Health-related effects of an intervention involving reduced working hours among women employed in the municipal eldercare

Abstract

Working conditions of eldercare are often characterized by a high workload and different organizational approaches have been suggested to reduce the potentially negative health-related effects of such conditions. One of the suggested approaches involves reducing the exposure to such working conditions by reducing the number of working hours. However, the knowledge of the concurrent health-related effects of working fewer hours remains limited. This longitudinal study set out to examine the health-related effects of an intervention involving daily working hours being reduced from ≥7 to 6 h. Data came from an organizational intervention that included 68 women employed in the municipal eldercare. All employees retained full-time salaries and additional personnel were employed by the organization in order to provide full services throughout the intervention period. A broad range of biomarkers and self-ratings in questionnaires were obtained before the intervention, 6 months after the change, and 12 months after the initial change. Measurements among referents (n = 19), matched with respect to type of work, were performed at corresponding time points. Results from repeated measures ANOVAs showed significant interaction effects for diastolic blood pressure and HbA1c with these effects being primarily related to changes among referents. No other significant group differences were found. Time-related effects emerged in both groups but cannot be attributed to the intervention as such. To conclude, this study showed none of the hypothesized health promoting or other consistent effects of reduced working hours.

Keywords:

• intervention
• health care services
• self-ratings
• biomarkers
• longitudinal

Introduction

Stress-related ill health, high sick leave figures, and organizational demands have contributed to an increased interest in new ways of organizing work and reducing the time employees spend at work (e.g. Åkerstedt et al., 2001; Voglino et al., 2022). This study set out to add to the field by investigating whether reduced working hours (RWHs) are associated with positive or negative effects across a broad range of health-related indicators.

In the Nordic countries, in particular, interventions focusing on work-time arrangements have been suggested a means to reduce adverse health effects and sick-leave figures among employees (Voglino et al., 2022). One of the most common types of such work-time arrangements involves a reduction of working hours from 8 h/day to about 6 h/day. In contrast to part-time work, work-time arrangements involving RWHs mean that employees work fewer hours but still retain their full-time salaries. Often additional employees are recruited by the organization to provide fulltime services (Brynja & Bildt, 2005; Voglino et al., 2022).

Within the public service and health care sectors in the Nordic countries, several so-called experiments have been carried out to investigate effects of RWH (for overviews, see Brynja & Bildt, 2005; Voglino et al., 2022). Evaluations of these experiments show positive social effects, particularly of a 6-h workday (e.g. Åkerstedt et al., 2001; Anttila et al., 2005).

Apart from positive social effects of a 6-h workday, the potential health-related benefits of RWH remain unclear. The 6-h workday has been associated with a considerable reduction of pain in the neck/shoulder area in employees within the public sector (Wergeland et al., 2003). However, evaluations of different types of RWH have only produced small effects on other health-related outcomes, such as fatigue and sleep with findings showing reduced fatigue, improved sleep quality, longer sleep duration, and less sleepiness (Åkerstedt et al., 2001; Schiller et al., 2017; see also Brynja & Bildt, 2005; Voglino et al., 2022). Similar findings, with statistically significant differences but small effect sizes, have been shown for self-reported stress as well (Schiller et al., 2017; Voglino et al., 2022).

One obvious reason for why RWH are assumed to promote health is that employees spend fewer hours at work, which decreases exposure to work and any associated risks for work-related ill health (von Thiele Schwarz et al., 2015). This assumption draws on the occupational health research findings showing consistent linkages between exposure to different risk factors in the work environment and various types of work-related health problems (e.g. Briner, 2000; Karasek & Theorell, 1990; Sverke et al., 2017; Theorell et al., 2015). Traditionally, research has focused on three categories of risk factors: ergonomic, physical, and psychosocial risk factors. These risk factors have been linked to negative health effects in terms of impaired physical and mental health, increased stress levels and so on (e.g. Karasek & Thorell, 1990; Sverke et al., 2017; Theorell et al., 2015). Additionally, at least three other explanations have been put forth to describe how RWH, in allowing more time for non-work activities, may promote health (Åkerstedt et al., 2001; Anttila et al., 2005; Brynja & Bildt, 2005; Malmberg et al., 2003; Voglino et al., 2022). First, with more time for non-work activities, RWH may facilitate combining different life domains relating to work and non-work (e.g. family life). Second, RWH may increase time for recovery from work stress and thereby have positive health effects. Third, RWH may increase time spent on health-promoting leisure activities. Thus, less time at work is assumed to result in more time spent on non-work activities and to improve health (von Thiele Schwarz et al., 2015; see also Voglino et al., 2022).

