item assessing non-restorative sleep was used.

Symptoms of OSA were estimated using two items from the Karolinska Sleep Questionnaire (Kecklund and Åkerstedt, 1992). These self-report items were used to identify at-risk individuals based on their own or their partner's reports on ‘snoring' and ‘breathing cessation’ during sleep. Participants were classified as having symptoms of OSA if they reported both these core symptoms either ‘sometimes (several times a month)’, ‘often (several times a week)’ or ‘always’. Based on these operationalizations of insomnia and OSA, four groups were created: (1) insomnia only (no OSA), (2) OSA only (no insomnia), (3) comorbid OSA and insomnia and (4) good sleepers (subjects not meeting criteria for insomnia or OSA).

As many OSA patients also experience daytime sleepiness, we created a separate operationalization of OSA (for sensitivity analysis), in which it also was required that the patients also reported being ‘tired or sleepy at work or during spare time’, ‘sometimes’, ‘often’ or ‘always’. This operationalization has also been used in a previous study from the same data material examining the effect of OSA on disability pension (Sivertsen et al., 2008). A similar definition based on the Hawaiian Sleep Questionnaire (the apnoea score) has been shown previously to identify 100% of the cases with severe sleep apnoea [apnoea–hypopnea index (AHI) > 40] and 75% of all sleep apnoea cases with AHI > 5, and an overall predictive accuracy of 88% for AHI > 10 (Kapuniai et al., 1988).

Potential confounders

Level of education was reported in four categories from less than 7 years of schooling to at least 4 years of higher education in college/university. We also collected data on marital/cohabitant status (dichotomized into living alone or with partner), smoking status (current smoker: yes or no) and weekly level of exercise: (1) no or easy physical activity 1 h week, (2) moderate physical activity 1–2 h week or (3) hard physical activity more than 2 h week. Alcohol consumption was operationalized using four categories based on weekly number of self-reported alcohol units (none, 1–2, 3–4 or ≥5 units week). Body mass index (BMI) was calculated from body weight by squared height from the clinical examinations. Blood pressure was measured as part of a clinical examination.

Questions on somatic diagnoses were framed in the form of: ‘Do you have or have you had (one or more of the following)’: asthma, allergy, myocardial infarction, stroke, diabetes, or angina. Positive responses on these items were considered self-reported diagnosis positive, coded dichotomously to indicate presence of that illness.

Symptoms of current anxiety and depression were measured using the Hospital Anxiety and Depression Scale (HADS) (Zigmond and Snaith, 1983), which is a self-report questionnaire comprising 14 four-point items, seven for anxiety (HADS-A) and seven for depression (HADS-D). No somatic items or items regarding sleeping difficulties are included. The HADS scores were used as continuous variables, reflecting symptom load of anxiety and depression.

Statistical analysis and models

spss for Mac 20 was used for all statistical analyses. Pearson's chi-square tests with Yates' correction were used to examine the association between insomnia and OSA. Chi-square tests were also conducted to examine demographic and clinical differences between the four groups (good sleepers, insomnia only, OSA only and comorbid OSA and insomnia). Multivariate logistic regression analysis was used to examine the relation between the sleep variables and sick leave. Only significant clinical and demographic variables from the initial group comparisons were entered as confounders in the logistic regression analyses. Additional logistic regression analysis was also conducted to examine the effect of OSA with and without daytime sleepiness. Participants who were granted disability pension awards during the follow-up (= 139) were excluded from the analyses. Results are presented as odds ratios (OR) with 95% confidence intervals (CI).

We also explored the additive interactions between insomnia symptoms and OSA by examining whether reporting both had a synergistic effect on sick leave. This was calculated using the algorithm suggested by Andersson et al. (2005), in which the synergy index (SI) is equal to calculation of [OR (AB)–1]/[(OR(Ab)–1p(OR(aB)–1)], where A and B denote the presence of two risk factors (insomnia and OSA) and a and b are designated as the absence of these risk factors. An SI of 1.0 implies perfect additivity and >1 indicates synergistic interaction.


The study protocol was cleared by the Regional Committee for Medical Research Ethics of Western Norway and approved by the Norwegian Data Inspectorate. Informed consent in writing was obtained from all subjects included in this study.


