Millennials: sickness presenteeism and its correlates: a cross-sectional online survey

Abstract

Objective

Sickness presenteeism (SP) is a well-documented phenomenon in the current workforce. However, little is known about the SP of future employees (Millennials). We investigated rate and propensity of presenteeism and health-related and work-related correlates in university students to obtain information about the relevance of SP in the future workforce. Sickness presenteeism (SP) refers to going towork while ill.¹

Design and setting

We administered a cross-sectional online survey in August 2018 involving self-assessments of health-related and work-related behaviour. As planned, we analysed rates, instances and propensities of presenteeism and absenteeism. The target population was Millennial university students.

Results

749 German university students, average age 24.7±3.6 years, 49.5% women, mean academic progress 4.5±2.7 semesters, <1% without work experience, 15.4±10.3 mean weekly working hours. Presenteeism rates in the complete sample were 64% at school and 60.4% at work. For the subsample of participants who reported sickness, presenteeism rates were 87.9% at school and 87% at work. Presenteeism propensity among them was 0.67±0.34 at school and 0.68±0.35 at work. SP at school and at work was significantly correlated with subjective health (r=0.29, p=0.000 and r=0.25, p=0.000) and locus of control (r=−0.20, p=0.000 and r=−0.26, p=0.000). Age and sex adjusted multiple regression analysis showed that subjective health explained 9.7% of the variance in SP at school (B=0.049, p=0.000) and 8.1% of the variance in SP at work (B=0.037, p=0.000).

Conclusions

The SP rates of the future workforce were in line with previous research on older employees, whereas the presenteeism propensities we found here were higher. Across all analyses, subjective health proved to be associated with presenteeism. Thus, the relevance of SP will remain constant or might even increase with regard to the future workforce. Programmes offering health education should be implemented to sensitise this generation as early as possible.

Strengths and limitations of this study.

• One of the first studies to investigate sickness presenteeism among individuals constituting the future workforce (Millennials).

• Analyses distinguish between individuals with health problems and those without.

• The study focuses not only on the presenteeism rate but also on presenteeism propensity.

• The cross-sectional data do not allow for causal conclusions.

• Data are based on self-assessments.

Introduction

Sickness presenteeism (SP) refers to going to work while ill.¹ The study of SP has gained in importance during the past two decades due to its association with sickness absenteeism, productivity, health and attitudes.² Although the vast majority of empirical studies of the phenomenon are cross-sectional surveys,³ important consequences of presenteeism for the organisation and the individual have been established in longitudinal research. Studies have reported negative effects for individual health.^4–12 Other negative effects of SP for the individual are productivity loss¹³ and a decline in work attitudes.^{14 15} Considering consequences for the employing organisation, there is empirical evidence for uncovered costs caused by productivity loss.¹⁶ SP has been reported independently of a variety of sociodemographic variables^{1 17–23} and for various kinds of professional groups.17 19 24–29 SP is a global phenomenon. Its prevalence has been documented in many countries in Europe,17 25 29 30 the USA,³¹ Canada,^{32 33} Asia³⁴ and the Middle East.³⁵ The presenteeism rates reported for study samples vary from 30% to >90%.

Due to the wide range of negative effects of SP for individuals and organisations as well as the global ubiquity of SP, the study of SP seems advisable. However, although SP rates are well documented for the current workforce, the majority of which belong to the generations of Baby Boomers and Generation X,^{36 37} empirical evidence concerning the SP behaviour of the future generation of workers is scarce.^38–40 Matsushita and colleagues³⁹ developed an instrument to measure presenteeism among students, but they did not report on presenteeism behaviour. Johansen³⁸ analysed motives for the presenteeism of students in secondary school. To our knowledge, the study by Kötter et al ⁴⁰ is the only one to date to investigate the presenteeism behaviour of students. They did not find any differences in prevalence rates between medical students and science, technology, engineering and mathematics students and reported higher levels of presenteeism for female than for male students. However, the prevalence rate they reported could not be compared with other studies because their way of measuring and aggregating presenteeism behaviour was uncommon. Current university students are predominantly classified as Millennials,^{36 41} who are said to have different attitudes towards work than older generations. In particular, Millennials have greater entrepreneurial interests,⁴² and they hold a ‘work-to-live’ attitude with a higher priority on work–life balance^{41 43} than previous generations. Especially, this latter trend suggests that presenteeism among Millennials should be lower than among older generations. In general, interests and work values are relatively stable³⁷ and are associated with behaviour. Thus, knowledge about the SP-related behaviours of current students should be useful for evaluating the significance and consequences of such behaviours for individuals, organisations^44–52 and society in the future.^53–57 Specifically, we wanted to know how common presenteeism is among students belonging to the generation of Millennials and whether they favour presenteeism over absenteeism when they are sick. We were particularly interested in the question of whether they differ in their behaviour with regard to study and work times and whether their behaviour is associated with other health-related variables. To investigate this, we conducted an online survey.

