Background
Sickness absence is costly to workers, employers, and society. In a 2008 report, the Danish Ministry of Employment presents that the yearly cost of sickness absence corresponds at least 37 billion DKK in unproductive wages and sickness absence subsidies alone in Denmark [
1,
2]. Moreover, sickness absence is a considerable risk factor for workers permanently exiting the labour market [
3‐
5].
Physical work demands – physical activity, movements, and postures at work – are amongst the dominant causes of long-term sickness absence [
6,
7]. Physical work demands such as stationary standing [
7,
8], sitting [
9,
10], forward bending of the trunk [
7,
8,
11], and arm elevation [
7,
8,
11] have been shown to be associated with sickness absence [
6‐
8,
12,
13].
However, there are three overall issues with the research literature on physical work demands and sickness absence: 1) physical work demands have mainly been measured using self-reports, 2) the analytical methods used in previous studies on physical work demands have predominantly ignored the compositional nature of the data, and 3) studies analysing sickness absence using register data, have generally omitted short-term sickness absence.
First, previous studies on physical work demands included self-reported information on physical work demands [
14,
15], that have been presented to be less accurate than technical measurements [
16‐
20] (e.g., for sitting time). Such technical measurements are in this case accelerometers attached to the body of study participants; accelerometers use accelerations of the body [
21] to measure physical activity, movements, and postures. Therefore, future studies investigating the association between physical work demands and sickness absence are likely to strengthen the field of research when using accelerometer measurements.
Second, the vast majority of existing studies analysing physical work demands have investigated the effect of each physical work demand ‘in isolation’ of other physical work demands. For example, by, e.g., investigating the health effects of sitting time without taking into account the time spent on all remaining demands such as standing, or resting. Time-use on various physical work demands is constrained or fixed by nature – summing up to 100% – (or for example 8 h). Therefore, the proportion of time spent on physical work demands carries relative information, is co-dependent. Addressing this special property of data on physical work demands requires special statistical methodology – Compositional Data Analysis (CoDA) [
22‐
25]. Only recently, a limited amount of studies have used Compositional Data Analysis approaches to address the compositional property of physical work demands [
24,
26]. However, none of them have investigated the association between physical work demands and sickness absence. Therefore, future studies investigating the association between time spent on various physical work demands and prospective sickness absence using a Compositional Data Analysis approach are needed.
Third, the previous studies have often used sickness absence using self-reports that have less validity than sickness absence information from national registers in Nordic countries [
27]. Studies using national register-data on sickness absence have predominantly used long-term sickness absence (see, e.g., [
8,
12,
28‐
30]). Nevertheless, physical work demands has also been associated with short-term sickness absence [
31], and, like sickness absence overall [
32‐
35], is likely to be placing a considerable economic burden on workplaces and society. Studies investigating both long-term and short-term sickness absence register data are thus warranted.
Aim
The purpose of the present article is to present the protocol for the ‘The technically measured compositional Physical wOrk DEmands and prospective association with register-based Sickness Absence study (PODESA)’. Specifically, to counter the above-presented challenges of previous studies, PODESA will be the first study to investigate the association between technically measured compositional data on physical work demands and prospective register-based data on short- and long-term sickness absence. PODESA will investigate the following hypothesis:
The composition of physical work demands is associated with prospective sickness absence.
Statistical analyses
We expect to analyse the association between time-use in various physical work demands and prospective long-term sickness absence primarily using regression models such as time-to-event analyses based on a Compositional Data Analysis approach. This entails several steps: First, we will transform the 24-h compositional data on physical work demands and physical activity behaviour at leisure using an appropriate log-ratio method. Second, depending on the analytical definition of the outcome, we will adopt time-to-event methods (as the main analyses, we expect to use Cox time-to-event regression on sickness absence data from the DREAM sickness absence register). The models will be adjusted for variables such as age, sex, body mass index, smoking status (for similar covariates, please see, e.g., [
8]). Third, to enable understanding how time spent in various postures and movements at work is associated with prospective sickness absence, we expect to use isotemporal substitution models [
54] indicating association between reallocation of time-use in various physical work demands and the change in probability of prospective sickness absence.
Moreover, having access to detailed longitudinal register data from the two types of sickness absence registers enables us to code the outcome variable in several ways, such as binary, percentage of sickness absence, trajectories, and time-to-event, thus also enabling additional analyses further unravelling the relationship between physical work demands and sickness absence.
Discussion
The PODESA study comprises high quality data, including technical measurements of physical work demands from accelerometer data, which provide a more precise depiction compared with, e.g., self-reports. Furthermore, the combined PODESA cohort data and sickness absence register data enable us to conduct analyses not only on the links between singular exposures and sickness absence, but also on the association between relative time-use on specific physical work demands and prospective sickness absence, e.g., using Compositional Data Analysis. Additionally, having access to prospective sickness absence data on both short-term and long-term sickness absence enables us to exploit the longitudinal nature of the data in analyses using, for example, trajectories or time-to-event analyses.
Conversely, there are also potential weaknesses to discuss. First, despite having high-quality data on physical work demands, the PODESA cohort has no objectively measured data on changes in physical work demands over time; if there is a non-random change, e.g., due to an organizational change or change in worker instructions at the workplace, it could affect the prevalence and timing of prospective sickness absence. Second, as the cohort data stem from a Danish context, generalizability of findings to other countries in terms of, e.g., inter-country variations in occupational policy is limited. Third, as the majority of participants in the study are blue-collar workers, the findings will primarily be generalizable to this group of workers.
The findings from PODESA can be used to develop improved preventive workplace interventions for sickness absence. For instance, if the results show which combinations of physical work demands may increase – and which may decrease – the risk of sickness absence, this information can be used to design better future preventive workplace interventions for sickness absence. Results from PODESA are expected in 2019–2021.
Acknowledgements
The authors wish to thank Jan Høgelund for sharing his insights into sickness absence data registers.