Abstract
Background:
Shift work, a nonstandard work arrangement, is increasingly popular in the hospitality industry to meet consumer demands and boost productivity. However, its effects on the health and well-being of hotel workers in the United States remains underexplored.
Objective:
We aimed to explore the effects of shift work on mental health, sleep, and chronic fatigue among hotel workers.
Methods:
We collaborated with a community advisory board to develop a survey and recruit participants. Individuals aged 18 years or older, working in hotels in a Midwest state, completed surveys online or on paper. Data analysis included descriptive statistics, linear and logistic regression, and stepwise moderation to examine the impact of work schedules on chronic sleep problems, fatigue, and mental health outcomes.
Results:
Data from 518 participants were analyzed. The majority identified as white and female, with an average age of 32.05 years. Most were employed in housekeeping or front desk positions. Regression analyses showed that those working nonstandard hours were significantly more likely to report chronic fatigue and screen positive for posttraumatic stress disorder. Paid overtime was significantly associated with decreased symptoms of depression and anxiety. Significant interactions by job type were also observed, with front desk workers reporting increased depression with more overtime compared to other job types.
Conclusion:
These findings address a crucial gap in understanding shift work’s impact on US hotel workers’ health and well-being. Future longitudinal studies should explore shift work’s effects on sleep, fatigue, and mental health. Tailored interventions and policies may mitigate the adverse effects of shift work.
Keywords: shiftwork, non-standard work arrangements, hotel workers, service workers, low-wage, tourism, hospitality, sleep, fatigue, mental health
Shiftwork is characterized by work outside of the standard daylight hours (eg, 7 AM-6 PM; 8-hour work day), long hours exceeding 40 hours a week, overtime, rotating shifts, and irregular and varying days.1 The National Occupational Research Agenda identified shiftwork as a priority area in the health and safety of the growing US workforce.2 Such a mode of nonstandard work arrangement is commonly adopted in the health care, manufacturing, and law enforcement industries.3 However, shift work has gained momentum in other industries such as hospitality and tourism with the goal of maximizing revenues, being responsive to consumer demands, and increasing productivity. Yet, we lack understanding of the impact of shift work among workers in the hospitality industry. Hotel workers exemplify service workers who receive low wages and have high job demands. However, the unique nature of their stressful work environment in hotels, work stressors, and environmental exposures differ, requiring industry-specific explorations and tailored interventions.
Although viewed as a lucrative avenue for employers, shiftwork can be detrimental to the health and well-being of employees, resulting in negative health effects such as mental fatigue,4 biological dysregulation of sleep-alertness patterns,5 poor cardiovascular health,6 metabolic syndrome, diabetes, obesity,7 decreased safety,8 increased risks for accidents and injuries,9 poor health behaviors,10 and worse job satisfaction.11 Given the promising aspects of shiftwork and the fact that some industries need shift workers to function (eg, nurses are needed for the night shift to care for patients throughout the night), this type of nonstandard work arrangement will not be fully eliminated. Mitigating strain and supporting shift workers will ensure optimized health.
We lack understanding of the experience of US hotel workers engaged in shift work and their health outcomes. The few published works on the impact of shift work on hotel workers’ health has heavily stemmed from an international context,12–14 with scant attention to US hotel workers. In this paper, we address this gap by reporting findings on shift work and associated outcomes of mental health, sleep, and fatigue among a group of hotel workers located in a mid-western region of the United States.
Methods
A community advisory board (CAB) was comprised of 5 hotel workers, the president of a Local Union, and a union organizer assisted in the survey development and recruitment for this cross-sectional descriptive study. Research staff and CAB members distributed the study flyer in hotels. The study flyer was also posted on social media platforms, particularly on Facebook. We worked with a university-based research support group to promote social media postings to maximize visibility of the project specifically to those who were working at or had affiliations with hotels and motels in the region. To participate in the study, individuals had to (1) be 18 years or older, (2) work in that particular Midwest State, and (3) work at a hotel or motel in any capacity. The survey was available via both Qualtrics and a paper format since prior work indicated that some workers prefer to complete the surveys on hard copies. Data collection occurred between February 2022 and February 2023. Each individual received $25 for their participation.
