Information and communication technology demands at work: the association with job strain, effort-reward imbalance and self-rated health in different socio-economic strata

Data from the Swedish Longitudinal Occupational Survey of Health (SLOSH) were used. SLOSH aims to examine connections between work participation, work environment, social situation and health/well-being (Magnusson Hanson et al. 2008, 2015; Statistics Sweden 2014). The study started in 2006 when previous respondents of the 2003 the Swedish Work Environment Survey (SWES) were invited to participate in a follow-up survey. This population has been followed up biennially thereafter, and participants also from SWES 2005–2011 have successively been invited to respond to questionnaires. In the present study, cross-sectional data collected in 2014 were used (Fig. 1). In the 2014 data collection, 38,657 SWES participants were invited, of which some had been asked earlier and some asked for the first time to respond to SLOSH follow-up questionnaires. In total, 20,316 persons (response rate 52.6 %) responded. Out of those who were invited to participate, men, younger people and people with low SES were somewhat less likely to respond (Statistics Sweden 2014). Out of the 20,316, those who were working less than 30 % in paid work or not at all (e.g. retired or on long-term sick leave), or reported that they did not use ICT in their work, were excluded, leaving 14,757 gainfully employed men (6342 [43.0 %]) and women (8415 [57.0 %]), 20–75 years of age, recruited from the entire Sweden as our analytical study sample (Fig. 1). A detailed description of the data collection of SLOSH has been published elsewhere (Magnusson Hanson et al. 2008, 2015; Statistics Sweden 2014).

SLOSH has been approved by the Regional Research Ethics Board in Stockholm, and the present study has also received complementary approval by the Regional Research Ethics Board in Linköping #2014/355-31.

ICT demands were measured by a scale specifically developed for SLOSH, based on previous qualitative work by Johansson-Hidén et al. (2003). The scale was introduced by the following ingress in the questionnaire; “New technology and more flexible working conditions have changed working life for many people. Technology can be very helpful, but also contribute to new types of stress. Estimate to what extent you are stressed by…”. The ingress was accompanied with the following six items: “…too many calls and emails”, “…claims to be available on work-related issues during work hours”, “…claims to be available on work-related issues during leisure time”, “…claims to give immediate answers to emails and telephone calls that require a lot of work”, “…constantly being interrupted by the telephone and email” and “…computers and other equipment that fail to work properly”. The response options were rated on a Likert scale from 1 (I do not have access to this at work) to 5 (very much). The population mean score was calculated and used as cut-off values for high and low ICT demands (high ICT demands were defined as strictly above the median, and low ICT demands were defined as equal to or below the median).

Job strain was measured by the demand-control questionnaire (DCQ), comprising the dimensions of demands and control (Karasek and Theorell 1990). The dimension of demands was covered by five items, e.g. “Do you have to work very intensively?”. Control was covered by six items, e.g. “Do you have a choice in deciding how you do your work?”. The response options were rated on a Likert scale from 1 (no, hardly ever) to 4 (yes, often). The population mean response scores for the demands and control dimensions were computed and used as cut-off values for high or low scores of demands and control (“low demands” were defined as equal to or below the median, and “high demands” were defined as strictly above the median; “low control” was defined as strictly below the median, and “high control” was defined as equal to or above the median). “Job strain” was defined as a combination of high demands and low control and was compared with all other combinations of the demands and control dimensions.

ERI was measured by the short version of the ERI questionnaire, which comprised the dimensions of effort and reward (Siegrist 2013). Effort was covered by three items, e.g. “I have constant time pressure due to a heavy work load”. Reward was covered by seven items, e.g. “Considering all my efforts and achievements, I receive the respect and prestige I deserve at work”. The response options were rated on a Likert scale from 1 (strongly disagree) to 4 (strongly agree). ERI was defined as a combination of high effort and low reward, calculated by dividing the effort dimension with the reward dimension and using the ratio of 1 as cut-off value, with ratios strictly above 1 considered as ERI and ratios equal to or below 1 considered as no imbalance (Siegrist 2013).

