The association between self-efficacy and sick-leave among men and women: a cross-sectional study of the general working population in Sweden

Background: The aim of this study was to investigate if low self-efficacy was associated with increased risk of sick-leave, in a general population of employed women and men. The aim was also to analyse differences in self-efficacy concerning age, education, income, and socio-economic position. MeThods: This cross-sectional study was based on data collected in western sweden, 2008. The study population consisted of 2,900 employed sick-listed individuals (e-sL) and 2,649 random working population individuals (r-WP). Both mailed questionnaire, including the general selfefficacy scale (gse) and register data on age, education, income and socio-economic position were used. a continuous mean score of the total gse was calculated for each individual. a low gse-score indicated low general self-efficacy. resuLTs: Lower general self-efficacy had an increased odds ratio (or) of belonging to a sick-listed general working population among both men (or=1.60; 95% cI 1.32–1.94) and women (or=1.26; 95% cI 1.08–1.47). The or remained significant after adjustments for socio-demographic variables. Yet, men in the r-WP and women in both the r-WP and e-sL with lower education, income or socioeconomic position had lower general self-efficacy compared with those in each cohort with higher education, income or socio-economic position. concLusIons: Low self-efficacy was associated with increased probability of belonging to a sicklisted general working population. although more research is needed, it seems highly relevant to take both self-efficacy and socio-economic factors into account in preventive and rehabilitation work targeting people on sick-leave.

THE ASSOCIATION BETwEEN SELf-EffICACy ANd SICk-LEAVE

InTroducTIon
The costs of sick-leave in Sweden is high and over half a million people, approximately 10% of the Swedish work-force, are absent from work due to illness and disability [1].Sick-leave does not only affect the economy of the society but also has deleterious consequences on people's quality of life [2].Several studies have also shown an association between lower socio-economic status and higher rates of sick-leave [3][4][5].However, little is known on how personal characteristics, such as self-efficacy, influence sick-leave, especially in relation to socio-economic status.
Self-efficacy is a concept that has been considered to have an impact on health behaviour and rehabilitation [6][7][8].The concept has a central role in Bandura's social cognitive theory [6,9] and refers to the idea that the person's own confidence in changing behaviours or situations is decisive on the outcome.In other words, the stronger the self-efficacy, the higher are the possibilities of success.Research focusing on self-efficacy and health has indicated that health behaviour among the chronically ill can be improved and that self-efficacy can be used as a tool to improve the outcome of chronic diseases [7].These studies concern self-efficacy in specific domains or situations.Other studies have focused on a more generalised sense of selfefficacy that refers to a broad sense of selfconfidence in dealing with stressful situations and challenging demands [10,11].
Sick-listed workers' self-efficacy and expectations have also been linked to return to work possibilities [12][13][14], and high selfconfidence and belief in returning to work and self-efficacy predict an early return to work.Yet, little research has focused on whether there initially are any differences between newly sick-listed employees and employees in a general working population.In one study, targeting men and women sick-listed due to musculoskeletal disorders, low self-efficacy, in terms of a low sense of mastery, was associated with prolonged sick-leave [15].In another study, Labriola et al. [16] found that general self-efficacy was lower in a group of employed sick-listed people compared to the general working population.These are the only two studies on the association between sick-leave and self-efficacy.To find tools preventing longterm sick-leave we do need to further explore the relevance of self-efficacy for sick-leave.
The socio-economic status in terms of level of education, occupational class and income has in previous studies been related to sickleave [3][4][5].In a register-based study in Finland, a high level of education, the occupational class of the employee, such as manager or professional, and high income were associated with lower sick-leave among both men and women [4].Similar findings were reported by Laaksonen et al. [5], who found that sick-leave was more common among manual workers than among managers and professionals in both men and women.Little is known on perceived self-efficacy with respect to diverse socioeconomic groups.In one study on women with breast cancer, a higher educational level was found to be associated with better general self-efficacy [17].No information is available on how socio-economic factors may affect the association between sick-leave and self-efficacy in the general working population.
Research on the association between general self-efficacy, socio-demographic factors and sick-leave is limited, especially studies targeting a general working population of men and women.The main aim of this study was to investigate if low self-efficacy was associated with increased risk of sick-leave, in a general population of employed women and men.The aim was also to analyse differences in selfefficacy concerning age, education, income, and socio-economic position.