Measures
The study included individuals in the following job categories: manager, housekeeper, food, bellhops, and front desk. We used a skip pattern approach whereby some questions were not asked of certain participants. For example, those who worked in the food or front desk category were not asked about room cleaning. For this manuscript, everyone (across all the job categories) was asked about work hours, mental health, sleep health, and fatigue.
Demographics.
Participants reported on their gender identity, race, age, sexual orientation, marital status, work location, job type, education, hourly pay, and annual income. We used single-item questions to capture demographic information, except for education, hourly pay, and annual income data, which utilized ordinal scales.
Work schedule.
Participants were asked whether they worked full-time or part-time. They were also asked “What shift do you typically work?” and given the option of “day shift,” “evening shift,” or “night shift.” Participants were asked how many hours per week they normally work for 3 categories: standard hours, paid overtime, and extra hours without pay. Another question asked “Do you normally work any hours outside the ‘typical’ working week? (Early mornings before 8 AM, evenings after 5 PM, or weekend work)” with a dichotomous option of “Yes” or “No.”
Sleep and fatigue.
Participants were asked “Please tell us if a healthcare provider has ever told you that you have any of these chronic conditions.” They were specifically asked to select 1 of 3 options: “Yes, and I am currently getting treatment,” “Yes, but I am not currently getting treatment,” and “No.” The options they could endorse included a list of 11 chronic health conditions. For this study, we looked specifically at “chronic sleeping problems” and “chronic fatigue or low energy.”
Mental health.
Depression symptoms were assessed with the Patient Health Questionnaire-9 (PHQ-9).15 The PHQ-9 consists of nine 4-point Likert scale items (1 = not at all, 2 = several days, 3 = more than half the days, 4 = nearly every day). Summing the items composes the total score, with higher scores indicating greater depression symptom severity. The PHQ-9 demonstrates excellent internal reliability (ɑ = 0.89), test-retest reliability (intraclass correlation = 0.84), criterion validity, and construct validity.15 In our sample, we obtained satistfactory internal reliability (ɑ = 0.87).
Anxiety symptoms were assessed with the General Anxiety Disorder-7 (GAD-7) scale.16 The GAD-7 consists of seven 4-point Likert scale items (0 = Not at all, 1 = Several days, 2 = More than half the days, 4 = Nearly every day). Summing across the scores generates a total score, with higher scores indicating greater anxiety symptom severity. The GAD-7 has demonstrated excellent internal reliability (ɑ = 0.92), test-retest reliability (intraclass correlation = 0.83), construct validity, and convergent validity (r range 0.72–0.74).16 In our sample, we replicated good internal reliability, (ɑ = 0.87).
The potential presence of posttraumatic stress disorder (PTSD) was screened via the primary care-PTSD screen for DSM-5 (PC-PTSD-5).17 The PC-PTSD-5 screens for the presence of a Criterion A traumatic stressor (yes/no). If no traumatic event is endorsed, the total score is 0; otherwise, the participant is then instructed to answer 5 more questions (yes/no) corresponding to DSM-5 defined symptoms of PTSD.17 The sum of “yes” responses on these 5 items is the total score. For our study, we chose to maximize optimal sensitivity (ie, our ability to detect true PTSD), thus selecting a cutoff score of 3 to screen positive for PTSD.17 This cutoff demonstrated a sensitivity of 95% and specificity of 85%. Overall, the PC-PTSD-5 has demonstrated excellent diagnostic accuracy (Area Under the Curve [AUC] = 0.941), test-retest reliability (r = 0.83), and predictive validity against the Clinician-Administered PTSD scale (ie, CAPS; r = 0.83).18
Data Cleaning Procedure
Participants completed an online Qualtrics survey with all the above measures or an in-person paper version, which was then subsequently data-entered. Before analysis, the full data set was first cleaned to exclude duplicate IP addresses and non-Michigan residents (by ZIP code). Further, the larger battery contained other questions that were used to maintain accuracy of the data. Inconsistent responses (eg, different responses to duplicate items), illogical responses (eg, rating satisfaction with a work benefit they did not endorse), and irrelevant responses (eg, open entry responses that did not answer the question coherently) contributed to an ongoing tally for each participant. This tally resulted in 218 (29.86%) participants without any invalid responses, 344 (47.12%) with 1 invalid response, and 168 (22.47%) with 2 or more invalid responses. With the understanding that participants may be prone to make occasional invalid responses, we made a decision to set a cutoff such that participants with 2 or more invalid responses were excluded from the analysis.