But work may not always involve health risks. For instance, comparisons of unemployed and employed individuals show that work has positive effects on health and well-being (e.g. Jahoda, 1979; van der Noordt et al., 2014). In addition to financial support, paid work has been argued to provide a goal-directed activity and a daily structure that includes social contacts with co-workers. The resulting positive effects include increased self-esteem and feelings of mastery along with personal development, increased physical health, and well-being (e.g. Briner, 2000; Jahoda, 1979; van der Noordt et al., 2014). Thus, working fewer hours may not only decrease exposure to potential health hazards but may also reduce exposure to the potentially positive effects of work, such as social relationships at work.

With few exceptions (Åkerstedt et al., 2001; Anttila et al., 2005; Schiller et al., 2017; Wergeland et al., 2003), evaluations of health-related effects of RWH seldom include systematic longitudinal studies with measurements before, during, and after the implementation of RWH and with referents. Often a restricted range of health-related outcomes are used (Brynja & Bildt, 2005; Voglino et al., 2022). Moreover, it remains unclear whether the effects of RWH are direct or mediated via psychosocial factors. It then follows that the theoretical understanding of various health-related effects of RWH is poor which, in turn, calls for further studies of direct effects including both subjective and objective indicators of health-related outcomes (e.g. Voglino et al., 2022; von Thiele Schwarz et al., 2015).

The present study set out to examine the direct effects of RWH, including a 6-h workday, on factors pertaining to various aspects of health in women employed by the municipal eldercare, with measurements carried out before the change, after 6 months, and after 12 months. Such direct effects have been suggested in previous research of similar RWH interventions (e.g. Åkerstedt et al., 2001; Wergeland et al., 2003). To extend previous research focusing on a few health-related indicators, the present study covered a broad range of health-related outcomes including biomarkers, self-ratings of health, and psychosocial work characteristics that are known to influence health (e.g. Karasek & Thorell, 1990; Lundberg, 2002, 2005; Sverke et al., 2017). To avoid relying solely on self-reports, the chosen outcome measures included both individuals’ subjective experiences (self-ratings in questionnaires) and objective assessments (biomarkers). The self-ratings were selected to cover exposure to psychosocial work characteristics and to health-related measures such as recovery and work/home interference (WHI). Additionally, the measures covered positive as well as negative aspects of health (Ryff & Singer, 2000), such as recovery and stress. The different biomarkers were selected to allow the study of both health promoting and health damaging changes across multiple systems. This means that hormones such as dehydroepiandresterone and prolactin as well as cardiovascular measures such as heart rate (HR) were included (e.g. McEwen, 1998; McEwen & Stellar, 1993). Also, most of the included biomarkers are associated with major diseases: blood pressure and blood lipids are related to cardiovascular problems while anthropomorphic measures (e.g. waist/hip ratio), and different measures of metabolism (e.g. glucose) are related to metabolic disorders (Lundberg, 2002, 2005; McEwen, 1998). The purpose of including a broad range of outcome measures in a single study was to explore whether RWH had direct positive or negative effects on these health-related outcomes.

Drawing on previous studies (Åkerstedt et al., 2001; Anttila et al., 2005; Schiller et al., 2017; Wergeland et al., 2003), it was hypothesized that the intervention group would exhibit overall time-related improvements in the health-related measures. Specifically, we expected to replicate previous findings showing positive effects of RWH on work/home interference, musculoskeletal symptoms, fatigue, and stress. But, keeping in mind that existing studies on RWH have presented inconclusive findings, with small effect sizes, no specific hypotheses were formulated for the remaining study variables, including the biomarkers. Also, we were interested in exploring any potentially negative effects. To strengthen the research design (Kristensen, 2005; Ruotsalainen et al., 2006; Scharf et al., 2008), a reference group was included to control for time effects. So, another aim of the present study was to compare employees being part of the organizational intervention to referents employed within the same organization. Following previous research (Voglino et al., 2022), the intervention group was expected to show overall positive changes in the health-related study variables when compared to referents.

Materials and methods

Participants and organizational setting

This study forms part of a larger research project investigating different non-randomized organizational interventions, initiated by the organizations, with longitudinal designs and referents. This particular study was set in a municipal eldercare organization. A larger group of employees were included in the intervention while a smaller group of referents were recruited from the same organization. Despite being fewer, the referents held the same jobs and performed comparable work tasks within the same organizational milieu meaning that the matching of the intervention and referents targeted organizational and work characteristics at the unit level.