Sample characteristics

The prevalence of insomnia and OSA was 5.1 and 8.4%, respectively. There were no significant gender differences in the prevalence of insomnia, but more males than females reported OSA (13.9% versus 3.5%, χ2 = 259.3, df = 1, P < 0.001). The demographic and clinical characteristics of the study sample are presented in Table 1. In short, 4.5% of the participants reported having insomnia without OSA, 7.7% had OSA without insomnia and 0.6% reported having both sleep disorders. Compared with those having insomnia or OSA only, participants with both conditions were less educated, smoked more cigarettes, consumed more alcohol, had higher BMI and more often had asthma, allergies, strokes and depressive symptoms (see Table 1 for group comparisons).

Table 1. Baseline demographical and clinical characteristics of participants with and without self-reported symptoms of insomnia and obstructive sleep apnoea (OSA) in the Hordaland Health Study, Norway
CharacteristicsGood sleepersInsomnia onlyOSA onlyComorbid insomnia and OSAOverall P-value
  1. a

    Data presented as mean [95% confidence interval (CI)].

  2. b

    1 unit equals approximately 12 g ethanol.

n, %6007, 87.2309, 4.5534, 7.742, 0.6 
Male, %<0.001
Agea42.6 (42.5–42.6)42.5 (42.3–42.7)42.6 (42.4–42.8)42.9 (42.4–43.3)0.540
Living with partner76.165.077.764.3<0.001
Education, %
1–3 years higher20.017.519.121.4
≥4 years higher18.623.613.12.4
Current smoker, %<0.001
Alcohol consumption, %b
0 units week27.925.220.016.7<0.001
1–2 units week67.769.671.769.0
3–4 units week4.
Physical exercise, %
No or easy28.534.632.033.30.108
Body mass index, %
Asthma, %
Allergy, %12.511.314.825.70.050
Diabetes, %
Angina, %
Stroke, %
Myocardial infarction, %
Depressiona2.9 (2.8–2.9)5.8 (5.4–6.3)3.5 (3.3–3.7)5.5 (4.4–6.6)<0.001
Blood pressure (systolic)a129.5 (129.1–129.9)127.9 (126.3–129.4)134.4 (133.2–135.6)131.0 (126.0–135.9)<0.001

Association between insomnia and OSA

The prevalence of OSA was 12.0% among subjects with insomnia, compared to 8.2% among non-insomniacs. Conversely, the prevalence of insomnia was 7.3% among subjects with OSA, compared to 4.9% among subjects without OSA (χ2 = 6.29, df = 1, P = 0.012). Fig. 1 illustrates the overlap of OSA and each of the insomnia symptoms.

Figure 1.

Overlap of insomnia symptoms and obstructive sleep apnoea (OSA). The Hordaland Health Study (1997–99). Error bars represent 95% confidence intervals.

Combination of insomnia and OSA as risk factor for sick leave

As detailed in Table 2, subjects with insomnia (but without OSA) were approximately twice as likely to be on subsequent sick leave compared to good sleepers (crude ORs 2.2 and 2.5 for 4 and 8 weeks sick leave, respectively). For subjects with OSA (but without insomnia) the associations were weaker, with crude ORs 1.2 and 1.1 (not significant) for both sick leave outcomes. For subjects fulfilling the criteria for comorbid insomnia and OSA, the risks were higher than merely summing the two specific disorders (crude ORs 3.1 and 2.7 for 4 and 8 weeks sick leave, respectively), indicating an additive interaction effect in the crude analyses for 4 weeks sick leave (SI = 1.5; see Fig. 2).

Figure 2.

Insomnia and obstructive sleep apnoea (OSA) and possible synergy effect as risk factors for subsequent sick leave (4 weeks) in the Hordaland Health Study, 1997–99.

Table 2. Combinations of obstructive sleep apnoea (OSA) and insomnia as risk factors for subsequent sick leave
 Sick leave duration
4 weeks8 weeks
Odds ratio95% CIOdds ratio95% CI
  1. CI, confidence interval.

  2. Adjusted for age, gender, education, smoking, alcohol use, body mass index, allergy, asthma, systolic blood pressure, angina, stroke and depression.