MeThods
This cross-sectional study was a part of the 'Health Assets Project' (HAP), a longitudinal study on health, sickness absence and work [18].Baseline data was collected by way of a postal questionnaire distributed between 15th of April 2008 and 30th of June 2008 in the Västra Götaland Region, Sweden.This region includes both urban and rural areas, with a population of approximately 1.6 million (17% of the Swedish population) at the time of the study.The postal questionnaire consisted of questions concerning socio-demographic factors, employment, general self-efficacy, sick-leave and working life.Two reminders were sent out.Additionally, socio-demographic variables such as gender, age and income were collected from the register of Statistics A pilot study and a cognitive test were performed on a sample from Statistics Sweden corresponding with the final target group of the questionnaire.HAP was approved by the Ethical Review Board in Gothenburg, Sweden: registration number 039-08.More detailed information on study procedure of HAP is published elsewhere [18].

study population
The study base comes from two different cohorts (19-64 years) originating from the general population in the Västra Götaland Region, Sweden: an employed sick-listed cohort (E-SL) and a random population cohort (R-WP) (Figure 1).The E-SL was identified by the Swedish Social Insurance Agency (SSIA).The random population cohort was identified by Statistics Sweden, and was randomly selected from the Register of the Total Population [19].The R-WP represented 0.5% of the population in the region (n=7,984).In Sweden, the first 14 days, except for one qualifying day, of sick-leave spells for the gainfully employed are covered by the employee.From day 15 the sickness benefits are paid by the SSIA.Irrespective of the reasons for the sick-leave, all consecutively newly registered sick-listed individuals (≥15 days) were identified and included by the SSIA during a period from 18th of February 2008 to 15 th of April 2008 (n=6,140).
All in all, 3,310 of the E-SL cohort and 4,027 of the R-WP cohort participated, leaving a response rate of 50.4% and 53.9% respectively.The drop-out pattern was similar for both cohorts.The drop-out rate was higher among men (drop-out = 55%), in the younger age-group 19-30 years (men 67%, women 49%), among people with the lowest income, ≤149,000 SEK/ year (men 67%, women 52%), and among those born outside Sweden and the other Nordic countries (men 64%, women 57%).
The study population of this study were those who, at the baseline data collection, stated that they were employed and working, and who also answered the General Self-Efficacy Scale (GSE) [10].So, from the study base 410 people (36% men and 64% women) were excluded in E-SL cohort and 1,378 (44% men and 56% women) in the R-WP cohort.The final study population consisted in E-SL of 2,900 (33% men and 67% women), and in R-WP of 2,649 (45% men and 55% women).The study population procedure is presented in Figure 1.THE ASSOCIATION BETwEEN SELf-EffICACy ANd SICk-LEAVE assessments Self-efficacy was the independent variable, measured by the GSE [10].GSE is an instrument assessing one's self-efficacy to cope with a variety of stressful situations and challenging demands.This instrument has been used in several international studies [11], and been found to have high validity and reliability [11,20].GSE contains ten items with four response categories ranging from 'exactly true' to 'not at all true'.For each individual the score for all items was summed up and divided by the number of items creating a continuous mean score ranging from 1-4.A low GSE-score indicated low general self-efficacy.Individuals with missing values on one or more items on GSE were excluded.

dependent variable
In the logistic regression models we estimated the OR for belonging to the employed sick-listed cohort.

statistical analyses
The analyses were performed using SPSS version 19 (Statistical Package for Social Sciences).All analyses were carried out separately for men and women [21,22].The internal consistency of GSE in E-SL, tested with Cronbach's alpha [23,24], was 0.92 for men and 0.91 for women.The corresponding values in the R-WP were 0.90 for both men and women.
Chi2-tests were used in the descriptive analyses to examine differences in the proportions between the E-SL and R-WP cohorts due to age, education, income, and socio-economic position [25].Differences in the mean scores of self-efficacy concerning age, education, income, and socio-economic position were then examined within the E-SL and R-WP cohorts respectively.This was done with one-way ANOVA analysis using Tukey's honest significance test [25].Logistic regression models were used to analyse the possible association of belonging to the E-SL cohort or not (dependent variable; E-SL=1 and R-WP=0) and low self-efficacy (independent variable: GSE score), as well as to be able to control for socio-demographic variables [26].First, a univariate logistic regression model was constructed for men and women respectively to analyse any association between self-efficacy and belonging to the E-SL (Univariate Model).To adjust for socio-demographic variables two forward multiple logistic regression models were created: Multiple Model 1 adjusted for age and Multiple Model 2 for education, income and socio-economic position.