Like most eldercare organizations, the vast majority of the employees were women and consequently this study included women only. In all, 116 employees, of whom 93 belonged to the intervention group, were invited to participate in the study. Of these employees, 109 volunteered to participate in the study resulting in an intervention group of 81 women and a reference group of 28 women. Of these, 89 (69 in the intervention group and 20 referents) took part in all three phases of the study. Participation was voluntary but encouraged by the employer to increase the quality of the research. However, the organization had no access to information about individual employees volunteering participation or refraining from being part of the research.

Initial analyses (results not shown) based on data from the first phase of analyzing the drop-outs (n = 22) who completed the first phase only, revealed no significant differences in demographic factors or health-related measures between those who fulfilled all three phases of the study and the others. The majority of drop-outs were due to change of jobs (n = 9) and parental leave (n = 4). Seven individuals actively declined to participate in the follow-ups and two never showed up at the health checkups. Data from two diabetics were excluded from the statistical analyses due to extreme values. The final sample consisted of 87 healthy women (68 in the intervention group and 19 referents).

Design and procedure

In the intervention group, weekly working hours were reduced from a traditional 6-week scheduling of work with 7 h/day to about 6 h/day. The intervention included both day and early night shifts. A few months before its start, the intervention was planned and then introduced to employees taking part in the study. These preparations were necessary since the organization had to implement the intervention including recruiting and training additional personnel to the organization to provide full caring services throughout the study period. At the same time, referent units with employees were recruited and invited to participate in the study.

During the intervention period a typical work-week included the following: 1 h of sports and exercise, 1 h of activities aiming to support the ongoing intervention, 1 h of administrative and representative duties including searching for information while the remaining time involved caring work. The hour of activities supporting the intervention was organized in the form of study circles by the employer to promote staff skills regarding communication and mental health among the elderly. Also, these study circles allowed employees to inspire each other to adopt a healthy lifestyle and share their experiences of how to make use of the extra leisure hours. All employees in the intervention group retained their full-time salaries. The referents were employed by the same organization but worked at another workplace in the same geographical area in the municipality. The referents followed their ordinary weekly work schedules which included 1 h of sports and exercise, 1 h of weekly administrative and representative duties including searching for information, and caring work. The referents were told that they were given an opportunity to get three health checkups during work hours, for free and thus get a close monitoring of their health status during the study period.

Employees taking part in the intervention and referents were given detailed oral and written information about the study design, research ethics, and procedures for the measurements involved (self-ratings in questionnaires and health-checkups) during separate sessions. This information was repeated prior to all three phases of the study.

Data were collected at three points in time during a period of 15 months: biomarkers and self-ratings in questionnaires were obtained about 3 months before the intervention (T1), 6 months after the change (T2) and 12 months after the initial change (T3). Efforts were made to match the reference group with the intervention group with regard to demographics (gender), work schedules, and work tasks prior to the intervention. However, due to the limited finances for the project, the number of women in the reference group was smaller than that of the intervention group. Measurements in the reference and intervention groups were performed at corresponding time points. This three-phase study design was chosen to allow detailed analyses of changes between time points (Taris & Kompier, 2003). Two of the time-points (T1 and T3) set to the same season to account for any seasonal variation in the study variables. Following previous studies (Anttila et al., 2005; Wergeland et al., 2003) and considering the overall length of the intervention, the time lag between phases was set to 6 months. Also, a 6-month time lag is often used in intervention research, including RWH studies.

The self-report questionnaire was completed at home and returned to the registered nurse who performed the health checkup. The health checkups took place in the morning in a secluded room at the workplace. For each participant, health checkups at different study phases were scheduled around the same time to minimize circadian variation within individuals. All participants were instructed to refrain from eating for 12 h prior to the checkup. Also, they were asked to rise at least 2 h before the health checkup was scheduled and in the meantime refrain from consuming caffeine and nicotine and to avoid intense mental or physical activity. Compliance with instructions was checked at each health checkup. Overall compliance was high. Questions regarding menstrual status (menstruating or not), chronic diseases (yes/no), and medication (yes/no) were also asked.

