Changes in attitudes to psychiatric disorders and increased distress?
One finding of this study was that the increase in sickness absence with psychiatric diagnoses occurred in both men and women, in all age groups, in all regions and in the diagnostic groups with the most sick-listed. This pervading increase implies that underlying causes were related to factors that affect men and women, different age groups and different regions in a similar way. Stansfeld et al.  suggested different explanations for the increase of sickness absence with psychiatric diagnoses. Among these were changes in attitudes to psychiatric disorders in the general population, affecting both patients and doctors. With a less stigmatising attitude to psychiatric disorders, more patients disclose such symptoms, doctors become more inclined to identify these diagnoses, and both patients and doctors accept to a greater extent the recording of a psychiatric diagnosis on the certificate. Stansfeld et al.  further suggested that the availability to physicians of better drugs (e.g. SSRI drugs) for the treatment of psychiatric disorders might lead to improved and increased identification of psychiatric disorders. Furthermore, there has been a steady increase in the proportion of sickness absence spells with a psychiatric diagnosis. In 1995, 11.6% of sickness episodes longer than 16 days had a psychiatric diagnosis; in 2000, 15.7%. In the same period the total number of sickness absence spells increased from 316,114 to 501,708. Thus, both overall and psychiatric sickness absence increased in the study period, but psychiatric sickness absence had a greater increase .
The increase in sickness absence with psychiatric disorders might also reflect an increase of psychiatric disorders or mental distress. Many researchers and public health reports in Scandinavia have expressed doubt about a significant increase in psychiatric disorders in the population, but have confirmed increased distress and a decrease in psychological wellbeing in general . Increased stress and decreased psychological wellbeing might lead to a decline in mental as well as physical health, such as increased pain, headache, and other musculoskeletal and gastrointestinal symptoms. As mentioned in the introduction, Sandanger et al.  found that the relation between health problems and sickness absence differed between psychiatric disorders and musculoskeletal disorders, with a suggested underreporting of sickness absence with psychiatric diagnoses. We therefore think that the most plausible explanations for the pervading increase in sickness absence with psychiatric disorders are an increase in stress in the population, and changes in attitudes of both patients and physicians leading to a higher number of people being sick-listed with a psychiatric diagnosis than previously. Whether this merely reflects a change in certification behaviour or whether there has also been a change in how individuals express their stress and low psychological wellbeing via different symptoms is difficult to ascertain.
A higher incidence among women remains and differences between the sexes increase
Another important finding of this study was that the cumulative incidence as well as the increase in cumulative incidences of sickness absence was highest in women aged 30–59 years. In the systematic review of research on sickness absence with psychiatric diagnoses, the higher incidence of women was the only finding that could be identified as evidence-based . The higher incidence corresponds well to the higher prevalence of anxiety and depressive disorders found in women compared to men . Several other explanations have been put forward, such as gender differences in health-seeking behaviour, in symptom presentation, symptom interpretation of both patient and physician, in psychological development and socialisation, and in exposure to risk factors such as sexual and violent abuse, other traumatic events, sole responsibility for children, and poverty [21–24]. The finding that women not only had a higher cumulative incidence but also the largest increase is important from a public health point of view, as it means that an increasing proportion of Norwegian women are affected by absence from work relating to mental illness. It is not likely that the change in attitudes discussed above had a specific effect on women aged 30–59, so we must look for other explanations. In both men and women the increase was greatest among ages when activity in the labour market is highest in combination with responsibilities for family and household, with increased stress in both sexes . However, the effect of the combination of paid and unpaid work seems to be higher in women, even if several studies have also shown that the health effect of combining paid and unpaid work is beneficial for women's (and men's) health. During the 1990s, workplaces in Norway, as in the rest of Europe, increased their efficiency by downsizing organisations, and this also affected typically female-dominated workplaces in health, child and elderly care, service professions and education, with increased demands placed both on employees and managers in these organisations. Organisational changes, management quality, low control and low social support are factors that have been shown to be associated with increased sickness absence in women [1, 6, 26, 27]. It is very plausible that the high increase of sickness absence with psychiatric diagnoses in women from 1994 to 2000 reflects increased demands in female-dominated workplaces, in combination with an unchanged distribution of unpaid work in the family and household . The increase in men probably reflects the same kind of increased demands but without the specific problems associated with the public sector (low wages, low control, less skilled managers, high emotional stress related to work with people rather than things) and without the high unpaid workload still associated with the female role in society. However, it is important to remember that we have no data on occupations in this study, and some of the gender differences might be explained by different distribution over socio-economic strata in men's and women's occupations. The role of work in sickness absence with psychiatric disorders need more attention in future studies , and of specific interest is health-related selection that can contribute to bias in cross-sectional studies on gender differences.
