The Swedish Annual Level of Living Survey
We used data from the Swedish Annual Level of Living Survey (SALLS), which has been conducted annually since 1974 by Statistics Sweden, the Swedish government-owned bureau of statistics. The survey comprises a representative, simple cross-sectional random sample of non-institutionalized individuals aged 16–84 years, drawn systematically by age group from the Swedish Total Population Register [
18], and the sample represents the entire population of Sweden. In this study, we included 2728 males and 2770 females aged 16–71 years, who were assessed every eighth year in 1980/81, 1988/89, 1996/97, and 2004/05. The sample included all who had answered at least once, and for each assessment the sample was completed with new individuals in the age span 16–23 years, and those who were older than 71 years were removed. The surveyed individuals were invited by letter to take part in the survey. Professional interviewers from Statistics Sweden conducted face-to-face interviews, usually at the respondents’ homes. Since 1979, there have been four main themes in the SALLS: social relations, work, health and the physical environment. Certain questions about health, employment, economic resources, working environment, education and housing are repeated every year in order to provide consistent information on important background variables, e.g., self-reported health, socioeconomic conditions and family type. The data are not publicly available and the use and analysis of the data require permission from Statistics Sweden.
We excluded individuals who had missing values for the variables weight or height (1%), or physical activity. Individuals with missing values for education were classified as belonging to the highest educational level. The variables sex, age, marital status, birth cohort and urbanization had no missing values.
Explanatory variables
To assess the longitudinal changes in different groups of age and birth cohort, three time-related variables were included: assessment period, age at the time of the assessment, and year of birth. Moreover, we included the following potentially explanatory variables for which previous studies have suggested an association with anxiety [
11,
14‐
17]: sex, education level, urbanization, marital status, smoking, leisure time physical activity, and body mass index (BMI). These variables were measured in each survey and included in the models as time-varying covariates. Education, smoking, cohabiting, physical activity and weight/height were self-reported, and the other variables were obtained from the register of the total population.
Assessment period comprised four categories: 1980/81, 1988/89, 1996/97, 2004/05.
Age at the time of interview was categorized into the following groups, reflecting the 8-year intervals between the assessments: 16–23, 24–31, 32–39, 40–47, 48–55, 56–63, and 64–71 years. Age was centered at 42 years in order to have a reference group within the studied age interval.
Birth cohort (based on year of birth), comprised groups born in 1910–17, 1918–25, 1926–33, 1934–41, 1942–49, 1950–57, 1958–65, 1966–73, 1974–81, and 1982–89. Birth cohort was centered at 1950.
Sex: All analyses were made separately for males and females.
Education level: Education level (comparable over the entire study period) was categorized as: (1) high (theoretical high school and/or college, ≥12 years); (2) intermediate (practical high school, i.e., vocational school, 10–11 years); and (3) low (compulsory school or less, ≤9 years).
Urbanization: Residence in: (1) the three largest cities in Sweden; (2) medium-sized towns (population > 90,000); and (3) small towns (population 27,000–90,000) and rural areas.
Marital status was dichotomized as married/cohabiting and all others.
Smoking was dichotomized as (1) non-smokers, comprising never smokers, occasional smokers and former smokers (regardless of when they quit), and (2) daily smokers.
Leisure time physical activity was based on a question about how much physical activity the person does during leisure time, with five options from (1) “basically nothing” to (5) “regularly rather strenuously at least twice a week.” The five options were dichotomized as (1) more than once a week (options 4–5) and (2) no or some physical activity, at most once a week (options 1–3).
BMI was calculated as self-reported weight (kg)/height (m)2. Subjects with missing values for either weight or height (1%) were excluded, to be able to calculate BMI. BMI was categorized into (1) normal weight (20.0–24.9 kg/m2), (2) overweight (25.0–29.9 kg/m2), and (3) obesity (≥30.0 kg/m2).
Statistical analysis
In the analysis, descriptive statistics were used to present the distributions of the variables (Table
1), as well as unadjusted prevalence of anxiety according to the explanatory variables. Differences/trends were considered as significant if p was <0.05. No adjustment of
p-values was done. We tested trends by applying a method suggested by Cuzick, which was implemented in STATA as nptrend [
19] (Table
2).