Data Analysis
Descriptive statistics were computed as summary statistics and/or percentages. Work schedule variables (continuous and dichotomous) were considered independent variables and used to model the presence of either chronic sleep problems or chronic fatigue via logistic regression such that if the participants endorsed a chronic health condition it was rated as “Yes.” Work schedule variables were analyzed to model each mental health outcome via linear and logistic regression. For each model, we also used stepwise moderation analysis to add in job type and its interaction with each independent variable to see whether these associations would vary by job type. The best fitting model was selected by Akaike Information Criterion (AIC) and likelihood ratio tests, with only coefficients from the best fitting model (baseline, main effects, or interaction) being interpreted. For all analyses, significant findings (α = .05) were corrected via the Benjamini-Hochberg procedure to control for the false discovery rate, with only corrected p values reported. All analyses were run using the R statistical software (v4.4.1; R Core Team, Vienna, Austria).
Results
Descriptive Statistics
After data cleaning, 518 participants’ responses remained. A majority of participants identified as white (71.4%) and female (52.1%). Participants were on average 32.1 years old (SD = 6.3). Most participants worked in housekeeping (41.5%) or front desk positions (24.3%). Job locations were mostly hotels (92.3%) versus motels or hotels connected to a casino. See Table 1 for more details.
Table 1.
Sociodemographic Characteristics of Participants.
| Characteristics | Full sample (n = 518) |
|
|---|---|---|
| n (except for age) | % (except for age) | |
| Gender | ||
| Female | 270 | 52.12 |
| Male | 243 | 46.91 |
| Trans woman | 2 | 0.39 |
| Trans man | 2 | 0.39 |
| Nonbinary | 1 | 0.19 |
| Age | M = 32.05 | SD = 6.31 |
| Race | ||
| White | 370 | 71.43 |
| Black/African American | 91 | 17.57 |
| Hispanic/Latino/Mexican American/Mexican | 41 | 7.92 |
| American Indian/Alaska Native | 11 | 2.12 |
| Native Hawaiian/Other Pacific Islander | 2 | 0.39 |
| Other | 1 | 0.19 |
| Sexual orientation | ||
| Straight or heterosexual | 476 | 91.89 |
| Bisexual | 18 | 3.47 |
| Lesbian or gay | 17 | 3.28 |
| Prefer not to say | 3 | 0.58 |
| Pansexual | 1 | 0.19 |
| Marital status | ||
| Married | 287 | 55.41 |
| Single | 183 | 35.33 |
| Partnered | 20 | 3.86 |
| Divorced | 19 | 3.67 |
| Separated | 7 | 1.35 |
| Work location | ||
| Hotel | 478 | 92.28 |
| Motel | 28 | 5.41 |
| Hotel connected to a casino | 10 | 1.94 |
| Job type | ||
| Housekeeping | 215 | 41.51 |
| Front desk | 126 | 24.32 |
| Manager/supervisor | 82 | 15.83 |
| Kitchen | 40 | 7.72 |
| Laundry | 34 | 6.56 |
| Bellhop | 17 | 3.28 |
| General cleanera | 2 | 0.39 |
| Education | ||
| Eighth grade or less | 5 | 0.97 |
| Some high school | 23 | 4.44 |
| High school graduate or General Education Development (GED) | 102 | 19.69 |
| Some college or 2-year degree | 136 | 26.25 |
| 4-year college graduate | 163 | 31.47 |
| More than 4-year college degree | 89 | 17.18 |
| Hourly pay | ||
| <$10.00 | 26 | 5.02 |
| $10.01-$12.00 | 52 | 10.04 |
| $12.01-$14.00 | 41 | 7.92 |
| $14.01-$16.00 | 25 | 4.83 |
| $16.01-$18.00 | 121 | 23.35 |
| $18.01-$20.00 | 86 | 16.60 |
| $20.01-$25.00 | 71 | 13.70 |
| $25.01-$30.00 | 48 | 9.27 |
| >$30.00 | 38 | 7.34 |
| Prefer not to sayb | 10 | 1.93 |
| Annual income | ||
| <$20 000 | 14 | 2.70 |
| $20 000-$39 999 | 114 | 22.01 |
| $40 000-$59 999 | 177 | 34.17 |
| $60 000-$79 999 | 122 | 23.55 |
| $80 000-$99 999 | 44 | 8.49 |
| >$100 000 | 41 | 7.92 |
| Prefer not to sayb | 5 | 0.98 |
| No responseb | 1 | 0.18 |
Abbreviations: M: mean; SD: standard deviation.
