Differences in Self-Rated Worker Outcomes Across Stress States: An Interim Analysis of Hybrid Worker Data

Abstract

Stress experiences can have dire consequences for worker performance and well-being, and the social environment of the workplace is a key contributor to worker experience. This study investigated the relationship between hybrid workers’ self-ratings of productivity, mood, and stress with perceptions of positive (eustress) and negative (distress) stress states. We hypothesized that self-ratings would vary across combinations of eustress and distress experiences and that these differences would differ based on the social context. Ecological momentary assessments (EMA) were used to obtain ecologically valid data at four data points each workday across a 4-month study period in a cohort of seven office workers. Findings aligned with the Yerkes–Dodson law, such that higher states of arousal were associated with greater self-perceived productivity, and higher stress magnitudes were found when distress existed. Compared to other states, eustress was associated with higher productivity in work-related activities and better mood across all activity types.

Introduction

Although typically perceived as a singular and negative concept, individual responses to stressors result in various states of arousal and subsequent consequences. When associated with pressure, overwhelming experiences, and negative emotions, the high state of arousal that can occur is best described as distress, a stress state consequential to mental and physical health (Duchaine et al., 2020; Gilbert-Ouimet, 2012; Leger et al., 2022). Alternatively, boredom is an extremely low state of arousal, which may be viewed as a lack of any perceived stress. Eustress, or positive stress, exists between these two arousal levels and provides motivation through appropriate challenges and opportunities. The Yerkes–Dodson Law suggests that extreme states of arousal (i.e., boredom and distress) are associated with low levels of productivity and the existence of eustress provides engaging experiences that positively impact worker performance (Yerkes & Dodson, 1908).

In the wake of the COVID-19 pandemic, the widespread adoption of hybrid work formats has led to many performance-impacting stressors due to both physical and social aspects of work, such as the burden of adapting to new communication formats and expectations (Fukumura et al., 2021; Kniffin et al., 2021). Unfortunately, little is known about effectively differentiating between distress and eustress in hybrid work settings or how these stress states affect performance and well-being as work contexts shift (Aldoney et al., 2023). With rapid advancements in environmental and wearable sensing, artificial intelligence, and other technologies, there are growing opportunities to examine factors and features associated with aspects of human performance and well-being within real-life contexts.

Ecological momentary assessment (EMA) methods have been widely used to measure stress and its effects (Yang et al., 2018). Indeed, a recent systematic review of smartphone-based EMA of well-being lists a need for future work investigating individual differences in overall patterns of well-being (de Vries et al., 2021). We previously validated EMA methods for investigating the effect of various stress states in a controlled laboratory setting, finding that trends in self-rated productivity and arousal across stress states generally agreed with the Yerkes–Dodson Law (Awada et al., 2024). The following paper describes the findings of an interim analysis of data obtained through an observational study translating our laboratory-based methods to ecologically valid, real-life office worker contexts. We examined emerging associations between stress states and self-ratings of productivity, mood, and general stress among a preliminary cohort of hybrid office workers and explored the influence of activity context as work-related versus non-work-related and social versus solo.

Methods

Study Design

Daily EMA data were obtained over 4-months with a cohort of seven hybrid office workers. The protocol for the observational study was approved by the university’s institutional review board, and all participants provided informed consent before participating.

Recruitment and Participants

Inclusion in the study required participants to work 30 hr a week at a computer workstation, plan to maintain full-time employment for 4 months, have a dedicated workspace to perform their job duties, and work at the university. Members of the research team distributed information about the study and a link to a screening survey to individuals within their network and encouraged sharing with additional qualified individuals. Following responses to inclusion criteria, the screening survey was used to obtain worker demographic and job-related information, including responses to the Job Content Questionnaire (JCQ), a validated assessment to categorize the decision authority (DA, e.g., control) and psychological demand (PD) of a person’s job (Karasek et al., 1998).

Initial recruitment efforts resulted in 40 interested participants. We aimed to maximize heterogeneity in this initial, preliminary cohort to explore the variability in random effects across different worker types for generating valid algorithmic stress detection models. Specifically, we first considered diversity in work roles and structure based on job title, JCQ ratings of authority and demand, and percentage of remote work, followed by diversity in worker demographics of age, race/ethnicity, and gender identity.