Table 1
Distribution (%) of the different variables according to sex and assessment period (longitudinal samples of the Swedish population from 1980/81, 1988/89, 1996/97, and 2004/05) in individuals aged 16–71 years
Age (years) |
16–23 | 16.3 | 16.3 | 12.8 | 12.7 | 16.2 | 16.0 | 12.0 | 13.1 |
24–31 | 17.6 | 15.5 | 16.5 | 13.1 | 15.9 | 15.8 | 16.5 | 13.0 |
32–39 | 19.6 | 16.1 | 16.5 | 16.6 | 17.0 | 15.8 | 16.6 | 16.2 |
40–47 | 13.2 | 18.1 | 16.1 | 15.2 | 12.3 | 16.8 | 16.1 | 16.1 |
48–55 | 10.6 | 12.4 | 17.9 | 14.5 | 12.9 | 11.1 | 16.5 | 15.3 |
56–63 | 12.4 | 9.5 | 11.3 | 17.1 | 13.0 | 12.2 | 11.0 | 15.4 |
64–71 | 10.3 | 12.1 | 8.9 | 10.8 | 12.7 | 12.3 | 11.3 | 10.9 |
Education level |
High | 30.8 | 35.8 | 43.6 | 52.4 | 20.2 | 29.4 | 40.9 | 54.0 |
Intermediate | 26.4 | 31.3 | 30.9 | 26.8 | 31.1 | 36.0 | 33.6 | 26.8 |
Low | 42.8 | 32.9 | 25.5 | 20.8 | 48.7 | 34.6 | 25.5 | 19.2 |
Urbanization |
Large cities | 30.8 | 31.0 | 31.2 | 34.3 | 30.5 | 30.0 | 31.4 | 33.5 |
Medium towns | 31.5 | 33.8 | 36.5 | 36.0 | 32.6 | 34.7 | 37.4 | 36.1 |
Small towns | 37.7 | 35.2 | 32.3 | 29.7 | 36.9 | 35.3 | 31.2 | 30.4 |
Marital status |
Married/cohabiting | 65.5 | 64.5 | 64.8 | 64.6 | 68.2 | 67.7 | 67.6 | 65.6 |
All others | 34.5 | 35.5 | 35.2 | 35.4 | 31.8 | 32.3 | 32.4 | 34.4 |
Smoking |
Non-smokers | 65.9 | 73.1 | 81.8 | 86.5 | 68.9 | 71.3 | 75.8 | 81.3 |
Daily smokers | 34.1 | 26.9 | 18.2 | 13.5 | 31.1 | 28.7 | 24.2 | 18.7 |
Leisure time physical activity |
None to once a week | 67.9 | 64.7 | 61.3 | 54.7 | 75.0 | 70.6 | 64.8 | 51.9 |
More than once a week | 32.1 | 35.3 | 38.7 | 45.3 | 25.0 | 29.4 | 35.2 | 48.1 |
Body mass index |
Normal | 66.6 | 63.0 | 54.3 | 47.8 | 75.8 | 74.0 | 67.0 | 64.0 |
Overweight | 29.2 | 31.9 | 38.7 | 42.3 | 19.8 | 20.9 | 26.2 | 26.3 |
Obesity | 4.2 | 5.1 | 7.0 | 9.9 | 4.4 | 5.1 | 6.8 | 9.7 |
Table 2
Unadjusted prevalence of self-reported anxiety (%) in individuals aged 16–71 years and tests for trends in the different explanatory variable groups, presented separately according to sex and assessment period (longitudinal samples of the Swedish population from 1980/81, 1988/89, 1996/97, and 2004/05)
A mixed logistic model with random intercepts was applied to test the change in prevalence of anxiety for age groups and cohorts in the four assessment periods (Table
3). The effect of time period does not need to be estimated for a longitudinal panel study, as age and time express the same effect. Including random slopes did not improve the model. There was a highly significant interaction between age and cohort, and this interaction was included in the models. We tested all other interactions between time and each risk factor, but none of these was significant. The unadjusted model included age, cohort, the age-by-cohort interaction, and age-squared. In a second model, adjustments were made for all explanatory variables, i.e. education level, urbanization, marital status, smoking, leisure time physical activity, and BMI. Age was centered at 42 years and cohort was centered at 1950. Odds ratios (ORs) with 95% confidence intervals (95% CIs) were calculated separately according to sex. Finally, adjusted prevalence of anxiety was calculated by using the models in Tables
3 and
4 (Table
5). Birth cohort trends (change per birth year) and age trends (change per year) were also calculated. The trends for each age group and cohort were estimated by applying a linear regression model with time as the independent variable and with the estimated proportions in Tables
3 and
4 as the dependent variable.