An inclusive labour market
Another possible explanation for the increasing rates is that during the study period Norway had falling unemployment rates, from 4.9 % in 1995 to 3.4 % in 2000 . With lower unemployment rates the possibility for individuals with health problems to get a job is higher. The proportion of individuals with health problems in the work force thus increase which might contribute to high sickness absence rates. Individuals with higher vulnerability due to health problems are likely to be the most sensitive to a more demanding labour market. It is possible that individuals with a vulnerability for mental illness are more sensitive to such changes, especially in a work life, which places increasing demands on the individuals to perform well both regarding social and cognitive functions .
Regional inequity, selection or differences in health?
The capital Oslo had high cumulative incidences. An increased risk for alcohol and drug problems in urban areas as well as an increased risk for psychoses might contribute to these findings, even if these diagnostic groups had low cumulative incidences. Stansfeld et al.  found an underreporting of psychoses on medical certificates so people with alcohol and drug problems and psychoses might be sick-listed with other diagnoses. The higher urban incidence has been explained by selection of individuals with these health problems into large cities, or by exposure to living conditions more characterised by poverty, low social support, high criminality and access to drugs [29–31]. Andersson et al.  found an increased risk for disability pensions with psychiatric diagnoses in men in Oslo, but not in women. Psychoses and substance abuse among men was in that study found to be more prevalent in Oslo, while the distribution of neurotic and somatoform disorders showed no regional differences. For women there were no regional differences in psychoses and alcohol and drug disorders, while neurotic disorders were more prevalent in semi-rural areas. It is possible that the high sickness absence found among women in Oslo will lead to increased disability pensions, as several studies have found that an important risk factor for future disability pension is earlier sickness absence [32, 33]. To explain regional differences, information on the distribution over occupations in different regions would have been helpful in order to add in controls for possible selection or confusion associated with labour market factors and occupational roles. Another factor of interest is the access to mental health care and rehabilitation.
This study was performed based on a national social insurance register covering almost the whole Norwegian population. Apart from earlier studies in Norway by Hensing et al.  and Nystuen et al. , there are no other general population-based studies on sickness absence with psychiatric diagnoses that have included such a large number of individuals. As the major part of the population is included, the selection bias common in studies based on specific occupations or workplaces is not an issue in this study . The validity of diagnoses on sickness certificates has not been studied to a very large extent, but it can be hypothesised that the specificity of psychiatric diagnoses (considered as a diagnostic group) is high. Brage et al.  found few changes of psychiatric diagnoses in the long-term sick-listed. It was more common that individuals originally sick-listed with musculoskeletal diagnoses were later sick-listed with psychiatric diagnoses. The validity between different diagnoses is probably lower. Stansfeld et al. [6, 11], as mentioned, found an underreporting of psychoses on sickness certificates, and Hensing et al.  did not find any alcohol diagnoses on certificates in a study of sickness absence in women with alcohol problems.
The limitation of this register-based study is that we have access to a limited number of factors to study. We have no information on occupations, working conditions or family situation, which could have contributed to explaining the differences found between women and men more fully.
In the study period 1994–2000, sickness absence episodes shorter than 14 (16) days were not recorded in the registers, and could thus not be included in the study. The duration of sickness absence episodes varies with diagnosis, and is longer for psychiatric diagnosis than other diagnostic groups (with the exception of cardiovascular disorders), indicating that shorter sickness absence episodes are infrequent in persons with psychiatric diagnoses . The exclusion of episodes shorter than 14 days therefore gives an underestimate of the real cumulative incidence of sickness absence, but is of limited importance.