All health checkups followed a specific protocol and involved sampling of blood and measurement of cardiovascular parameters and waist/hip ratio (WHR). The choice of biomarkers was guided by the aim to include commonly used indicators covering different physiological systems while also having little variation over the menstrual cycle (e.g. Ahmad et al., 2002; Theodorsson et al., 2018). Systolic blood pressure (SBP; mmHg), diastolic blood pressure (DBP; mmHg) and HR (beats/min) were measured as indicators of cardiovascular activity. Blood samples were used to determine blood lipids (mmol/l) associated with atherosclerosis and cardiovascular disease, including total cholesterol (TC), triglycerides, high-density lipoproteins (HDLs) and low-density lipoproteins (LDLs). Also, blood samples were used to determine the following biomarkers: dehydroepiandesterone sulfate (DHEAS) which is a functional hypothalamic–pituitary–adrenal (HPA) axis antagonist, glucose to measure levels of blood sugar, glycosylated hemoglobin (HbA1c) as an integrated measure of glucose metabolism during the previous 30–90 days, and prolactin which is a hormone sensitive to sleep and stress. For most biomarkers, lower values indicate good health whereas the opposite holds for HDL and DHEAS. In all, 25 ml of venous blood was sampled and then left to coagulate for a minimum of 30 min and a maximum of 120 min before being centrifuged at room temperature (15 min, 1000 g) and transported to a commercial laboratory for chemical analysis. WHR reflects adipose tissue deposition and metabolism. Waist circumference (cm) was measured at the narrowest point between the rib and the iliac crest and hip circumference (cm) at the maximal buttocks and was used to calculate WHR. A semiautomatic device (Boso, Bosch + Sohn, Germany) was used to assess SBP, DBP and HR 3 times after 5 min of rest and with 5-min rests in-between. The median value of the three recordings was used in the statistical analyses to reduce the influence of potential technical errors.

When the project was completed, all study participants were informed about the overall results of the study. The project was approved by the Swedish Central Ethical Review Board (Ref. No. 04-036/4) and follows its standards.

Questionnaire

In addition to details on demographic factors (gender, age, marital status, children, education, current occupation), the questionnaire included previously validated Swedish translations of measures of health-related factors and psychosocial work characteristics.

Perceived stress

Overall perceptions of stress were measured by the 14-item version of the Perceived Stress Scale (PSS; Cohen et al., 1983; Eskin & Parr, 1996). All items were rated along a 5-point scale ranging from Never (1) to Very often (5) and a sum score was computed with high scores indicating higher levels of perceived stress. The internal consistency coefficient (Cronbach’s alpha) varied between 0.81 and 0.88 for the three points in time.

Recovery

Recovery from work (Aronsson et al., 2003; Gustafsson et al., 2008; von Thiele et al., 2006) included both fatigue (4 items) and recovery (4 items). All items were rated along a 5-point scale ranging from Never (1) to Very often (5). For both fatigue and recovery, sum scores were computed with high scores indicating high fatigue and poor recovery respectively. The internal consistency coefficients varied between 0.71 and 0.81 for fatigue and 0.77 and 0.85 for recovery for three measurements.

Self-rated health

Self-rated health was assessed by a single-item asking the respondents to rate their current health status compared to that of other individuals of the same age. As with the other measures, ratings were made on a 5-point scale ranging from Excellent (1) to Very poor (5) (Eriksson et al., 2001).

Symptom reports

Physical symptoms were measured by a modified version of the QPSNordic (Dallner et al., 1999) that covers nine general symptoms (heartburn, nausea, stomach ache, palpitations, coughs, colds, and headaches, extreme fatigue and sleep disturbances), and the Standardized Nordic Questionnaire (Kuorinka et al., 1987) that includes musculoskeletal symptoms or pain in the neck, shoulders, upper back, lower back, hand/wrist and leg/knees/feet. For each symptom, respondents were asked to indicate whether they had experienced any symptom or pain during the past 6 months. For both measures, sum scores were computed with high scores indicating more symptoms.

Interference between work and home

WHI, that is, whether working life interferes with home life (WHI), was assessed by two items (Frone et al., 1992). Both items were rated on a 7-point scale ranging from Very seldom (1) to Very often (7) and the ratings were added into a WHI-score with high scores indicating high interference. The internal consistency coefficients varied between 0.65 and 0.75 for the three measurements.