When studying presenteeism, two methodological issues need to be borne in mind: previous studies have usually not differentiated between individuals suffering from health events—who are thus able to show SP—and individuals without health problems—who per definition cannot show SP behaviour.⁵⁸ Therefore, these studies have provided information about the prevalence of SP in the workforce but not about the prevalence among sick employees. In addition, it is necessary to distinguish between SP rates and SP propensity.⁵⁹ SP rates are measures of the frequency of attending work while ill. They depend on the number of health events and are an indicator of health status or vulnerability to illness.⁶⁰ SP propensity is the percentage of days worked while ill over the total number of days that the person was sick,³² and it reflects an individual’s probability of opting for sickness presence rather than sickness absence in the case of illness.⁶⁰ Thus, it offers additional information with regard to the decision-making process of the individual. Both aspects were considered in the current study.

Methods

Study design and population

Based on Krejcie and Morgan’s⁶¹ specifications for the calculation of a representative sample, we calculated a required sample size of 384 by taking into account a CI of two SDs, the degree of accuracy of 0.05 and a total population of 2.9 million students in Germany.⁶² During the 2018 summer term, we asked 15 random German institutions of higher education if they would allow us to send our survey to their students. Of these, four institutions agreed to forward the survey. In the email that we sent to the students, we asked participants to forward the survey to other potentially interested persons, which led to a further distribution of the survey and thus in total, students from nine German institutions of higher education took part in the online survey. Participants were informed that the survey was anonymous and voluntary and that they could withdraw their participation at any time. Of the 1291 individuals who began taking the online survey, 791 provided complete data.

Measures of presenteeism and absenteeism

Because many students do not only study full time but are also employed in paid part-time jobs to contribute to their living, we collected data on presenteeism and absenteeism separately for school and work. To measure SP frequency, students were asked how often during the past 12 months they went to classes when the state of their health would have justified staying home.^{1 9 63} The same question was posed for going to work. Whereas fixed response categories such as ‘never’, ‘once’, ‘2 to 5 times’ and ‘more than 5 times’ have been used in other studies,1 30 63 64 we used an open-question format. The reason for this was that the former is seen as too crude for such a low base rate behaviour and might convey information about what is perceived as normal and might thus bias the answers.⁶⁵

Sickness absence (SA) was also measured with two items asking students to indicate how often during the past 12 months they stayed home due to sickness instead of going to classes/work.

Health-related and personality-related factors

As potential correlates of SP, we collected subjective measures of individual health, locus of control, professional self-efficacy and extent of paid work. Because in Germany health-related issues and objective health status are very sensitive topics, we decided to use a subjective measure of individual health, which is known to be a robust predictor of several health outcomes, such as functional ability, healthcare utilisation and morbidity.⁶⁶ Locus of control is ‘a fundamental individual difference variable that reflects individuals' beliefs about the degree of control they have over events in their lives’ (Galvin et al, p. 820).⁶⁷ People with an internal locus of control are convinced that their fate depends on their decisions, efforts and behaviours. Individuals with an external locus of control believe that their lives are controlled by external forces such as powerful others, chance or luck. The construct is often applied in the area of health psychology.^68–73 Self-efficacy is one’s belief in one’s competence to perform intended actions. Efficacy beliefs influence the self-regulation of motivation through goal challenges and outcome expectations and the types of activities people choose.⁷⁴ Self-efficacy and locus of control (especially, health locus of control) are seen as personal health resources that determine health-related behaviour.⁷³ Both concepts have repeatedly been used in studies of the health-related behaviours of students.^{73 75–78} In addition, we thought the extent of paid work would be a relevant correlate because it is an indicator of workload, which is known to be related to presenteeism.³