Due to limited sample size, general cleaners were excluded from all analyses.
Individuals who selected “prefer not to say” or did not respond were excluded from final analyses.
Most participants reported working full-time (n = 468, 90.3%) and during the day shift (n = 388, 74.9%); evening shift was endorsed by 17.8% (n = 92), while night shift was 7.3% (n = 38). Participants reported working on average 30.62 standard hours (SD = 15.59), 5.65 overtime hours (SD = 9.46), and 2.08 extra hours without pay (SD = 6.54). Most participants (n = 303, 58.5%) reported normally working outside of the “typical” work week (eg, early morning before 8 AM, evening after 5 PM, or weekend work).
Self-reported sleep and fatigue are described in Table 2. Most participants did not report receiving a diagnosis of chronic sleeping problems or fatigue/low energy. Of those who did, a larger percentage reported not receiving care for their sleep or fatigue problems (16.99% vs 6.37%). Mental health outcomes for the sample are summarized in Table 2. On average, our participants scored in a range demonstrating mild depression symptoms (9.7) and mild anxiety symptoms (7.7). Seventeen percent of participants screened positive for PTSD.
Table 2.
Sleep, Fatigue, and Mental Health by Job (N = 518).
| Health conditions | Full sample | Housekeepers | Front desk staff | Manager/supervisor | Kitchen workers | Laundry workers | Bellhops |
|---|---|---|---|---|---|---|---|
| Chronic sleeping problems, n (%) | |||||||
| Yes and treatment | 49 (9.46) | 23 (10.95) | 6 (4.76) | 8 (9.76) | 9 (23.08) | 1 (2.94) | 1 (5.88) |
| Yes and no treatment | 109 (21.04) | 46 (21.90) | 21 (16.67) | 20 (24.39) | 11 (28.21) | 8 (23.53) | 3 (17.65) |
| No | 354 (68.34) | 141 (64.14) | 99 (78.57) | 54 (65.85) | 19 (48.72) | 25 (73.53) | 13 (76.46) |
| Chronic fatigue or low energy, n (%) | |||||||
| Yes and treatment | 33 (6.37) | 19 (9.09) | 2 (1.59) | 1 (1.22) | 7 (17.95) | 3 (8.82) | 0 (0) |
| Yes and no treatment | 88 (16.99) | 34 (16.27) | 20 (15.87) | 13 (15.85) | 8 (20.51) | 6 (17.65) | 7 (41.18) |
| No | 390 (75.29) | 156 (74.64) | 104 (82.54) | 68 (82.93) | 24 (61.54) | 25 (73.53) | 10 (58.82) |
| Mental health outcomes | |||||||
| Depression, mean ± SD | 9.68 ± 5.43 | 10.17 ± 5.49 | 9.08 ± 4.87 | 7.22 ± 5.38 | 11.69 ± 5.49 | 11.18 ± 5.69 | 12.18 ± 3.68 |
| Anxiety, mean ± SD | 7.66 ± 4.66 | 8.02 ± 4.94 | 7.25 ± 4.12 | 5.94 ± 4.27 | 8.85 ± 4.99 | 8.94 ± 3.85 | 9.31 ± 4.16 |
| PTSD, n (%) | 81 (17.80) | 53 (27.32) | 15 (13.76) | 11 (15.28) | 1 (3.03) | 0 (0) | 0 (0) |
Note. Depression is measured by the PHQ-9, anxiety the GAD-7, and PTSD the PC-PTSD-5.