Psychosocial work characteristics

The measures of psychosocial work characteristics (Karasek, 1979; Karasek & Thorell, 1990) covered four factors: job control (2 items), job demands (5 items), skill discretion (4 items) and social support (6 items). All items were rated along a 4-point scale ranging from No, almost never (1) to Yes, very often (4) for all factors but social support that ranged from No, completely disagree (1) to Yes, completely agree (4). As suggested by previous research (Schreurs & Taris, 1998), the four factors were analyzed separately. For all factors, sum scores were computed with high scores indicating poor conditions. The internal consistency coefficients varied between 0.67 and 0.73 for job control, 0.72 and 0.75 for job demands, between 0.34 and 0.58 for skill discretion and between 0.82 and 0.87 for social support for the three measurements.

Statistical analyses

Baseline group differences (intervention group vs. referents) were checked using t tests for independent measures or χ² tests where appropriate. Repeated measures of analysis of variance (ANOVAs) were then performed to compare the intervention group with referents on all outcome measures. For self-ratings, the analyses included all three points in time. Since biomarkers and all other measures originating from the health checkup, show consistent seasonal variations (Garde et al., 2000; Ockene et al., 2004; van Anders et al., 2006), only data from T1 and T3, collected during the same season, were included in the statistical analyses. Significant effects were followed by post hoc tests. Due to missing data, there are small variations (n = 1–3) in sample sizes between the analyses.

Results

Demographic description of the groups

Apart from the women in the intervention group being significantly older than the referents (t (85)=2.56, p < 0.05; intervention group: M = 46.6, SD = 8.9; referents: M = 40.3, SD = 12.0), no significant differences were found in the other demographic factors. The majority (62.5%) of the women in the intervention group were married or living with a partner. The corresponding figure among referents was 50.0%. Most of the women in both groups had children living at home (intervention group: 59.7%; referents: 53.6%). Also, the vast majority of the women in both groups had completed upper secondary school (intervention group: 92.4%; referents: 88.8%) while the others had a university degree.

Comparing the intervention group and the referents

Tables 1 and 2 include descriptive statistics for self-ratings and biomarkers for the different study phases in the intervention and reference groups and summarizes results from the analyses that were performed to compare the two groups on all outcome measures. Initial analyses of baseline group differences for all outcome measures, including study participants taking part in all study phases, showed significant differences in DHEAS only (t = −3.15, p < 0.01), with referents having significantly higher levels (Table 2).

Table 1 Descriptive statistics for self-ratings for the intervention and reference-group before, after 6 months and after 12 months, together with results (F values and p values) of repeated measures ANOVAs.

	Intervention group						Referents
	Before		6 months		12 months		Before		6 months		12 months
Measure	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	Ftime	Fint
Health-related measures
Perceived stress	39.6	7.8	39.8	7.3	39.3	8.1	41.2	7.4	44.5	6.5	45.2	7.8	1.71	1.94
Recovery	9.8	3.2	9.5	3.1	10.2	3.1	10.9	3.0	11.4	2.7	11.5	2.8	1.78	2.75
Fatigue	13.8	2.9	13.3	3.0	13.6	2.9	14.5	3.0	15.2	2.8	14.6	3.2	0.09	1.64
General symptoms	3.4	2.1	3.9	2.2	3.3	2.2	4.4	2.2	4.5	2.0	4.5	2.4	1.17	0.78
Musculoskeletal symptoms	3.6	1.6	3.4	1.8	3.8	1.8	3.6	1.9	3.4	1.6	2.8	1.7	1.56	4.32*
Self-rated health	2.2	0.9	2.3	0.8	2.3	0.8	2.4	0.7	2.6	0.6	2.5	0.8	0.20	0.80
Work-related measures
Job control	4.6	1.5	4.6	1.4	4.6	1.5	5.2	1.5	4.9	0.9	4.8	1.4	0.52	0.54
Job demands	15.5	2.6	15.9	2.4	16.1	2.2	15.6	1.8	16.9	1.9	16.2	2.1	3.32*	1.02
Skill discretion	11.1	1.7	11.1	1.6	11.3	1.7	11.3	1.5	11.4	1.5	11.2	1.2	0.10	0.50
Support	18.7	3.2	18.5	3.1	18.6	3.0	16.9	2.6	15.9	2.6	17.5	3.7	3.25*	1.89
Work/home interference	5.7	3.7	4.7	3.3	5.5	3.4	6.8	2.1	7.2	3.3	7.6	3.2	0.80	1.16

F_time: F value main effect of time; F_int: F value interaction time × group.

*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Table 2 Descriptive statistics for biomarkers for the intervention and reference-group at the first measurement, after 6 months and after 12 months, together with results (F values and p values) of repeated measures ANOVAs including the first and final measurements.