With regard to students’ health status, we used two variables based on a sequence of subjective assessments. Subjective health status was calculated as the mean of two items (‘How do you rate your general susceptibility to disease?’; ‘How do you rate your health in comparison with your age group?’). The scale was a slider ranging from 0 (‘very low’; ‘very bad’) to 100 (‘very high’; ‘very good’). We used the 4-item short scale by Kovaleva et al ⁷⁹ to measure locus of control (eg, ‘When I really try, I will succeed’) and three items from the short scale of occupational self-efficacy⁸⁰ (eg, ‘I am not sure I have all the abilities necessary for my job’). Both locus of control and occupational self-efficacy are stable characteristics that are relevant to presenteeism.^{65 81 82} Response categories for the seven items ranged from 1 (‘does not apply at all’) to 5 (‘fully applies’).

Employment and demographic variables

Participants stated whether they had worked before they started their university programme, and they were asked if and how many hours they worked while attending school and during semester breaks. The survey asked for demographic information on gender, age, study major, pursued degree and name of university. Employment and demographic variables were measured with one item each.

Data analysis

The number of health events was calculated as the sum of the SP and SA frequencies.^{33 60} We computed presenteeism propensity as SP frequency divided by the number of health events for university and work settings separately. Hence, SA propensity was 1-SP propensity.⁶⁰ Descriptive statistics (means, percentages and SDs) were initially used to tabulate results. Because the variables were not normally distributed, we used nonparametric tests (ie, the Wilcoxon test for dependent samples and the Mann-Whitney U test for independent samples) to analyse group differences and Spearman’s rho correlation coefficient to describe associations between variables. Bivariate correlations were calculated between participants’ sickness, presenteeism rate, presenteeism propensity, subjective health, locus of control, occupational self-efficacy and number of hours spent working per week. Age and sex adjusted multiple linear regressions using ordinary least squares were computed to test the influence of the independent variables subjective health and locus of control (selection method: Enter; consider both variables) on the dependent variables presenteeism at school and at work. We used adjusted R² for variance explained and report regression coefficients and standard errors. For all analyses, two-sided p values <0.05 were considered statistically significant. We performed all analyses with IBM SPSS (V.24, Armonk, New York, USA). Only participants from the target population who completed the survey were included in the analyses. However, it was not compulsory to provide an answer to each question because some of the questions did not apply to all participants (eg, those referring to working while studying). Thus, analyses of subgroups differed in sample sizes. Participants who had missing data on one of the investigated variables were excluded from the respective analysis.

Ethics approval and informed consent of participants

Before beginning the study, participants were informed that their participation was voluntary and that they could withdraw their participation at any time while working on the questionnaire. They were informed that no personalised data would be collected, that their data would be used for scientific purposes only and that it would be stored and analyzed anonymously so that it would not be possible to establish any connection between them and their answers.

Patient and public involvement

Neither subjects nor public were involved in the study.

Results

A total of 791 questionnaires were completed. In accordance with the goal of the study, the analyses were restricted to participants belonging to the generation of interest, the Millennials, which include people who were born in 1981 or later.^{36 41} Further, cases of long-term absence or presence were excluded from the analyses because these were likely to have been caused by different modalities.⁶⁰ As was done in other studies, we defined 60 days as the threshold.^{60 83} More than 99% of respondents reported fewer than 60 days of SA or sickness presence. Thus, the final sample comprised 749 participants (49.5% women) between the ages of 19 and 37 years with an average age of 24.7±3.6 years. Their mean academic progress was 4.5±2.7 semesters. Less than 1% of participants had no work experience, and the mean number of hours spent working per week was 15.4±10.3. For a summary of the demographic variables, see table 1.

Table 1.