Table 3
Odds ratios (ORs) with 95% confidence intervals (95% CIs) for self-reported anxiety in males aged 16–71 years in Sweden in 1980/81–2004/05, estimated using mixed models with random intercepts
Fixed effects | | | | | |
Rate of change | | | | | |
OR by age1
| Age centered at 42 years | 1.04 | 1.03–1.05 | 1.05 | 1.04–1.06 |
Agec*cohortc | | 0.9994 | 0.998–1.000 | 0.9993 | 0.9987–1.00002 |
Agec-squared | | 0.998 | 0.997–0.999 | 0.998 | 0.997–0.999 |
OR by cohort2
| Cohort centered at 1950 | 1.03 | 1.02–1.04 | 1.04 | 1.03–1.05 |
Education level | High | | | 1 | Reference |
Intermediate | | | 1.01 | 0.79–1.28 |
Low | | | 0.96 | 0.74–1.23 |
Urbanization | Large cities | | | 1 | Reference |
Medium towns | | | 0.91 | 0.72–1.16 |
Small towns | | | 0.68 | 0.53–0.88 |
Marital status | Married/cohabiting | | | 1 | Reference |
All others | | | 2.03 | 1.65–2.51 |
Smoking | Non-smokers | | | 1 | Reference |
Daily smokers | | | 1.71 | 1.38–2.14 |
Leisure time physical activity | More than once a week | | | 1 | Reference |
None to once a week | | | 1.67 | 1.39–2.02 |
Body mass index | Normal | | | 1 | Reference |
Overweight | | | 0.88 | 0.72–1.09 |
Obesity | | | 1.23 | 0.85–1.79 |
Variance components | Variance | Standard error | Variance | Standard error |
| Var (cons) | 3.33 | 0.38 | 3.02 | 0.35 |
Table 4
Odds ratios (ORs) with 95% confidence intervals (95% CIs) for self-reported anxiety in females aged 16–71 years in Sweden in 1980/81–2004/05, estimated using mixed models with random intercepts
Fixed effects | | | | | |
Rate of change | | | | | |
OR by age1
| Age centered at 42 years | 1.04 | 1.03–1.05 | 1.05 | 1.04–1.06 |
Agec*cohortc | | 0.998 | 0.997–0.999 | 0.998 | 0.997–0.999 |
Agec-squared | | 0.998 | 0.997–0.999 | 0.998 | 0.997–0.999 |
OR by cohort2
| Cohort centered at 1950 | 1.03 | 1.02–1.04 | 1.03 | 1.02–1.04 |
Education level | High | | | 1 | Reference |
Intermediate | | | 0.93 | 0.76–1.15 |
Low | | | 1.29 | 1.04–1.59 |
Urbanization | Large cities | | | 1 | Reference |
Medium towns | | | 0.95 | 0.78–1.15 |
Small towns | | | 0.80 | 0.65–0.98 |
Marital status | Married/cohabiting | | | 1 | Reference |
All others | | | 1.81 | 1.53–2.14 |
Smoking | Non-smokers | | | 1 | Reference |
Daily smokers | | | 1.61 | 1.35–1.93 |
Leisure time physical activity | More than once a week | | | 1 | Reference |
None to once a week | | | 1.43 | 1.24–1.65 |
Body mass index | Normal | | | 1 | Reference |
Overweight | | | 1.04 | 0.87–1.25 |
Obesity | | | 1.31 | 0.97–1.77 |
Variance components | Variance | Standard error | Variance | Standard error |
| Var (cons) | 3.39 | 0.29 | 2.98 | 0.27 |
Table 5
Adjusted prevalence (%) of self-reported anxiety based on the adjusted models in Tables
3 and
4, and annual change in anxiety (ΔANX per year by age and cohort, test of trend) in males and females aged 16–71 years, presented according to age, cohort (birth year), and assessment period (longitudinal samples of the Swedish population from 1980/81, 1988/89, 1996/97, and 2004/05)
We analysed non-response carefully, and found differences in distributions between responders and non-responders. We have taken this fact into consideration by including the post stratification variables (age, marital status, urbanization and education) in the models. By including the post-stratification variables, no weights are necessary, according to Nordberg [
20]. Nordberg means that this way of compensation for missing data might even be better than calculating weights based on the same variables. In an additional analysis, we used sampling weights.
STATA version 13 [
21] was used for the statistical analyses.