	Intervention group						Referents
	Before		6 months		12 months		Before		6 months		12 months
Measure (unit)	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	Ftime	Fint
Blood lipids
Total cholesterol (mmol/l)	5.2	1.1	5.3	1.1	5.5	1.1	4.8	1.2	5.1	1.2	5.2	1.5	23.82****	0.04
Triglycerides (mmol/l)	1.1	0.5	1.1	0.6	1.2	0.4	1.2	0.8	1.2	0.6	1.4	0.6	7.39**	0.50
HDL (mmol/l)	1.7	0.4	1.8	0.5	1.8	0.5	1.6	0.2	1.6	0.2	1.7	0.3	24.72****	0.11
LDL (mmol/l)	3.0	0.9	3.0	0.9	3.1	0.9	2.8	1.0	2.9	1.0	2.9	1.1	4.99	0.02
Cardiovascular measures
SBP (mmHg)	117.9	18.4	120.7	20.3	117.7	18.8	108.4	14.9	114.3	11.7	111.1	11.6	1.26	1.39
DBP (mmHg)	79.3	12.6	83.6	13.4	78.7	11.4	73.4	12.6	79.6	10.8	78.0	11.1	4.74*	7.71**
Heart rate (beats/min)	66.5	10.6	66.6	9.1	67.2	10.9	65.7	13.1	67.8	11.9	69.0	12.4	3.43	2.28
Metabolic markers
Glucose (mmol/l)	5.2	1.0	5.0	0.7	5.4	1.3	5.6	1.8	4.6	1.3	6.0	2.3	5.51*	0.45
HbA1c (mmol/l)	4.5	0.5	4.4	0.5	4.5	0.5	4.8	0.7	4.7	0.9	4.7	0.9	23.90****	11.25***
Waist/hip ratio (cm)	0.85	0.07	0.85	0.1	0.83	0.07	0.82	0.1	0.85	0.1	0.80	0.07	4.50*	0.02
Neuroendocrine markers
DHEAS (mmol/l)	4.1	1.8	4.4	2.1	4.4	2.1	6.2	3.8	5.7	3.8	6.9	4.4	15.47***	1.06
Prolactin (µmol/l)	8.6	5.6	9.9	7.7	9.5	6.0	6.6	2.2	7.5	3.5	7.1	3.3	1.14	0.10

F_time: F value main effect of time; F_int: F value interaction time × group.

HDL: high-density lipoproteins; LDL: low-density lipoproteins; SBP: systolic blood pressure; DBP: diastolic blood pressure, HbA1c: glycosylated hemoglobin, DHEAS: dehydroepiandesterone sulfate.

*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

For self-reports (Table 1), there was a significant interaction effect (partial eta²=0.06) for musculoskeletal symptoms showing a significant decrease among referents between T1 and T2 (p < 0.05), but no significant differences in the intervention group (Table 1). For self-ratings of psychosocial work characteristics, no significant interaction effects emerged. However, there was a small (partial eta²=0.04) but significant time effect for job demands with ratings at T1 being significantly lower than those at T2 (p < 0.05) and T3 (p < 0.05) in the whole group.

As for the biomarkers (Table 2), significant interaction effects emerged for DBT and HbA1c. The interaction effect for DBT showed that levels in the intervention group were fairly stable over time, while there was an increase in the reference group from the first to the final measurement. In contrast, there were no significant changes in the other cardiovascular measures. However, in both groups mean levels of DBT, SBT, and HR were within the healthy range at all points of measurement.

For HbA1c, there was a significant decrease among referents between the first and the final measurement while levels in the intervention group were fairly stable over time. For the other metabolic measures (Table 2), there was a significant increase in glucose levels in the whole study group from the first to the final measurement. This change was paralleled by a significant decrease in WHR. However, the mean values for HbA1c, glucose and WHR reported in Table 2 are all within the healthy range.

Additionally, significant time effects emerged for TC, triglycerides, HDL, LDL, and DHEAS. In respect of blood lipids, TC levels increased significantly over time with levels at T3 being significantly higher than those at T1. A similar pattern emerged for triglycerides, HDL and LDL. However, for all blood lipids, mean values reported in Table 2 are within the healthy range.

As for neuroendocrine markers, levels of DHEAS also increased significantly over time (Table 2) and this increase remained when taking into account the significant differences between groups in baseline levels (analysis not shown). In contrast, there was no significant changes in prolactin. However, in both groups, mean levels of DHEAS as well as prolactin were within the healthy range at all time points.