Descriptive data of study participants. Number of incidents and percentages

	All study participants			Considered participants*
	n	Mean±SD	%	n	Mean±SD	%
Total number (N)	1773		100	749		100
Mean age (years)		24.87±4.70			25.70±3.61
Female	672		37.9	371		49.5
Mean age (years)		24.68±5.30			24.25±3.51
Male	703		39.6	368		49.1
Mean age (years)		25.21±4.37			25.17±3.64
Desired degree and current semester
Bachelor	1087		61.3	580		77.4
1	87		8.0	19		3.3
2	300		27.6	161		27.8
3	33		3.0	20		3.4
4	249		22.9	129		22.2
5	37		3.4	22		3.8
Six or more	363		33.4	225		38.8
Master	305		17.2	166		22.2
1	38		12.5	18		10.8
2	73		23.9	38		22.9
3	37		12.1	20		12.0
Four or more	149		48.9	90		54.2
Course of study
Computer science	193		10.9	109		14.6
Technology	223		12.6	120		16.0
Engineer	132		7.4	75		10.0
Architecture	70		3.9	37		4.9
Business administration	166		9.4	87		11.6
Psychology	140		7.9	64		8.5
Social work/Education	79		4.5	49		6.5
Legal studies	15		0.8	10		1.3
Design	8		0.5	2		0.3
Journalism	17		1.0	12		1.6
Mathematics/Chemistry	23		1.3	18		2.4
Communication	41		2.3	17		2.3
Media	51		2.9	27		3.6
Other study programmes	37		2.1	18		2.4
Not specified	578		32.6	104		13.9

*Participants included in the analyses were those who completed the questionnaire and belonged to the generation of Millennials.

Sickness

The descriptive results for the number of days of sickness in the past 12 months can be found in table 2 and the correlations with other factors in table 3. It turned out that participants who worked while attending school reported a significantly larger number of days of sickness during school time than during work time. Further, the correlation between days of sickness during school time and days of sickness during work time was significant (Rho=0.60, p=0.000). Furthermore, days of sickness during school time and during work time were both significantly correlated with subjective health, locus of control, occupational self-efficacy and for work time, also with number of hours spent working per week. With regard to gender, women reported significantly more days of sickness during school time and during work time than men. 27.2% of the participants did not report any health event during the last 12 months during school time. Concerning work days, 30.3% reported no health events.

Table 2.

Means for days of sickness in the complete sample

Average days of sickness	School time		Work time		Analysis of means
	n	Mean±SD	n	Mean±SD
During school time	749	4.49±6.54	671	4.31±6.51	*
During work time	671	4.31±6.51	671	3.56±5.65	School time versus work time for working students: Z=3.586, p=0.000
Female	371	4.99±6.82	326	4.29±6.29	Female versus male for school time: Z=−3.571, p=0.000 work time: Z=−3.249, p=0.001
Male	368	3.96±6.27	336	2.90±4.93

Note. Z: Wilcoxon test for dependent samples and Mann-Whitney U test for independent samples.

*Since not all subjects work during their studies, an inferential statistical comparison is forbidden.

Table 3.

Correlations for days of sickness in the complete sample

Correlations	School time		Work time
	Rho*	P value	Rho*	P value
Subjective health	0.37	0.000	0.33	0.000
Locus of control	−0.19	0.000	−0.12	0.005
Occupational self-efficacy	−0.14	0.000	−0.05	0.207
Number of working hours	−0.003	0.425	0.26	0.000

*Spearman’s Rho for correlations of not normally distributed variables.

SP frequencies and rates

Referring to the complete sample, as is typical for the reporting of SP rates,1 9 23 30 the descriptive values are reported in table 4 (left side). For the further exploration of SP behaviour and its correlates, the analyses were restricted to participants who reported at least one incident of sickness. The reason for this is that only individuals who reported some kind of illness during the past 12 months were able to show presenteeism (see table 4, right side). As results show, 72.8% of the sample were left for analyses regarding SP at school (n=545) and 68.6% with regard to work (n=468). Therefore, after excluding ‘healthy’ participants, the SP rates rose to 87.9% during school time and 87% during work time.

Table 4.

SP rates and frequencies for participants without and with incidents of sickness

	Complete sample				At least one incident of sickness
	n	SP rates	Mean incidents SP ±SD	Mean incidents SA±SD	n	SP rates	Mean incidents SP±SD	Mean incidents SA±SD	Analysis of means
									Z*	P value
School time	749	64%	2.94±5.00	1.52±2.99	545	87.9%	4.03±5.47	2.09±3.34	−9.582	0.000
Work time	682	60.4%	2.36±4.32	1.19±2.38	468	87%	3.41±4.86	1.69±2.70	−9.254	0.000

*Wilcoxon test for dependent variables.

SA, sickness absence; SP, sickness presenteeism.