resuLT
A higher proportion of men in the E-SL cohort were in the oldest age group (51-64 years), compared with men in the R-WP cohort (33% versus 47%).Compared with the men in the R-WP, a lower proportion of men in the E-SL had a university education, high income (≥300 000 SEK/year) and was employed in nonmanual work (Table 1).
Among women in the E-SL cohort, a higher proportion was in the oldest age group compared with the women in the R-WP cohort (42% versus 36%) (Table 1).A lower proportion of the women in the E-SL had a university education than the women in the R-WP had (40% versus 45%).Compared with the women in the R-WP, a lower proportion of women in the E-SL had an income of ≤149,000 SEK/ year and ≥300,000 SEK/year (Table 1).Of the women in the E-SL, a higher proportion was employed in skilled/non-skilled manual work compared with women in the R-WL (46% versus 37%).differences in general self-efficacy concerning socio-demographic characteristics within the employed sick-listed cohort and within the random working population cohort Among men, table 2 shows that those in the R-WP with the lowest education had a e 9 0 0 6 -4 THE ASSOCIATION BETwEEN SELf-EffICACy ANd SICk-LEAVE significantly lower GSE compared with those who had a university or a high school education in the R-WP.Men in the R-WP with an income of 150,000-299,000 SEK also had a lower GSE than those with the highest yearly income (≥ 300,000 SEK).Compared with men employed in higher non-manual work, men in the R-WP employed in skilled/non-skilled manual work had a lower GSE.No such differences were found in the E-SL.
In the E-SL cohort, women with a lower education had a lower GSE than women with a higher education (Table 2).This was also observed among the women in the R-WP.
Women in E-SL with a lower yearly income had a lower GSE compared with women with the highest yearly income (≥ 300,000 SEK) in this cohort.This pattern was also seen among the women in the R-WP.In the E-SL cohort, women employed in lower socio-economic positions had a lower GSE than women in higher non-manual work.Women in the R-WP, employed in skilled/non-skilled manual work had a lower GSE compared with women in the R-WP employed in non-manual work (Table 2).

dIscussIon
The main finding of this study was that low general self-efficacy showed an association with increased probability of belonging to a sicklisted general working population in both men and women.This finding remained significant after the adjustment for socio-demographic variables.No study, so far, has been conducted on the importance of general self-efficacy dimensions in relation to sick-leave in a newly sick-listed cohort.Our findings do, however, confirm the results from Labriola et al. [16], where general self-efficacy was measured with parts of the general self-efficacy scale in a general population, and from Busch et al. [15], who measured self-efficacy in a sample with nonspecific chronic musculoskeletal disorders.Labriola et al. [16] discuss the possibilities that low general self-efficacy among the sicklisted is a consequence of the sick-leave itself, rather than a cause for it.This aspect is important to consider, especially for healthcare professionals working with sick-listed clients.A study on a full-time multidisciplinary program using self-efficacy training showed that sickleave declined after intervention, as compared to matched controls [27].Other studies on interventions directed to chronic disability or specific diagnoses have also shown that self-efficacy can be used as a tool to improve chronic health outcomes [7,28].However, more research, particularly intervention studies, using self-efficacy as a tool, and targeting a sick-listed population, is required.
In this study, the socio-demographic characteristics differed between the two study groups: in the E-SL the participants were older, had lower educational levels, income and socioeconomic positions compared to the R-WP.The results of the social gradient on sick-leave correspond with several studies [3][4][5].However, we also found that the level of self-efficacy differed between different socio-demographic groups.In the R-WP, we found lower selfefficacy among both men and women with lower educational levels, income and socioeconomic positions.Additionally, women in the E-SL with lower income and lower socioeconomic positions had a lower self-efficacy.Previous research [15,16] on sick-leave and selfefficacy did not include detailed information on the influence of socio-economic status.Whereas the social gradient on sick-leave is known, no previous research on general self-efficacy differences between socio-demographic groups has been carried out.Our findings are novel and contribute to an identified knowledge gap.
Self-efficacy was defined by Bandura as people's perception of their abilities to organize and carry through with certain actions in order to achieve specific goals [6,9].Based on Bandura's social cognitive theory and self-efficacy, health education programs have been successful in improving quality of life and functional capacity in chronically disabled people [7].Additionally, some evidence has been found for rehabilitation programs targeting sick-listed people that show self-efficacy training to be successful in reducing sick-leave [27].Since sick-leave contributes largely to both the direct and indirect costs for society as well as for the individual [1,29] much more research is needed in this field.Especially, evaluating rehabilitation programmes targeting people on sick-leave and taking both self-efficacy and socio-economic factors into account seem to be highly relevant.