Insomnia only2.191.71–2.802.461.92–3.16
OSA only1.190.97–1.471.120.90–1.40
Comorbid insomnia and OSA3.131.55–6.302.731.38–5.42
Fully adjusted*
Insomnia only2.071.60–2.692.311.78–3.01
OSA only1.431.15–1.781.371.08–2.74
Comorbid insomnia and OSA2.421.17––4.35

When adjusting for confounding factors, the ORs were reduced significantly for insomnia (adjusted ORs 2.1–2.3), but increased for OSA (adjusted OR 1.4). The joint contribution of comorbid insomnia and OSA was stronger than having either insomnia or OSA alone when using 4 weeks as the outcome, but there was no significant additive interaction (see Fig. 2 for details). There were no significant additive interaction effects when using 8 weeks sick leave as the outcome measure, neither in the crude or adjusted analyses.

OSA with and without daytime sleepiness

Logistic regression analyses were also used to investigate the effect of OSA with and without daytime sleepiness on sick leave. OSA without daytime sleepiness was not associated significantly with subsequent sick leave (4 or 8 weeks) in either crude or adjusted analyses, whereas OSA with daytime sleepiness was associated strongly with sick leave [4 weeks adjusted OR = 1.73 (95% CI: 1.36–2.124) and 8 weeks adjusted OR = 1.57 (95% CI: 1.21–2.03)].


The main aim of this study was to explore the separate and combined effect of insomnia and OSA on subsequent sick leave. In short, we found that having both conditions increased the risk compared to having either insomnia or OSA. While the crude analyses showed a significant synergy effect of comorbid insomnia and OSA on 4 weeks sick leave, this effect was reduced to a non-significant level after adjusting for confounding factors, and there was no interaction using 8 weeks as the outcome.

Previous studies on the association between insomnia and OSA have found a high level of overlap, with figures ranging from 42 to 55% (Beneto et al., 2009). However, the nature of this relationship remains poorly understood, and it has been suggested that the causality works both ways. For example, insomnia is likely to develop in an OSA patient as a result of repeated awakenings causing dysfunctional sleep behaviours. Conversely, it has been speculated that insomnia may lead to more nocturnal arousals and superficial non-rapid eye movement (NREM) in phase 1, which again may increase the instability of the upper airway, and thereby exacerbate OSA symptoms (Series et al., 1994). However, if this was the case in all OSA and insomnia patients, it is difficult to explain why some display comorbid insomnia and OSA while others do not. An overlap between the two conditions in the current study of only 7–12% is considerably lower than in previous studies. One possible reason for this relatively low level of overlap may be a measurement issue. First, although the operationalization of insomnia resembles both the DSM-IV criteria and Research Diagnostic Criteria (Edinger et al., 2004), no items on non-restorative sleep were included. Non-restorative sleep is also one of the core symptoms of OSA, and also not included in our operationalization of OSA, and this should explain parts of the low overlap. Secondly, OSA was operationalized based on two self-reported items, and not polysomnography, which is the gold standard. However, a previous study using a similar evaluation of OSA predicted an apnoea index of >10 with a sensitivity of 83% and a specificity of 63% (Kapuniai et al., 1988), and another study found self-reported symptoms of snoring and breathing stops to yield a very high level of specificity for OSA (99%), although the sensitivity was somewhat lower (Bliwise et al., 1991). Nevertheless, under ideal circumstances, self-report data should be validated by polysomnographic readings, although this is rarely performed in such large population studies. Another explanation for the low overlap of insomnia and OSA may relate to the study sample. Both insomnia and OSA increase significantly with age, and the limited age range in the current study (40–45 years), in addition to the sample being relatively healthy, may also lead to an underestimation of the association between the two conditions. A final reason might be that some of the previous studies on the OSA–insomnia overlap have been conducted on clinical samples, possibly comprising study participants with higher morbidity, and thus more likely to have comorbidities of correlated conditions.

The current study found that both insomnia (without OSA) and, to a lesser extent, OSA (without insomnia), were associated with subsequent sick leave. The effect of insomnia was higher than OSA, both in the unadjusted analyses and after controlling for sociodemographic factors and other symptoms and conditions. Although the reported effect sizes corroborate earlier studies on the same issue, including two studies from the same data set (Eriksen et al., 2003; Lallukka et al., 2011; Overland et al., 2008; Salo et al., 2010; Sivertsen et al., 2006, 2008, 2009a,b; Sjosten et al., 2009), the current study is the first to explore whether the combined effect of insomnia and OSA constitutes a synergetic effect for sick leave. We found some evidence of this in the unadjusted analyses. However, when adjusted for confounders, we found no additive interaction effect: the effect on sick leave of having comorbid insomnia and OSA was no stronger than merely adding the risks for each disorder. As such, the additive interaction found in the unadjusted analyses was explained by confounding factors.