The difference between SP incidents and SA incidents during school time was significant. When experiencing illness during school time, 39.6% always chose to go to classes, whereas 12.1% never did. With regard to work, the pattern was quite similar: SP incidents during work time differed significantly from SA incidents. 41.9% always chose to go to work in spite of illness, and 13% never did.

Correlates of SP

With regard to SP at school, we found a significant correlation with SP at work (Rho=0.49, p=0.000). Further correlations for SP during school and work time are listed in table 5.

Table 5.

Correlates of sickness presenteeism among the participants with incidents of sickness

	Correlations
	SP during school time			SP during work time
	n	Rho*	P value	n	Rho*	P value
SA during school time	545	0.14	0.001
SA during work time				468	0.11	0.022
Subjective health	466	0.29	0.000	394	0.25	0.000
Locus of control	470	−0.15	0.001	398	−0.17	0.001
Occupational self-efficacy	469	−0.15	0.001	398	−0.12	0.013
Number of working hours	545	0.01	0.763	468	0.13	0.004

*Spearman’s Rho for correlations of not normally distributed variables.

SA, sickness absence; SP, sickness presenteeism.

In order to further investigate the two most important correlates, we computed multiple linear regressions using ordinary least-square method and variables selection method ‘Enter’ (consideration of both variables). First, we checked for whether the assumptions for the application of the method were fulfilled.⁸⁴ Data were collected through independent random sampling. Inspection of scatterplots of the independent variables against the dependent and unstandardised predicted values against studentized residuals indicated that the assumption of linearity between independent and dependent variables was met. In addition, the assumptions with respect to homoscedasticity and multicollinearity were met. Outliers were identified in two ways. First, all subjects with values above or below three SDs of studentized deleted residuals were eliminated from the analyses. Second, on the basis of the leverage index, all subjects with values above the critical value (ie, 2*p/n)⁸⁵ were excluded from the analyses. Thus, 419 subjects were left for the analysis of presenteeism at school and 323 subjects for the analysis of presenteeism at work (table 6).

Table 6.

Age and sex adjusted multiple linear regression models in sample of the participants with incidents of sickness

Variables	School time		Work time
	B (95% CI)	SE	B (95% CI)	SE
Subjective health	0.049 (0.033 to 0.066)	0.008	0.037 (0.021 to 0.054)	0.009
Locus of control	−0.580 (−1.326 to 0.165)	0.379	−0.375 (−1.155 to 0.405)	0.396
Sex	0.593 (−0.146 to 1.331)	0.376	−0.526 (−1.270 to 0.217)	0.378
Age	0.065 (−0.051 to 0.180)	0.059	0.074 (−0.042 to 0.191)	0.059
R2	0.105		0.093
Adjusted R2	0.097		0.081
F value	F(4,414)=12.203, p=0.000		F(4,318)=8.137, p=0.000

Models include the variables Subjective Health and Locus of Control.

Analyses are adjusted for sex (female=0, male=1) and age.

Ordinary least-square method. Variables selection method: Enter (consideration of all variables). Linearity, multicollinearity and homoscedasticity are given.

Durbin-Watson ratio: 2.054 (school time), 2.160 (work time).

B represents regression coefficient.

Analyses revealed significant models. Subjective health explained 9.7% of the variance of presenteeism frequencies for school time and 8.1% for work time, indicating that presenteeism frequencies were higher for participants with good subjective health in comparison with those with poor subjective health. In both models, the effect of locus of control was not significant.

Presenteeism propensity

Presenteeism propensity, which reflects an individual’s probability of opting for presenteeism instead of absenteeism when suffering a health event, was 0.67±0.34 at school and 0.68±0.35 at work. Presenteeism propensity at school was significantly associated with presenteeism propensity at work (Rho=0.52, p=0.000). Of course, there were significant correlations between SP frequency and presenteeism propensity; however, because the formula for presenteeism propensity contains SP frequency, it would not be productive to report them.

The probability of going to classes when sick was significantly higher than the probability of staying home (Z(544)=-10.042, p=0.000). With regard to work time, the pattern was quite similar: the probability of going to work while ill was significantly higher than the probability of staying home (Z(467)=-9.444, p=0.000). There were no gender differences in presenteeism propensity during school or work time.