Abbreviations: GAD-7: General Anxiety Disorder-7; PC-PTSD: primary care-posttraumatic stress disorder; PHQ-9: Patient Health Questionnaire-9.
Work, Sleep, and Fatigue
Of the logistic regressions modeling sleep, the interaction model fit best (AIC = 564.17 vs 595.13 baseline and 588.17 main effects) and fit significantly better than the nested main effects model, F[16, 465] = 56.00, p < .001. Thus, only coefficients from the interaction model were interpreted (Table 3). While assumptions of multicollinearity were met in this model, participants who worked as bellhops had to be removed from the model due to a small sample size (n = 17), which introduced bias in the estimation; likewise, interactions with part-/full-time work and shifts had to be excluded in the interaction model due to violations in the assumption of a fully represented data matrix. With every 1-hour increase in overtime, the odds of reporting chronic sleep problems multiplied by 1.13 times (p = .001). In contrast, every 1 hour increase in extra hours without pay decreased the odds of reported sleep problems by 11% (p = .017). Several interactions were significant. First, front desk staff (odds ratio (OR) = 1.05) and kitchen workers (OR = 1.08) who reported working more standard hours also demonstrated a significantly increased probability of chronic sleep problems (p = .042 and p = .018, respectively). Second, front desk workers who reported more overtime demonstrated a significantly decreased probability of chronic sleep problems (OR = 0.79, p = .002; Figure 1). Third, front desk (OR = 1.26) and kitchen workers (OR = 2.02) who reported more extra hours without pay demonstrated a significantly increased probability of chronic sleep problems (p = .004 and p = .005, respectively; Figure 2). Lastly, kitchen workers (OR = 0.02) and managers/supervisors (OR = 0.23) who reported working outside of typical work hours were significantly less likely to report chronic sleep problems (p = .005 and p = .028, respectively; Figure 3). Notably, all these effects were observed after statistically controlling for shift schedule and job type, which did not additively contribute a significant amount of variance to explaining sleep problems.
Table 3.
Work Schedule Associations With Sleep and Fatigue.
| Variables | Chronic sleeping problems |
Chronic fatigue or low energy |
||||
|---|---|---|---|---|---|---|
| OR | SE | 95% CI | OR | SE | 95% CI | |
| Intercept | 0.27 | 0.43 | 0.11–0.63 | 0.12 | 0.49 | 0.04–0.31 |
| Part-time work | 0.66 | 0.43 | 0.27–1.48 | 0.54 | 0.47 | 0.20–1.30 |
| Shift | ||||||
| Evening shift | 1.10 | 0.30 | 0.60–1.97 | 1.51 | 0.32 | 0.80–2.79 |
| Night shift | 0.93 | 0.46 | 0.36–2.22 | 1.44 | 0.46 | 0.56–3.47 |
| Standard hours | 0.99 | 0.01 | 0.97–1.01 | 1.01 | 0.01 | 0.99–1.03 |
| Overtime hours | 1.13**** | 0.03 | 1.06–1.21 | 1.03 | 0.03 | 0.96–1.10 |
| Unpaid hours | 0.89* | 0.05 | 0.81–0.97 | 0.93 | 0.06 | 0.80–1.03 |
| Outside of work week | 2.60 | 0.35 | 1.33–5.26 | 2.68* | 0.37 | 1.33–5.67 |