As expected, we found that patients with OSA who also reported daytime sleepiness had a significantly stronger risk of subsequent sick leave compared to OSA patients without daytime sleepiness. This corroborates the previous finding that sleepiness is a stronger risk factor for disability pension than the items on snoring and breathing cessations (Sivertsen et al., 2008).

There are important limitations to the present study. First, insomnia was established by self-report rather than clinical diagnosis. However, the questionnaire in this study (Karolinska Sleep Questionnaire) is relatively analogous to the criteria for insomnia as specified by DSM-IV, and the prevalence rate found in the present study is similar to previous reports (Ohayon, 1997), suggesting a reasonable estimate of insomnia in this age cohort. However, in addition to not assessing non-restorative sleep, we measured insomnia on only one occasion, and as sleep problems fluctuate commonly over time, it would be better to have repeated measurements over a long time-period in order to identify chronic insomniacs. Also, assessing daytime impairment caused by the insomnia symptoms was limited to work impairment only. An important addition to the DSM-IV from previous versions was the inclusion of a clinical significance criterion to almost half of all the categories (including insomnia), which required that symptoms cause clinically significant distress or impairment in social, occupational or other important areas of functioning. A similar impact on family, social or occupational life is also required in the latest ICD-10 criteria for insomnia.

Another important limitation of the present study is the measurement of OSA. Rather than employing the gold standard of a clinical diagnosis based on polysomnographic recordings, the present study is based on a brief self-report questionnaire used to categorize people into two groups: OSA or no OSA. The prevalence estimate (8.4%) is similar to that found in other epidemiological studies based on the general population in similar age cohorts (Bresnitz et al., 1994; Kripke et al., 1997; Ohayon et al., 1997; Partinen, 1995; Young et al., 1993). Nevertheless, the use of self-reported symptoms to measure OSA remains problematic, because people are often unaware of their behaviour during sleep. However, polysomnographic data are not easily obtained in epidemiological studies, and the use of self-reports by patients are often the only feasible way of acquiring information about this. One way of improving the validity of such reports is by also including spouse-reported information on snoring and breathing cessations. The Karolinska Sleep Questionnaire is based on sleep problems also reported by the person's spouse, and thus attempts to improve the validity of symptoms. However, as noted by Grunstein et al. (1995), a potential source of misclassification into OSA and non-OSA groups may relate to the presence of a current home partner, and people sleeping alone may, as such, be more likely to be misclassified as not having OSA. In our analysis, adjusting for marital/cohabitant status did not attenuate the effect, and such a misclassification would tend to produce an underestimation in the health differences between the OSA and non-OSA groups, given that living alone is associated characteristically with poorer health (Grunstein et al., 1995).

Another limitation is that the list of confounding factors will most probably not capture and fully attenuate all possible confounding from chronic somatic or psychiatric conditions. Self-report instruments are prone to error and residual confounding cannot be ruled out, and screening for psychiatric morbidity was limited to symptoms of anxiety and depression only.

Other limitations of the present study include the limited age range (40–45 years), and that our sample included participants with both higher education and annual income compared to non-responders, while non-responders were more likely to receive disability benefits.

The strengths of the present study include a relatively large population with a high participation rate at baseline. Also, both exposure and outcome assessments should be relatively unbiased, as neither participants nor administrators were aware of this specific research hypothesis (as follows from this study being a secondary analysis of existing data). This reduces the possibility of information being biased by selective symptom presentation in order to gain access to, or avoid, benefits. Thirdly, data on work status at baseline and at follow-up were ascertained from the National Insurance Administration. These data are complete (those people moving to other parts of the country after participating in HUSK are still registered), and should not have been influenced by exposure status.


These findings demonstrate that people with insomnia and symptoms of OSA are at increased risk of subsequent sick leave. Having both conditions increased the risk compared to having either insomnia or OSA. There was a significant synergy effect of comorbid insomnia and OSA in the crude analyses, which was explained by confounding factors. Nevertheless, both conditions are likely to be significant contributors to increased social security costs and reduced productivity and family income. Our findings warrant replication using a better-validated measure of OSA and insomnia in a sample with a wider age range. If maintained, such findings would support an additional focus on social and occupational outcomes for these patient groups.

Disclosure Statement

None of the authors have any conflicts of interest to declare.