Discussion

The aim of the survey was to obtain information about the presenteeism behaviour of studying Millennials who constitute the future workforce. While more than one-fourth of participants reported no incidents of sickness during the past year, those who reported at least some kind of illness reported a larger number of days of sickness during school time than during work time. SP rates for the complete sample were 64.4% at school and 60% at work and for those participants reporting at least one health event, 87.9% at school and 87% at work. Thus, the SP rates of the future workforce were in line with previous research on older employees. Presenteeism propensities were 0.67 for school and 0.68 for work and were thus higher than reported in other studies. SP at school and SP at work were highly associated. Further, SP was positively correlated with good subjective health and an external locus of control and negatively correlated with occupational self-efficacy. Across all analyses, including multiple linear regressions, subjective health was the variable with the strongest association with SP. Results of the study show that the relevance of SP will remain constant or might even increase with regard to the future workforce.

In this survey, SP rates for the complete sample were on an average level in comparison to a recent research among German employees that has reported great variability in rates ranging from 35%⁸⁶ to >90%.⁸⁷ The levels found here were in the same range as studies that were conducted with employees in the Netherlands,²⁸ Scandinavia⁸⁷ and other European countries.²³ The proportion of participants who reported no health event during the past year was higher than in other studies.⁶⁰ This higher rate might be due to the fact that the sample we investigated was younger than the usual workforce samples and the fact that age was positively related to health issues.^88–90 In our analyses, we followed Navarro et al’s⁵⁸ recommendation to separately identify the SP prevalence for the subsample of individuals who reported some kind of illness and were thus able to show SP. Hence, the rates increased by more than 35%–40%. This increase was lower than in the study by Navarro et al,⁵⁸ who reported SP rates of 23% for the complete sample and a rate that was more than doubled for those with health problems (53%). However, in our survey, SP levels were much higher. The distinction between SP rates for the complete sample and the subsample of those who experienced health problems provides answers to two separate questions. The first (the whole sample) indicates the economic significance of the problem, whereas the second (the subsample) can provide human resource management and corporate health management with relevant information as to how many employees need support and what kinds of measures might be helpful.^{90 91} Therefore, both methods of calculation are useful; however, in terms of health interventions, it is more appropriate to consider the part of the workforce reporting some kind of illness.^{58 90 91}

The presenteeism propensities for school and work were identical, which means that participants’ decision making did not differ between the two settings. Propensities showed that on average, in two out of three cases, participants who were sick chose to attend class or go to work and thus showed significantly more SP than SA. To date, only a few studies have explored presenteeism propensities.^{32 60} A Canadian study reported a propensity of 0.50 for employees,³² and an Austrian study reported a propensity of 0.59, also for employees.⁶⁰ The propensities found in the current study for students appeared to be slightly higher (0.67 and 0.68). This finding seems to contradict the focus on work–life balance that has been reported for the generation of Millennials.41 43 92 93 The high levels of presenteeism propensity might be explained by students who might be worried that they will not be able to keep up if they miss too many classes.⁹⁴ With regard to their jobs, the high level might be due to their work conditions, that is, most of them are probably paid on an hourly basis with the consequence that they will not earn anything if they call in sick. A more detailed view of the data revealed that the proportion of participants who always decided for SP when they were sick was more than three times higher than those who always chose SA. Millennials’ attributes and expectations offer further potential explanations for their high attendance rates.⁹⁵ Millennials favour close relationships at work⁹⁶ and prefer team work inter alia in order to avoid risks.⁹⁴ Furthermore, they are interested in the acquisition of skills.⁹⁷ Thus, their decision to embrace presenteeism might reflect their desire not to let down team members (either at school or at work) and to expand their competencies.⁸⁷ Future studies should investigate whether this high propensity is due to differences in ‘jobs’ (ie, studying full time and working part time versus working full time) or due to different generations. We suggest that a survey be administered to full-time employees who belong to the generation of Millennials.

The high correlation between SP at school and SP at work indicates that both kinds of behaviour are determined by individuals’ health status, which is reflected in the number of days they were sick and the positive correlation with their subjective assessment of their health. Miraglia and Johns³ found that individuals with poorer general health exhibited more presenteeism because the sheer number of health events was higher, a finding that was supported by the data in the current study. In addition, the positive correlation between SP and SA also found here has been well established,1 6 17 21 63 65 although it seemed particularly high in this study.