| Current job | ||||||
| Front desk | 0.44 | 1.00 | 0.05–2.85 | 1.39 | 1.05 | 0.16–10.24 |
| Kitchen | 0.63 | 1.21 | 0.04–5.51 | 1.14 | 1.28 | 0.06–10.99 |
| Laundry | 0.88 | 1.28 | 0.04–8.39 | 1.81 | 1.34 | 0.07–19.26 |
| Manager/supervisor | 2.22 | 0.70 | 0.55–8.74 | 5.51* | 0.78 | 1.19–25.59 |
| Interactions | ||||||
| Standard hours × front desk | 1.05* | 0.02 | 1.00–1.10 | 1.01 | 0.02 | 0.96–1.06 |
| Standard hours × kitchen | 1.08* | 0.03 | 1.02–1.15 | 1.01 | 0.03 | 0.96–1.07 |
| Standard hours × laundry | 1.03 | 0.04 | 0.97–1.12 | 1.02 | 0.04 | 0.951.11 |
| Standard hours × manager/supervisor | 1.01 | 0.02 | 0.97–1.04 | 0.97 | 0.02 | 0.93–1.01 |
| Overtime hours × front desk | 0.79*** | 0.07 | 0.69–0.89 | 0.80* | 0.08 | 0.68–0.93 |
| Overtime hours × kitchen | 0.96 | 0.07 | 0.87–1.20 | 1.03 | 0.06 | 0.94–1.21 |
| Overtime hours × laundry | 0.86 | 0.10 | 0.65–1.02 | 0.79 | 0.17 | 0.53–1.04 |
| Overtime hours × manager/supervisor | 1.04 | 0.07 | 0.91–1.20 | 0.94 | 0.09 | 0.78–1.11 |
| Unpaid hours × front desk | 1.26*** | 0.07 | 1.10–1.46 | 1.33** | 0.09 | 1.13–1.63 |
| Unpaid hours × kitchen | 2.02** | 0.22 | 1.38–3.31 | 1.67* | 0.21 | 1.17–2.67 |
| Unpaid hours × laundry | 1.09 | 0.30 | 0.53–1.91 | 1.19 | 0.28 | 0.67–2.08 |
| Unpaid hours × manager/supervisor | 0.89 | 0.10 | 0.68–1.08 | 1.18 | 0.11 | 0.95–1.50 |
| Outside of work week × front desk | 0.39 | 0.62 | 0.12–1.33 | 0.37 | 0.68 | 0.10–1.43 |
| Outside of work week × kitchen | 0.02** | 1.29 | 0.00–0.16 | 0.13 | 1.12 | 0.01–1.08 |
| Outside of work week × laundry | 0.30 | 1.12 | 0.03–3.33 | 0.87 | 1.39 | 0.07–14.89 |
| Outside of work week × manager/supervisor | 0.23* | 0.64 | 0.07–0.81 | 0.10** | 0.75 | 0.02–0.43 |
Note. 95% CIs are significant if they do not include 1.00.
The reference group worked full-time, worked the day shift, and are housekeepers; ORs are compared against them. Bellhop was removed due to fitted probabilities being close to 0 or 1.
Abbreviations: CI: confidence interval; OR: odds ratio; outside of work week: working outside of the “typical” working week; SE: standard error; unpaid hours: extra hours without pay;.
p = .001,
p < .005,
p = .005,
p < .05.
Figure 1.
Effects of overtime work hours on sleep problems by job.
Note. Relationship was significant for front desk workers. Bands reflect 95% confidence intervals, with narrower bands indicating more certainty of the direction of the relationship.
Figure 2.
Effects of extra hours without pay on sleep problems by job.
Note. Relationship was significant for kitchen workers. Bands reflect 95% confidence intervals, with narrower bands indicating more certainty of the direction of the relationship.
Figure 3.
Effects of working outside of the typical work week on sleep problems by job.
Note. Relationships were significant for kitchen workers and managers/supervisors. Bars reflect 95% confidence intervals, with narrower bars indicating more certainty of the direction of the relationship.