In sum, the comparisons between the intervention group and the referents showed few significant interaction effects (musculoskeletal disorders, DBT and HbA1c). However, significant changes were found among referents but not in the intervention group. Positive effects including increasing levels of HDL and DHEAS and decreasing WHR were found in the whole study sample. Similarly, findings including negative effects characterized by increasing levels of blood lipids and glucose, also emerged in the whole study sample.

Discussion

Contrary to the initial hypotheses, the present study on RWH in a municipal eldercare organization showed no clear health promoting effects or other consistently positive or negative effects of RWH on the broad range of outcome measures included. This was despite the RWH including activities supporting health promotion. Specifically, comparisons of the intervention group and the referents resulted in two significant interactions (DBT and HbA1c), with further examination of these interactions showing significant changes among referents (Table 2) but not among the women taking part in the intervention.

As for self-reports, the results showed reduced levels of musculoskeletal symptoms among referents but not in the intervention group. Additionally, the present study showed no significant group differences in recovery, fatigue or WHI. This means that the present results run contrary to previous research (Åkerstedt et al., 2001; Anttila et al., 2005; Wergeland et al., 2003), which has shown some effects of RWH on comparable self-reported outcome measures in similar groups of employees. However, in contrast to the Wergeland et al. (2003) study, which assessed pain in a specific area, the present study used a composite measure of musculoskeletal symptoms. Yet, separate analyses (not shown) of the different body areas included in the composite measure produced no significant findings. Similarly, the other self-report measures used in the present study do not correspond exactly to the ones used in previous research but it should still be possible to compare conceptually the overall findings for similar concepts: for instance, the work–family conflict findings reported by Anttila et al. (2005) can be compared to the WHI measure included in the present study.

In the present study, most positive and negative effects were time-related and emerged in the whole study sample. When the analyses were rerun separately for the two groups, results (not shown) were in line with those from the repeated measures ANOVAs comparing the two groups over time and, consequently, we chose to report the results from the between-groups ANOVAs. These results show the value of a study design including a reference group, albeit smaller than the intervention group but still consisting of women having the same work tasks and being employed within the same municipal eldercare organization in the same geographical area (cf. Kristensen, 2005; Ruotsalainen et al., 2006; Scharf et al., 2008). Thus, despite limiting generalization to other occupational settings, including such a realistic group of matched referents sharing the overall organizational milieu while investigating a broad range of health-related indicators, is an addition to the existing research on RWH (cf. Brynja & Bildt, 2005; Schiller et al., 2017; Voglino et al., 2022).

Besides including referents, the present study was based on a repeated measures design consisting of three phases. This allows comparing individuals to themselves over time, which increases power even in smaller samples (cf. Tabachnik & Fidell, 2006). Also, each study phase included assessment of both self-reports and biomarkers. Biomarkers are objective measures not influenced by any shared response bias that may increase during the concurrent assessment of self-ratings of work characteristics and health-related measures (e.g. Hurrell et al., 1998). While the broad range of measures can be seen as a strength, it is also a potential drawback since the number of measures increase the number of statistical tests performed. But, in relating the overall pattern of findings to the individual results for each measure, the risk of overestimating results for a single measure may be avoided. However, there are potential caveats in repeated measures of biomarkers over time since seasonal changes may influence the results (Garde et al., 2000; Ockene et al., 2004; van Anders et al., 2006). In the present study, such seasonal effects were reduced by the first and third study phases being performed at the same time of the year and with the statistical analyses of biomarkers targeting specifically these two phases.

Characteristics of the study group, including age, gender, work schedules along with sleeping patterns and breaks at work, as well as life outside work may have influenced the present findings. Despite referents being younger, there were no consistent and statistically significant differences between the RWH and referents in the study variables before the intervention. This is of particular importance for biomarkers where age-related differences may be an issue. Yet age, or other confounders, may explain the differences seen in DHEAS-levels only. While the homogeneity in age and gender may contribute to the overall lack of significant effects, other factors including potential changes in sleeping patterns and breaks at work, various life events, menopausal status, and health status may have influenced the findings. Variations in menopausal status within women are not likely to have had any major effect on the results since most women reported that they retained their status throughout the study period. However, changes in sleep and breaks during work were not monitored meaning that their specific effects remain unknown. Yet, no significant effects emerged for prolactin, which is sensitive to sleep and stress, suggesting that there were no effects on sleep and stress. In relation to what can be expected in healthy working populations, most women rated their health as good or fairly good. Moreover, at a group level, the descriptive statistics for the biomarkers included in Table 2, show that levels were within the healthy range, and not clinically significant. Overall, such a homogeneity with respect to the different health-related indicators may provide little room for improvement in relation to the intervention. In particular, this may be the case in smaller samples as the one studied here. This means that the smaller sample size and the fact that the study sample included an overall healthy group, may explain the lack of health-promoting effects of RWH.