In our correlational analyses, SP was positively associated with good subjective health and an external locus of control and negatively correlated with occupational self-efficacy. The correlation between SP and subjective health status has been well established.⁶⁵ In addition, the age and sex adjusted multiple linear regressions with the two most important correlates (ie, subjective health and locus of control) revealed that subjective health is a significant predictor for SP in both contexts. This means the better participants perceive their health the more SP they report. This has to be taken into consideration in future measures to manage SP. There is only one earlier study we know of that investigated the relationship between health locus of control and presenteeism among employees in all age groups, and it found lower SP for people with an internal health locus of control.⁹⁸ Although we measured general locus of control, we found the same pattern: that is, a higher SP rate for individuals with an external locus of control. Research has shown that university students score higher on internal locus of control than on external locus of control⁷⁸ and that an internal locus of control and a high level of self-efficacy⁹⁹ are associated with academic achievement.⁷⁵ Thus, students with an external locus of control might worry that they will miss too much if they do not attend classes when they are ill. Self-efficacy has been postulated to be associated with SP.^{2 81} A study involving a sample of Chinese full-time workers established a positive correlation between SP and self-efficacy and found that self-efficacy could buffer the negative effect of SP on health.⁸² By contrast, in our study, participants with low occupational self-efficacy showed higher levels of SP. The difference in findings may have occurred because Lu et al ⁸² measured general self-efficacy, whereas we used occupational self-efficacy. In addition, research has shown that Millennials are high in self-efficacy,^{93 95} and thus, they might be convinced that they can manage to make up any work they miss when they miss school or work when they are sick. By contrast, participants with low occupational self-efficacy are not sure whether they have the necessary competencies to meet expectations and thus opt for SP in order to work as much as possible and meet their work and study goals. In the future, it would be interesting to investigate these contradictory results.

Strengths and weaknesses

Pertaining to the goal of the study to investigate SP among the generation of Millennials, a strength of the survey is that it provided information about the SP behaviour of the future workforce, a group that has rarely been investigated with regard to SP. In addition, to our knowledge, the study is one of the first to explore SP among students who belong to the generation of Millennials. Further, since many of the participants were not just studying but were also working part time, this study allowed us to assess SP behaviour in university and work settings at the same time and to compare them. However, it is obvious that the sample was restricted to students and thus does not allow generalisations about manifestations of SP behaviour among full-time employees in these birth cohorts. Additional advantages of the study are that we calculated propensities and made a distinction between individuals with and without health problems. In most studies to date, participants have been asked to quantify the number of times they have engaged in SP but not the number of times they have engaged in SA or the total number of health events they have experienced. Thus, the prevalence of SP could be reported only for the complete sample. Shortcomings of the survey pertain mostly to methodological issues: because we applied a cross-sectional design, we were not able to identify the direction of the influence between subjective health and SP. Another weakness is that all measures were self-reports, and thus, their reliability might suffer from common method bias and overreporting due to social desirability.⁶⁵ Due to missing data and the non-applicability of questions, some of the analyses were restricted to a much smaller number of participants than the total sample size. In addition, as in many other studies,³ participants were asked to recall incidents of sickness presence and absence over a 1-year period, and therefore, the data might be biased by false recollection.¹⁰⁰ However, because the self-reported measure of days of sickness¹⁰⁰ and the subjective measure of health status¹⁰¹ demonstrated a high level of correspondence with objective data, we used the self-report method as a practicable way to obtain the relevant data.

Conclusion

The study’s results are in accordance with the recent findings that the prevalence of SP is continuously rising,¹⁰² which means the issue will remain an important one. Further, they indicate that especially subjective health is associated with presenteeism behaviour of Millennials. SP and the associated risks are not restricted to the current workforce but are prevalent among those who are currently enrolled in a university. One should think about implementing health education concerning SP for students who represent the future workforce and soon-to-be managers in order to sensitise them to the management of SP before they join the full-time workforce. A future study should be conducted in order to compare the results of this survey of Millennials attending a university with full-time working Millennials. Further, presenteeism propensity is important for gaining insights into individuals’ decisions to engage in presenteeism or absenteeism when ill. With regard to methodology, future studies should make a distinction in prevalence rates between the complete workforce and those with health problems because these two rates provide answers to different questions.