Data Collection

EMA surveys were distributed via SMS text messages using the Qualtrics distribution tool with Twilio integration to monitor text message delivery fidelity and failures. Data collection took place between June and November 2023. Participants received four EMA surveys each workday during the study period (i.e., Monday through Friday). The first and last surveys were scheduled to match each participant’s work start and end times. The two surveys distributed in the middle of the day were sent at randomly generated times between the first and last daily surveys. Surveys were not distributed on national holidays or days participants identified as being on vacation or off work.

Measures

Participants categorized their activity in the past 5 min as work-related or non-work-related and identified the social nature of the activity. For work-related activities participants selected among the options of “working on my own,” “one to one meeting (online/in-person),” or “group meeting (online/in-person).” The options for non-work-related activity included “I was alone,” “I was with one other person,” or “I was with two or more people.” Participants rated their experience of the current activity using the rating structure of the Valencia Eustress-Distress Appraisal Scale (VEDAS; Rodríguez et al., 2013). These two questions were phrased as, “Would you describe this activity as a source of opportunity?” and “Would you describe this activity as a source of pressure?” Participants selected from the options of 1 = Very definitely IS NOT, 2 = Definitely IS NOT, 3 = Generally IS NOT, 4 = Generally IS, 5 = Definitely IS, or 6 = Very definitely IS. Finally, using an anchored visual analog scale ranging from 0 to 100, participants rated perceptions of productivity (unproductive to productive), mood (bad to good mood), and stress magnitude (no stress to complete stress).

Data Analysis

Data Preparation

Responses to the social context of activity were coded into a binary variable. For the work-related context question, the option “Working on my own” was considered solo (Activity = 1), while options “One to one meeting” and “Group meeting,” either in-person or online, were considered social (Activity = 2). Using VEDAS ratings, we constructed “stress states” following the criterion applied by Awada et al. (2024). Ratings on the eustress (E) and distress (D) scales were converted to binary categories such that ratings of 1 to 3 were “is not” and ratings of 4 to 6 were “is.” The four possible combinations of the binary categorizations were coded as the following stress states: Boredom (E =is not, D =is not); Eustress (E =is, D =is not); Eustress-Distress Coexistence (E =is, D =is); and Distress (E =is not, D =is).

Statistical Analyses

Our analyses focused on testing for significant differences in ratings of productivity, mood, and stress between the four stress states, considering work versus non-work activities and then social versus solo activities within those contexts. First, we conducted two-way factorial ANOVAs to examine productivity, mood, and general stress ratings, comparing work versus nonwork activity within the four stress states. Significant interactions between work status and stress state were noted; thus, we conducted individual MANOVAs for work and non-work activities to examine differences in ratings between stress states. Further dividing activities by social context within the work context resulted in low-frequency cells for non-work activity stress-state ratings; thus, we only examined social context within work-related activities. To account for unequal cell sizes, we used Scheffe’s Test for post-hoc testing of individual interactions at an alpha level of .05.

Results

Our diverse worker cohort ranged in age from 27 to 52 and included 4 females and 3 males, 3 of whom were white, 2 black, 1 Hawaiian/Pacific Islander, and 1 multiracial; 3 participants identified as Hispanic or Latinx. Five individuals had administrative or support staff positions, and two had academic or faculty appointments. Hybrid work arrangements varied across participants, with the time spent working from home on a typical work week ranging from ad hoc to 75%.

We distributed 2,130 surveys across the 4-month study period to our 7-person cohort; we ended data collection with a sample of 1,970 responses (i.e., 92.5% response rate). After accounting for missing VEDAS and activity data, 1,936 complete responses were categorized into stress states and coded by work and social contexts (Figure 1).

Figure 1.

Flowchart displaying the frequencies of final complete EMA responses within the four stress states by social and solo contexts within work and non-work activities.

Differences in Mean Productivity, Mood, and Stress Across Stress States Considering Work Versus Non-Work Activities

A display of preliminary statistical analysis findings for work (a) and non-work (b) activities is provided in Figure 2.

Figure 2.

Distribution of productivity, mood, and stress self-ratings (0–100) during (a) work and (b) non-work activities across four stress states: boredom (blue, circle), eustress (red, cross), eustress-distress coexistence (green, X), and distress (brown, triangle).

