The data for this study was obtained from a national health examination survey. The study (Health 2000) was carried out in Finland, a north-eastern European country with about 5 million inhabitants. The fieldwork with data collection was carried out between September 2000 and July 2001. The two-stage stratified cluster sampling design was planned by Statistics Finland. The sampling frame comprised adults aged 30 years and over living in mainland Finland. This frame was regionally stratified according to the five university hospital regions, each containing roughly one million inhabitants. From each university hospital region or catchment area, 16 health care districts were sampled as clusters (80 health care districts in the whole country, including 160 municipalities). The 15 biggest health care districts in the country were all selected in the sample and their sample sizes were proportional to population size. The remaining 65 health care districts were selected by systematic probability proportional to size sampling in each stratum, and their sample sizes (ranging from 50 to 100) were equal within each university hospital region, the total number of persons drawn from a university hospital region being proportional to the corresponding population size. The 80 health care districts were the primary sampling units, and the ultimate sampling units were persons who were selected by systematic sampling from the health centre districts. From these 80 health care districts, a random sample of individuals was drawn using the data provided by Population Register Centre. Its population information system contains the official information for the whole country on the Finnish citizens and aliens residing permanently in Finland.
For this study, all the persons aged 30 or over (n = 8028) identified and selected by The Social Insurance Institution of Finland were contacted. Interviewers attended training sessions on the specific themes that were to be covered in the computer assisted interviews. During the interviews, the respondents were handed an information leaflet, an informed consent form for signing, and a questionnaire containing self-reports such as the SPAQ, the 15D, the GHQ-12 and the Beck Depression Inventory (BDI) that interviewees were asked to fill in and bring along to the health status examination.
Of the final sample of 7979 persons, 6986 (88%) were interviewed at home or institution face to face and 6354 (80%) attended the health status examination in a local health center or equal setting, while 416 took part in the health status examination at home or in an institution. Overall, 84% participated either in the health status examination proper or in the examination at home. All the methods are reported more in detail on the Internet site of the Health 2000 (for details, please see http://www.ktl.fi/health2000).
Health-related quality of life
The HRQoL was measured using two instruments, the 15D and the GHQ-12. The 15D instrument measures 15 dimensions including mobility, vision, hearing, breathing, sleeping, eating, speech, elimination, usual activities, mental function, discomfort and symptoms, depression, distress, vitality, and sexual activity . It contains five ordinal levels on each dimension, and the respondent is instructed to choose from each item the level which best describes the current health status. 15D is a generic, comprehensive, standardized measure which yields both a profile and a single index score. Higher scores indicate better levels of the HRQoL. The index of zero to one, representing the overall HRQoL, is calculated by using a set of population-based preference or utility weights. The 15D scores are highly reliable and can be generalized in Western-type societies (for further information, please see http://www.15D-instrument.net).
In addition to the 15D and its depression and distress dimensions, we wanted to assess more in detail the part of the HRQoL to which mental well-being contributes by using the 12-item GHQ. It is scored on a four-point Likert-like scale (less than usual, no more than usual, rather more than usual, or much more than usual), yielding a sum score ranging from 0 to 36. Higher scores indicate greater mental ill-being. The GHQ was developed in the 1970s with the purpose to evaluate mental health and has been applied in a range of settings and cultures . Its original version contains 60 items, but the instrument is available as shortened forms, like as the GHQ-12. This version evaluates whether the individual complains about a recent symptom or behavior. The GHQ-12 is documented well, easy to complete and valid as a screening tool . It is a valid measure of the psychological symptoms at population level, especially in the areas of anxiety and depression . According to the analysis of data derived from the Health 2000 Study, the threshold value of 4 was taken to indicate ill health (the scores of 0 to 4 assigned as low and those of 5 to 36 as high).