Moreover, the lacking effects of the reduction in working hours may be explained by the fact that all employees taking part in the intervention were aware of the intervention being restricted in time. Information about the intervention may also have influenced the measures before the change took place (anticipatory effects). Knowing that they would return to their ordinary working hours when the study period ended may have induced feelings of distress, which, in turn, may have influenced the results. The lack of effects of RWH reported in the present study may also be due to the intervention being too short, not allowing the disclosure of potential long-term effects and, more importantly, not allowing employees to adapt properly to the new conditions (e.g. including changes relating to sleep and stress). However, this seems unlikely considering that health-promoting activities facilitating adaptation (i.e. study circles) formed part of the intervention. Also, other comparable interventions involving RWH have been restricted in time (e.g. Schiller et al., 2017; Wergeland et al., 2003), but still shown some positive effects. Other explanations relating to the lack of effects involve the fact that in requiring employees to work fewer hours, RWH interventions allow more time for unpaid work at home. Spending more time on household tasks means that the total workload, including both paid and unpaid tasks, may have stayed the same.

Similar to previous studies of RWH (Malmberg et al., 2003; Voglino et al., 2022), all additional costs including retained full-time salaries and the recruitment, training and employment of new workers that are associated with the intervention were financed by subsidies from the municipality. These subsidies were temporary, spanning over the study period, and were necessary since cuts in salaries would not be accepted by the unions or employees. Thus, the present results give no information on what would have happened if the employer and/or the employees would themselves have had to finance the costs of RWH.

In showing no direct effects on the psychological or physiological status of the women in the intervention, the present study lends no consistent support to any of the assumed health-related benefits of RWH. However, it is unclear whether any positive effects of RWH that are linked to reduced exposure to risk factors have been counteracted by simultaneous reductions in the positive aspects of work, or whether the reduction from 7 to 6 h/day was too small to have positive effects across the different health-related outcomes. The overall findings may seem surprising when considering that this RWH-intervention also included study circles providing employees with opportunities to share experiences and ideas regarding health promotion. Perhaps this intervention component needs further development to properly promote health. In addition to studying different intervention components, it is important that future research collect data on what individuals actually do and prefer to do with their extra leisure time: some individuals may prefer to spend more time together with their families, others choose to exercise while a third group may opt for other activities or even choose to work full-time (Voglino et al., 2022). Importantly, knowing how individuals use the extra time is crucial for understanding the variation of health-related effects of RWH; a study within dentistry showed that the direct effects on health of a 2.5-h reduction of weekly work hours to be, at best, unclear when compared to other interventions (physical activity) and referents (von Thiele Schwarz et al., 2008). Ideally, subgroup analyses would allow for further in-depth analyses. Such secondary sub-group analyses would allow the identification of groups of individuals with, for instance poorer health, to distinguish potential effects among specific groups. Unfortunately, the sample size of the present organizational intervention group restricted such analyses. Obviously, larger samples, including additional organizations, would allow more powerful and fine-grained analyses of the ways in which RWH may promote health in different organizational settings as well as in relation to individual characteristics, choices, and preferences. This may also allow developing the theoretical rationale explaining why working fewer hours may have effects among employees in some organizations but not in others. Moreover, this may provide further knowledge regarding potentially negative health-related effects not identified in the current study.

To conclude, the present findings showed no consistent health-related effects of reducing the workday from about 7 h/day to about 6 h/day among women employed by the municipal eldercare, despite the 6-h workday adding structure and including health-promoting components associated with the intervention.

Acknowledgements

We are indebted to all the employees who volunteered to participate in this study. We also thank the nurses at the occupational health company who carried out the health checkups.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The data that support the findings of this study are available on reasonable request from the corresponding author. The data are not publicly available due to participants of this study not agreeing for their data to be shared publicly as this would potentially compromise the privacy of research participants.

Additional information

Funding

This research was supported by AFA Insurance in Sweden and by FORTE.