Productivity

Participants consistently rated work activities as more productive than non-work activities regardless of the stress state (p < .0001). The mean productivity ratings for stress states within work activity responses were 61.4 (SD = 17.6), 72.5 (SD = 22.2), 72.0 (SD = 17.6), and 62.8 (SD = 20.5) for boredom, eustress, eustress-distress coexistence, and distress, respectively. Among non-work activities, mean productivity fell between 42.4 and 44.7 for all stress states except eustress-distress co-existence, which was higher at 58.2 (SD = 22.3).

For work activities, a significant difference in productivity ratings was found by stress states (F[3, 1462] = 26.00, p < .0001); states where eustress was present had significantly higher productivity ratings than boredom (p < .0001) or distress (p = .04). No differences were noted in productivity across stress states for non-work activity (F[3, 347] = 2.39, p = .07).

Mood

Ratings of mood across the four stress states varied between work and non-work activities (p = .0002). Mood ratings were similar between work and non-work activities identified in states of eustress (work: M = 77.3, SD = 15.8); non-work: M = 79.8, SD = 18.9) and distress (work: M = 58.3, SD = 24.4); non-work: M = 58.9, SD = 26.3). However, average mood ratings in states of boredom were better during non-work activities (M = 63.6, SD = 23.3) than for boredom at work (M = 55.8, SD = 19.8), while states of eustress-distress coexistence had better mean mood during work activities (M = 65.2, SD = 19.0) than non-work activities (M = 61.4, SD = 24.4).

We also found significant differences in overall mood ratings between stress states within both work activities (F[3, 1460] = 54.20, p < .0001) and non-work activities (F[3, 430] = 13.86, p < .001). Across both work and non-work activities, mood was significantly better when participants identified as being in a state of eustress compared to all other stress states.

Stress Magnitude

Ratings of stress magnitude across the four stress states varied between work and non-work activities (p = .004). Boredom and eustress were the lowest, and the two states involving distress were highest in both activity contexts. States of boredom were rated as significantly more stressful during non-work activities (M = 35.1, SD = 23.7) than during work (M = 22.8, SD = 15.1), while eustress resulted in slightly higher stress during work (M = 29.5, SD = 19.2) than non-work (M = 25.5, SD = 18.4). Stress magnitude was the highest in a state of eustress-distress coexistence during work activities (M = 44.6, SD = 23.9). Average stress ratings were similar between distress at work (M = 41.1, SD = 23.2), and non-work activities identified as distress (M = 42.6, SD = 26.2) and eustress-distress coexistence (M = 39.8, SD = 24.9).

We also found significant differences in overall stress ratings between stress states within both work activities (F[3, 1320] = 62.60, p < .0001) and non-work activities (F[3, 323] = 5.16, p = .002). Within work activities, there was evidence for a difference in mean stress by stress state for boredom and eustress (p < .0001 and p = .005, respectively). Boredom had a significantly lower mean stress rating than eustress, eustress-distress coexistence, and distress categories (p = .04, p < .0001, and p = .0003, respectively). Eustress also had significantly lower mean stress than the eustress-distress coexistence category (p < .0001). Within non-work activities, eustress was rated significantly lower than boredom and distress categories (p = .02).

Differences in Mean Productivity, Mood, and Stress Across Stress States Considering Social Context of Work Activity

Productivity

The interaction between stress state and social context was statistically significant for productivity (p = .003). By social context, there was evidence for a difference in mean productivity by stress state for solo activity (F[3, 1100] = 25.76, p < .0001). Within solo activity, there was evidence for a difference in mean productivity by stress state for eustress and eustress-distress coexistence (p = .0004 and p = .008, respectively). The mean productivity ratings within solo activity were 21.1 (SD = 12.8), 28.0 (SD = 18.4), 43.2 (SD = 23.8), and 36.5 (SD = 20.7) for boredom, eustress, eustress-distress coexistence, and distress, respectively. Eustress responses had significantly higher mean productivity than boredom and distress (p < .0001 and p = .006, respectively). Eustress-distress coexistence also had significantly higher mean productivity than boredom (p < .0001). No statistically significant differences in mean productivity by stress state were detected within social activity.

