This study used a difference-in-difference method to investigate the impact of the IA Agreement on the prevalence and mean duration of sickness absence separately for young men and women with musculoskeletal and psychological diagnoses, and to identify whether economic activity and graded SA modified these effects. Our results indicate that there are differences between those with and without the IA Agreement, as those working in companies with the IA Agreement tended to have a shorter mean duration of both musculoskeletal- and psychological-related SA. This result was even stronger when considering only those on full SA. The potential impact of the IA Agreement on men and women varied according to economic activity. The only clear trend in DID estimates was observed in the wholesale and retail economic activity, which showed consistent benefits for both prevalence and mean duration in both diagnoses and both genders for those working in companies with the IA Agreement.
Previous evaluations of the IA Agreement come from reports and peer-reviewed studies, and suggest either a positive effect [5, 8, 9, 23, 28], or no significant effect on overall SA [7, 10, 29]. Our results indicate a general beneficial contribution of the IA Agreement towards reduced duration of both musculoskeletal and psychological diagnoses in both genders, particularly in men, and a mixed contribution with regards to prevalence. However, few of these estimates were statistically significant, meaning that these trends could be due to chance.
If the trends can be attributed to the IA Agreement, our results could indicate that the measures included in the IA Agreement contribute more towards faster return to work than prevention of initial SA. This is supported by the fact that many of the IA-related measures focused on longer-term SA are related to maintaining contact with the individual and adjusting the workplace to ensure faster return to work [11]. Graded SA also has the same aim [4]. Full SA episodes were generally shorter than graded SA episodes for those with the IA Agreement compared to controls. This could indicate that IA companies facilitate for graded SA to ensure the individual can participate in working life, where the individual would ordinarily have continued with full SA.
Economic activities varied in how and to what extent the IA Agreement impacted SA prevalence and duration. As mentioned, only the wholesale/retail economic activity showed a consistent beneficial impact of the IA Agreement on both prevalence and duration, though many economic activities showed a beneficial trend with regards to mean SA duration. Only musculoskeletal SA appeared to show a trend towards an impact of the IA Agreement on full and graded SA, with shorter SA episodes on full SA and more frequent, longer SA episodes with graded SA. Economic activity seems therefore to have a modifying effect on any potential impact of the IA Agreement on SA, which is in line with previous studies and reports [5, 7, 10]. Potential explanations for differences between economic activities may lie in how much effort economic activities have put into implementing the IA Agreement [7], or through the degree of manual labour involved and the ease with which tasks can be adjusted [10]. We did not have information relating to potential differences in the level of effort available in this study, but it is possible that economic activities do have differing levels of effort into implementation [30]. We did find that economic activities that tend to be associated with manual labour (e.g. wholesale/retail and construction) also tended to show a beneficial impact of the IA Agreement. However, we also found a similar result for psychological-related SA, which is not necessarily correlated with manual labour and has been shown to respond differently to workplace interventions [18]. We did not find any clear trends relating to gender within the economic activities, indicating that economic activity may play more of an important role than gender when considering the effectiveness of IA-related measures.
Methodological considerations
This study used the DID method, which aims to observe and evaluate effects of quasi-experiments, such as the IA Agreement, where no large-scale RCTs are possible. However, DID includes assumptions that are very difficult to test in practice [22]. We were able to visually inspect the trend in all-cause SA in our study population prior to study start, but the young age of the study population meant we could not focus specifically on musculoskeletal and psychological SA. The distribution of SA duration is also skewed, which could introduce some bias into our results, though we chose the negative binomial regression method to try to account for this. This could be mitigated by using a DID approach that uses the median, though this requires a different method of analysis and additional assumptions [31].
We also controlled for variables that could cause the groups to have different levels in SA at baseline (e.g. mean company size) [5, 17]. We did not, however, have information on other potentially important confounders, such as sector (public or private) or employees’ work histories, which may influence group membership and level of SA and could thus have affected our estimates [17, 23]. Another important assumption underlying DID is that the intervention and control groups are well-defined [22], which includes the assumption that individuals cannot randomly switch group. We excluded those who switched group in 2000 or in 2005, as well as those in a different group in 2005 compared to 2000, but we included individuals who switched groups in the 4-year period between 2000 and 2005. When excluding those who changed IA status between 2000 and 2005, the results were similar to those of the original analysis. It is, however, important to follow up studies using DID with other analytical approaches, in order to understand more about the causal effects of interventions, including the IA Agreement.
We only included those in work for more than 10 months during the year (2000 or 2005), and we excluded people who switched group (intervention/control) or economic activity. These criteria could result in the exclusion of vulnerable individuals who have temporary contracts or who are struggling to find a secure and stable job, a situation that may be prevalent among our population of younger adults (aged 24–38). Applying these inclusion/exclusion criteria, though necessary to ensure proper exposure to the intervention, could limit the extent to which our findings can be applied to the general younger working population. In addition, younger individuals are less likely to experience SA compared with older adults [17] and therefore the SA levels in this study are not representative of the general working population. Finally, due to data limitations, only SA > 16 calendar days were included; this limits the generalisability of our findings to short-term SA (< 16 days).
Implications and future research
The first goal of the IA Agreement was to reduce SA by 20% from the 2001 level [5]. This goal was not reached [4]; however, the IA Agreement may still have contributed to meaningful reductions in SA, particularly for SA duration. An example of this is the reduction in mean duration of musculoskeletal SA, which was almost 17 days in men. In addition, IA companies appear to use graded SA to keep people in contact with the workplace during illness, which aids in achieving the overarching goal of keeping people in work [4].
The variation found between outcome measure, diagnosis, gender and economic activity in this study suggests that the overall impact of the IA Agreement is considerably heterogeneous. This indicates the importance of the economic activity-specific focus in the current IA Agreement [11], and suggests the potential relevance of focusing more on gender differences. Future studies should look closer at the reasons behind the heterogeneities; for example, whether differences are due to overall implementation of specific IA-related measures, which we did not have information on in this study, or due to variance in measure implementation that may depend on, for example, company motivation or job tasks. Looking closer at economic activities such as the wholesale/retail economic activity, which showed a consistent beneficial impact of the IA Agreement in this study, may provide further insights into what aspects of the IA Agreement contribute to SA reduction. Stratifying by occupational categories would also be useful, to study differences according to job tasks. Additionally, it would be beneficial for future studies to identify gender and economic activity differences in other samples, e.g. older samples or the whole working population.
Looking beyond Norway, the results indicate that other countries considering national interventions to reduce SA may find it useful to know that such interventions could have differential impact depending on the economic activity and gender. This would allow them to tailor the intervention accordingly. Our results also indicate that there may be variations in effects dependent on which SA outcome countries are interested in reducing (prevalence versus duration). Lastly, countries considering interventions to reduce SA are recommended to implement such interventions in a way that allows for proper evaluation, e.g. by random allocation of the intervention.