Cronbach’s alpha was calculated on all included work-related stress dimensions on the present study sample, resulting in values between 0.65 and 0.85, where the value of 0.85 represented ICT demands.

Suboptimal self-rated health was analysed as outcome measure and measured by a one-item question; “How would you rate your general state of health?”. The response options were rated on a Likert scale from 1 (very good) to 5 (very bad). In the analyses, suboptimal self-rated health (defined as responding very bad, rather bad or neither good nor bad to the question) was contrasted to good self-rated health (defined as responding very good or rather good to the question).

Sex, age, SES, lifestyle factors (smoking and leisure time physical activity), BMI and other types of work-related stress were considered to be potential confounders or mediating factors. SES was analysed in the categories: low SES (unskilled, semiskilled and skilled workers), intermediate SES (assistant and intermediate non-manual workers) and high SES (employed and self-employed professionals, higher civil servants and executives), classified in line with Statistics Sweden’s manual of the socio-economic index (Statistics Sweden 1982). Smoking was analysed in the categories never smoker, ex-smoker and smoker. Physical leisure time activity was analysed in the categories: low/occasional physical activity and regular physical activity. BMI was calculated by self-reported weight in kilograms/height in squared metres and classified into four categories: underweight (<18.50), normal weight (18.50–24.99), overweight (25.00–29.99) and obesity (≥30.00) (World Health Organization 2015).

Chi-square tests were conducted to analyse differences in proportions. ANOVAs were conducted to calculate differences between the mean values in different strata of SES. Spearman’s correlation coefficients were calculated to determine correlations between different continuous dimensions of work-related stress. Logistic regression analyses calculating odds ratios (OR) with 95 % confidence intervals (CI) were calculated to examine the association between dichotomised measures of work-related stress and suboptimal self-rated health. ICT demands, job strain and ERI were analysed separately in relation to suboptimal self-rated health by applying crude analyses and three different sequential regression models adjusted for age, sex and SES; age, sex, SES, lifestyle and BMI; and age, sex, SES, lifestyle, BMI and other types of work-related stress (e.g. job strain and ERI when analysing ICT demands). Alpha was set to 0.05. All analyses were carried out in the total study sample and stratified by SES. To test potential effect modification by SES, statistical interaction terms between the work-related stress measures and SES were included in the regression models. The software IBM SPSS Statistics 21 was used to calculate the results.

Statistically significant differences in the proportion of ICT demands, job strain and ERI were observed between different SES strata (Table 1). ICT demands were most prevalent among participants with high SES (59.8 %), followed by participants with intermediate SES (54.9 %) and low SES (29.1 %). In contrast, job strain was most prevalent among participants with low SES (29.5 %), followed by participants with intermediate SES (19.1 %) and high SES (10.9 %). ERI was most prevalent among participants with intermediate SES (53.1 %), followed by participants with low SES (50.8 %) and high SES (46.6 %). Statistically significant differences in the prevalence of suboptimal self-rated health were observed, and suboptimal self-rated health was most prevalent among participants with low SES (24.8 %), followed by participants with intermediate SES (18.2 %) and high SES (16.7 %). Social gradients in physical activity, smoking and BMI were also observed (Table 1).

In the total study sample, the continuous measure of ICT demands was correlated with all the work-related stress dimensions of job strain and ERI (Table 2). Some of the correlations were low, but the strongest correlation was observed between ICT demands and effort (r = 0.51 p ≤ 0.001), followed by demands (r = 0.42; p ≤ 0.001).

Statistically significant differences in the proportions of the dichotomised measure of ICT demands and job strain and ICT demands and ERI were observed (Table 3). Among participants with high SES, 80.7 % of those who reported job strain also reported high ICT demands. Among participants with intermediate SES, 75.2 % of those who reported job strain also reported high ICT demands, and among participants with low SES, only 36.3 % of those with job strain also reported high ICT demands. Concerning ERI, among participants with high SES, 77.8 % of those who reported ERI also reported high ICT demands, followed by 69.2 % among participants with intermediate SES and 40.0 % among participants with low SES (Table 3).