Of the logistic regressions modeling chronic fatigue or low energy, the interaction model fit best (AIC = 519.46 vs 544.20 baseline and 543.79 main effects) and fit significantly better than the nested main effects model, F[20, 480] = 64.33, p < .001. Thus, only coefficients from the interaction model were interpreted (Table 3). While assumptions of multicollinearity were met in this model, participants who worked as bellhops had to be removed from the model due to biases in the estimation; likewise, interactions with part-/full-time work and shifts had to be excluded in the interaction model due to violations in the assumption of a fully represented data matrix. Participants who reported working outside of the typical work week were 2.68 times more likely to report chronic fatigue (p = .014). Managers/supervisors were also 5.51 times more likely to report chronic fatigue than housekeepers (p = .033). Several interactions were significant. First, front desk staff who reported more overtime hours also demonstrated a significantly decreased probability of chronic fatigue (OR = 0.80, p = .011). Second, front desk (OR = 1.33) and kitchen workers (OR = 1.67) who reported more extra hours without pay demonstrated a significantly increased probability of chronic fatigue (p = .005 and p = .020, respectively). Lastly, managers/supervisors who worked outside of the typical work week were significantly less likely to report chronic fatigue versus those who did work a typical week, OR = 0.10, p = .005). Notably, all these effects were observed after statistically controlling for shift schedule, which did not additively contribute a significant amount of variance to explaining chronic fatigue.
Work and Mental Health
Limited sample sizes per bin resulted in the removal of interaction terms for part-/full-time work and shift in all mental health models. Due to bimodality in the distribution of depression and anxiety symptoms, the assumption of normality was violated in those linear regression models. Thus, quantile regression was used instead for these outcomes.
For depression symptoms, the third quartile (75th percentile) fit best versus other quartiles (AIC = 3136.64) and thus was used for all models. Then, the interaction model fit best (AIC = 3018.55) compared with the baseline (AIC = 3093.72) and main effects models (AIC = 3067.44); further, it fit the data significantly better than the main effects model (F[20, 461] = 8.43, p < .001). Thus, only the coefficients of the interaction model were interpreted (Table 4). Participants who reported working more standard (β = −0.02, p < .001) or overtime hours (β = −0.04, p < .001) also reported significantly less depression symptoms, whereas those who reported more extra hours without pay (β = 0.04, p = .031) or who worked outside of the typical work week (β = 0.39, p = .014) reported significantly more depression symptoms. Bellhops reported significantly lower depression symptoms versus housekeepers (β = −1.98, p = .019). After several interaction effects demonstrated significance, post hoc simple slope analyses were run, revealing several significant slopes. First, for bellhops who worked more standard hours, their depression symptoms were significantly lower (β = −0.55 p = .048). Second, housekeepers who worked more overtime reported significantly lower depression symptoms (β = −0.77, p = .003) whereas front desk staff who worked more overtime reported significantly higher depression symptoms (β = 0.60, p = .004). Last, housekeepers who worked outside of the typical work week reported significantly lower depression symptoms (β = −1.58, p = .001; Figure 4). Large effects included the following: bellhops versus housekeepers for standard hours; front desk staff versus housekeepers for overtime; and housekeepers for overtime and working outside of the typical work week. Generally, working outside of the typical work week had a medium effect on depressive symptoms. All other significant effects on depressive symptoms were small.
Table 4.
Work Schedule Associations with Depression and Anxiety Symptoms.