Methodological considerations
One of the most important strengths of this study is the selection and the use of the two cohorts.As far as we know, no studies are available focusing on self-efficacy in a cohort of consecutively newly sick-listed individuals representing a variety of occupations.Also, the association between self-efficacy and sick-leave THE ASSOCIATION BETwEEN SELf-EffICACy ANd SICk-LEAVE has not, until now, been analysed using a cohort of new sick-leave cases and a random cohort from the general working population.Although our study had a rather high and systematic dropout rate it was similar in both cohorts.Therefore, the comparisons between the cohorts should not be biased by this.The selection of the study base is also less biased than it would have been if only sick-listed individuals with a specific diagnose were included.The results from this study may be generalised to a larger general working population.This study is cross-sectional and conclusions regarding causality cannot be drawn.
Furthermore, our study used data collected with the GSE scale, which has been validated and has high reliability in several countries in different settings and populations [11,20].The GSE scale in our study also showed high internal consistency in the two cohorts for men and women respectively.However, it is important to be cautious with the validity of self-reported data.The GSE has answers on a pre-coded scale.There is a possibility that some of the participants may have felt pressed to choose a defined answer [24].It can, however, be argued that the attitude towards participating and answering the GSE is positive in this study population since the internal drop-out, i.e. missing values on the GSE, was low [30].It is also important to notice that the outcome originated from register data.
Missing responses on the GSE scale from one or more items meant exclusion of that person from our study.However, this procedure differs somewhat from the recommended way of handling missing responses on the GSE scale, which is to calculate a mean of the missing items if three or less answers are missing [10].Although this way of excluding people with missing values has been used for other questionnaires, such as the Short-Form 36 [24], the validity when using it in the GSE scale has not been fully evaluated [11,20].It can also be argued that there might be a reason why these individuals did not respond to certain questions in the GSE scale.This could have introduced a systematic bias if individuals who chose not to answer certain items of the GSE scale were from specific socio-demographic groups [24].Our way of handling missing responses resulted in 258 missing answers, compared with 113 if the recommended way had been used.

concLusIons
Low self-efficacy was associated with increased probability of belonging to a sicklisted general working population, in both men and women.Although a weaker socio-economic position meant lower self-efficacy, this did not affect the association between self-efficacy and sick-leave.Due to the large costs of sick-leave, it ought to be relevant to take both self-efficacy and socio-economic factors into account in preventive and rehabilitation work, targeting people on sick-leave.To be able to investigate the causality of the association between self-efficacy and sickleave, prospective studies are needed.
Funding: The collection of the data has been financed by the Social Insurance Agency.

ConFliCt oF interest:
The authors declare that there is no conflict of interest.
Public Health -2014, Volume 11, Number 2 THE ASSOCIATION BETwEEN SELf-EffICACy ANd SICk-LEAVE Sweden.

figure 1 fLOW
figure 1 Dependent variable: Belonging to the employed sick-listed cohort = 1and belonging to the random working population cohort = 0; Independent variable: low General Self-Efficacy; 1) Included only: currently sick-listed in the employed sick-listed cohort and not currently sick-listed in the random working population cohort; Univariate Model = un-adjusted odds ratios; Multiple Model 1 = Controlled for age; Multiple Model 2 = Controlled for socio-economic variables (education, income and socio-economic position); CI, 95% confidence interval; OR, odds ratio; * p<0.05, *** p<0.001 e 9 0 0 6 -6 O RIGINAL ARTICLES Epidemiology Biostatistics and Public Health -2014, Volume 11, Number 2

TAbLe 3
LOgiSTiC regreSSiON MODeL ANALYSeS THe ASSOCiATiON beTWeeN LOW SeLf-effiCACY AND SiCK-LeAve