ICT demands were associated with suboptimal self-rated health in crude analysis (OR 1.35 [CI 1.24–1.46]) and also after adjustment for age, sex, SES, lifestyle and BMI (OR 1.49 [CI 1.36–1.63]) (Table 4). When adding adjustment for job strain and ERI, the association between ICT demands and suboptimal self-rated health was attenuated but remained statistically significant (OR 1.18 [CI 1.07–1.30]). However, job strain (OR 1.93 [CI 1.74–2.14]) and ERI ((OR 2.15 [CI 1.95–2.35]), adjusted for age, sex, SES, lifestyle and BMI) showed fairly stronger associations with suboptimal self-rated health.

When the analysis was stratified by SES, the association was somewhat stronger between ICT demands and suboptimal self-rated health among participants with intermediate SES (OR 1.62 [CI 1.42–1.86]), followed by participants with low SES (OR 1.39 [CI 1.18–1.63]) and high SES ((OR 1.36 [CI 1.10–1.68]), adjusted for age, sex, lifestyle and BMI) (Table 4). Similar and consistent patterns were observed in the crude and all adjusted analyses. However, test for statistical interaction between ICT demands and SES was not statistically significant in any of those models.

Concerning job strain, the association between job strain and suboptimal self-rated health was somewhat stronger among participants with low SES, followed by participants with intermediate SES and high SES (Table 4). Similar and consistent patterns were found in the crude and all adjusted analyses. Concerning ERI, the association between ERI and suboptimal self-rated health was rather similar in the different SES strata (Table 4). However, neither for job strain nor for ERI, the test for statistical interaction with SES was statistically significant.

The present study contributes to an up-to-date picture of the work-related stress in general and in different SES strata in the modern working life, and new information about how ICT demands associate with other types of work-related stress and self-rated health. Generalisability of the results is strengthened by a relatively large sample size, including participants from both sexes, a wide range of ages, different SES strata and from different parts of Sweden, which embodies a rather representative cross section of the gainfully employed people in Sweden at the present time. The representativeness of the study sample is also, to some extent, supported by similar prevalence of background characteristics (e.g. self-rated health, lifestyle factors and BMI) as have been observed in other studies (Alvarez-Galvez et al. 2013; Padyab and Norberg 2014).

ICT demands were contrasted to work-related stress measured with the DCQ model and the short version of the ERI model, both measures that have been found valid and reliable, which strengthens the internal validity (Chungkham et al. 2013; Hokerberg et al. 2014; Leineweber et al. 2010). In addition, the internal reliability was strengthened by acceptable scores of Cronbach’s alpha at the work-related stress measures, not least ICT demands that showed a Cronbach’s alpha of 0.85.

The present study also has some limitations to consider. The ICT demands scale has not been validated at the time of writing. Even though this study provides some support that the concept of ICT demands is useful as a potential stressor of work-related stress, an actual validation study of the scale used is desired. Another concern with the ICT demands scale is that it does not provide a full explanation of why some people might experience high demands linked to some ICT-related working contexts. A further development and extension of the ICT scale, including a dimension that takes the individual’s resources regarding ICT use into account, is warranted, in order to get a more complete picture of ICT demands in relation to resources.

There are also some issues about using self-rated health as a health indicator. Measuring self-rated health by a one-item question provides an indication of the general health status of a person, but this measure can also be viewed as unspecific due to its association with a number of risk factors and diseases (Stenholm et al. 2014; Waller et al. 2015), and the present study does not provide information about specific health problems that contribute to poorer self-rated health.

Some matters concerning temporality, power, precision and generalisability should also be acknowledged. The temporality in the associations between work-related stress and suboptimal self-rated health cannot be established due to the cross-sectional design. Moreover, even though the power when analysing the total study sample is good, the power is lower in some of the SES stratified regression analyses, which affects the precision of the results. In addition, the influence of selection bias cannot be ruled out due to the somewhat low response rate, and the lower participation rate among men, younger people and people with low SES. This may limit the generalisability to the general working population. Hence, future studies based on other populations are needed to confirm the results.