Mood

The interaction effect between stress state and social context for mood was not significant, so we evaluated the effects of both predictors together in a two-way factorial model. We observed a similar trend for social and solo activity, with a slightly higher mean mood for social activity across stress states. We found evidence for a significant difference in mean mood by stress state adjusting for social context (F[4, 1446] = 40.16, p < .0001). There was no significant difference in mean mood between social and solo activity after adjustment for stress state (p = .10). The mean mood ratings adjusted for social context were 56.0 (SD = 19.8), 77.3 (SD = 15.9), 65.2 (SD = 19.0), and 58.3 (SD = 24.4) for states of boredom, eustress, eustress-distress coexistence, and distress, respectively. Adjusting for the social context of activity, eustress had a significantly higher mean mood than boredom, eustress-distress coexistence, and distress (p < .0001). Eustress-distress coexistence also had a significantly higher mean mood than boredom (p < .0001).

Stress Magnitude

The interaction effect between stress state and social context for stress was not significant, so we evaluated the effects of both predictors together in a two-way factorial model. We observed a similar trend for social and solo activity, with higher mean stress for social activity across stress states. We found evidence for a significant difference in mean stress by stress state adjusting for social context (F[4, 1309] = 52.38, p < .0001). There was also a significant difference in mean stress between social and solo activity after adjustment for stress state (p < .0001). The mean stress ratings were 22.8 (SD = 15.2), 29.5 (SD = 19.3), 44.6 (SD = 23.9), and 41.1 (SD = 23.2) out of 100 for states of boredom, eustress, eustress-distress coexistence, and distress, respectively. Adjusting for social context of activity, mean stress during distress stress states was significantly higher than during boredom (p = .0004) or eustress alone (p = .02). Eustress-distress coexistence also had significantly higher mean stress than boredom (p < .0001) and eustress (p < .0001).

Discussion

The findings of this preliminary analysis provide useful insights for the ongoing investigation into the relationship of positive and negative stress experiences to worker outcomes such as productivity and mood. We observed higher self-rated productivity during states of eustress and eustress-distress coexistence. This holds true to the Yerkes–Dodson relationship between performance and arousal, whereby positive stress can be associated with high performance and a feeling of accomplishment. After categorizing the data by work and social contexts, this relationship held true for work-related activities performed alone but not necessarily for work-related activities performed with others.

We also observed a common social assumption underlying the patterns in ratings of stress magnitude, that is, that “stress” is often conflated with distress. Indeed, stress magnitudes were higher for activities categorized as distress and lower during activities categorized as boredom or eustress. This finding aligned with the Yerkes–Dodson Law implication that boredom is a low arousal state and signals that participants may link their overall stress magnitude with distress more than eustress experiences. We observed additional differences in mean stress between stress states when dichotomizing the data by work context—within non-work activity, eustress had lower mean stress than boredom. It was interesting that this relationship was present in non-work activity, suggesting that boredom may be more stressful during non-work compared to work.

Trends in mood data support the idea that positive stress is distinguishable from negative stress. Mean mood was higher during eustress compared to boredom and both states involving distress. After adjusting for work and social context within work activity, mood’s relationship with stress state remained nearly the same, potentially serving as additional evidence for the strength of this trend in our data. Mood may be of interest for further studies of stress state data to consider how it coincides with distinguishable states of eustress versus distress.

This interim analysis faced multiple limitations. We could not assess the social context of non-work activity due to insufficient sample size. We must collect more data before gaining initial insights into the relationship between non-work activity and stress state. Overall, there is a need for more data to confirm all relationship insights presented. Of additional importance, the presented results have not considered individual effects. It will be a key next step to analyze the role of individual effects on the observed relationships.

Conclusion

Overall, this interim analysis provides key insights to support future analyses of the relationship between various worker outcomes and distinguishable states of stress. In response to the hypotheses proposed at the start of this manuscript, we confirmed significant differences between stress states for all three considered outcome variables of productivity, stress, and mood. Work and social contexts were most influential in the relationship between productivity and stress states, with less influence across stress states on ratings of overall mood. When we analyzed stress ratings by work context, some nuances emerged in the relationship between self-rated stress magnitude and stress states, specifically related to experiences of boredom between work and non-work contexts.

In future research, we will test all emerging patterns from this interim analysis in larger datasets as the multi-cohort study continues. For our team’s and other researchers’ efforts, it would be greatly beneficial to investigate the relationship between performance and worker stress states. Self-rated or objective performance measures would be useful for investigating the occurrence of the Yerkes–Dodson Law within different worker populations to further understand if and how these relationships hold true.