Seasonal changes in mood and behavior
Seasonal changes in mood and behavior were measured using items taken and adapted from the SPAQ . Two modifications were made to the original scoring as follows. Each item was scored from 0 to 3 (none, slight, moderate or marked change), not from 0 to 4 (none, slight, moderate, marked or extremely marked change), with the sum or global seasonality score (GSS) ranging from 0 to 18. Higher scores indicate greater seasonal changes. In addition, the SPAQ has a question: "If you experience changes with the seasons, do you feel that these are a problem for you?". This item was scored from 0 to 4 (none, mild, moderate, marked or severe problem), not from 0 to 5 (none, mild, moderate, marked, severe or disabling problem). The questionnaire was translated into Finnish and then back-translated in order to revise the linguistic accuracy. Since the seasonal changes in mood and behavior were assessed with a modified questionnaire, we tested earlier its psychometric properties and found them to be good in the adult population of ours , yielding a population-based distribution of the GSS across individuals similar to the original one . The modified questionnaire was thereafter applied for assessment using the cut-point of 7 (the scores of 0 to 7 assigned as low and those of 8 to 18 as high) which is similar to the original case-finding criteria .
Experienced exposure to illumination
Exposure to illumination was measured using two items which had not been validated earlier. Concerning the experienced indoors illumination, two items of the experienced lighting levels were analyzed. Poor lighting at home (yes or no) and insufficient lighting at work (not present or no problem, troubles to some extent, troubles quite a lot, or troubles exceedingly) were assessed as part of in the computer assisted interview. The sum of the scores on the two items was calculated and categorized for the analysis. Higher scores indicate poorer lighting conditions.
We decided that it was important to include a measurement of depression as an explanatory variable in the analysis. Therefore, we assessed the behavioral manifestation and symptom intensity of depression using a modification of the 21-item BDI  as adapted and validated for the Finnish population (for further information, see http://www.kela.fi), with a sum score ranging from 0 to 55. The modified questionnaire was thereafter applied for the case-finding definition using the cut-point of 9 (the scores of 0 to 9 assigned as low and those of 10 to 55 as high). Higher scores indicate more severe depressive symptoms. However, no diagnosis of depressive disorder can be assessed with the BDI.
Other variables used in the analysis of data were as follows. As part of the assessment, the participants filled in items concerning their leisure time exercise, alcohol use during the past 12 months, activities outdoors, and social activities. The intensity of physical exercise was categorized as follows: low (no strenuous exercise such as reading, watching television or handicraft), medium (lightly strenuous exercise such as walking or bicycling for four or more times a week), keep-fit (fitness training for three or more hours a week), and sport (sports for several times a week). The frequency of alcohol use was categorized as follows: none, low (once to six times a year), medium (once to four times a month), and high (twice to seven times a week). The frequencies of social activities (meeting relatives, friends or neighbors) and of activities spent outdoors (exercise, hunting, fishing, gardening or other outdoor recreation) were categorized as follows: low (less than once a year), medium (once a year to twice a month), and high (once to seven times a week).
The National Public Health Institute coordinated and implemented the study project in collaboration with the Ministry of Social Affairs and Health. It provided a written informed consent to each participant, giving a full description of the protocol before signing it. The procedures were according to the ethical standards of the responsible committee on human experimentation and with the Declaration of Helsinki, its amendments and revision.
The data were weighted to take into account the sampling design and to reduce the bias due to non-response. The R project for Statistical Computing (R, version 2.2.1) was applied for, and its survey Package, available through the Comprehensive R Archive Network family of internet sites http://www.r-project.org, was run for analysis of the stratified data using survey-weighted generalized linear models.
We wanted to know whether there was any association of the health-related quality of life with the seasonal changes in mood and behavior or with the illumination levels. To that end, first, a multivariate regression model using the indexed sum score on the 15D and another model using the categorized sum score on the GHQ-12 as the dependent variable were computed. For both models, the following explanatory variables: the sex, age in four categories (30 to 45, 46 to 60, 61 to 75 or 76 to 99 years), education in three categories (low, middle or high), marital status in two categories (living alone or with someone), area of living in two categories (the southern or northern part of Finland), physical exercise, alcohol use, the GSS in two categories (0 to 7 or 8 to 18), illumination levels experienced in two categories (not poor and not a problem, or poor or of trouble to any extent), activities outdoors, and social activities. In the former model, the BDI sum score in two categories (0 to 9 or 10 to 55) were in addition included as a covariate. Second, the two models in which the GSS was replaced by the six items of which the GSS is comprised were computed in order to elucidate which of the seasonal changes explained the association best.