| Variables | Depression | Anxiety | ||
|---|---|---|---|---|
| β | 95% CI | β | 95% CI | |
| Intercept | 1.33 | 0.81–1.56 | 1.19 | 0.72–1.44 |
| Part-time work | 0.05 | −0.08 to 0.25 | 0.27 | −0.03 to 0.59 |
| Shift | ||||
| Evening shift | 0.02 | −0.18 to 0.15 | 0.01 | −0.11 to 0.34 |
| Night shift | −0.03 | −0.28 to 0.45 | 0.34 | −0.01 to 0.65 |
| Standard hours | −0.02 | −0.02 to −0.01 | −0.02 | −0.02 to −0.00 |
| Overtime hours | −0.04 | −0.05 to −0.03 | −0.05 | −0.06 to 0.02 |
| Unpaid hours | 0.04 | 0.03–0.09 | 0.04 | −0.00 to 0.13 |
| Outside of work week | 0.39 | 0.22–0.58 | 0.22 | 0.06–0.56 |
| Current job | ||||
| Bellhop | −1.98 | −2.07 to 1.15 | −1.53 | −3.06 to 2.88 |
| Front desk | −0.37 | −0.70 to 0.25 | −0.04 | −0.53 to 0.34 |
| Kitchen | −0.56 | −1.62 to 0.32 | −0.47 | −1.81 to 0.42 |
| Laundry | −0.51 | −0.81 to 1.47 | −0.91 | −1.43 to −0.03 |
| Manager/supervisor | −0.64 | −0.94 to −0.16 | −0.99 | −1.36 to −0.21 |
| Interactions | ||||
| Standard hours × bellhop | 0.06 | −0.01 to 0.06 | 0.43 | −0.05 to 0.61 |
| Standard hours × front desk | 0.01 | −0.00 to 0.02 | 0.01 | −0.00 to 0.02 |
| Standard hours × kitchen | 0.02 | 0.01–0.04 | 0.02 | −0.00 to 0.03 |
| Standard hours × laundry | 0.20 | −0.00 to 0.03 | 0.02 | 0.00–0.03 |
| Standard hours × manager/supervisor | 0.02 | 0.00–0.03 | 0.02 | −0.01 to 0.03 |
| Overtime hours × bellhop | 0.05 | 0.04–0.14 | 0.14 | 0.04–0.03 |
| Overtime hours × front desk | 0.05 | 0.03–0.07 | 0.05 | 0.01–0.18 |
| Overtime hours × kitchen | 0.05 | 0.03–0.07 | 0.07 | −0.01 to 0.08 |
| Overtime hours × laundry | 0.01 | −0.01 to 0.16 | 0.06 | −0.00 to 0.08 |
| Overtime hours × manager/supervisor | 0.13 | 0.06–0.18 | 0.17 | 0.11–0.18 |
| Unpaid hours × bellhop | −0.03 | −0.06 to 1.80 | −0.03 | −0.05 to 0.20 |
| Unpaid hours × front desk | −0.03 | −0.07 to −0.01 | −0.03 | −0.08 to 0.01 |
| Unpaid hours × kitchen | 0.07 | −0.05 to 0.41 | 0.16 | 0.10–0.48 |
| Unpaid hours × laundry | −0.01 | −0.19 to 0.53 | 0.13 | −0.10 to 0.77 |
| Unpaid hours × manager/supervisor | −0.09 | −0.15 to −0.04 | −0.15 | −0.18 to −0.11 |
| Outside of work week × bellhop | −0.85 | −1.72 to −0.71 | −1.47 | −2.05 to 0.16 |
| Outside of work week × front desk | −0.68 | −1.12 to −0.41 | −0.59 | −1.00 to −0.36 |
| Outside of work week × kitchen | −0.86 | −2.06 to 0.10 | −1.10 | −2.14 to −0.86 |
| Outside of work week × laundry | 0.18 | −0.50 to 0.65 | 0.10 | −0.31 to 0.49 |
| Outside of work week × manager/supervisor | −1.67 | −2.03 to −0.89 | −1.22 | −1.80 to −0.65 |
Note. 95% CIs are significant if they do not include 0.
Quantile regression was used for these models, with bootstrapped estimates and confidence intervals provided; standard errors are likewise irrelevant. The reference group worked full-time, worked the day shift, and are housekeepers. Depression symptoms were measured by the PHQ-9 and anxiety symptoms by the GAD-7.
Abbreviations: β: standardized coefficient; CI: confidence interval; GAD-7: General Anxiety Disorder-7; outside of work week: working outside of the “typical” working week; PHQ-9: Patient Health Questionnaire-9; unpaid hours: extra hours without pay.
Figure 4.
Effects of working outside of typical work hours on depression symptoms by job.
Note. Depression symptoms were measured by the PHQ-9. Relationship was significant for housekeepers. Lines reflect 95% confidence intervals, with narrower lines indicating more certainty of the direction of the relationship.
Abbreviation: PHQ-9: Patient Health Questionnaire-9.
For anxiety symptoms, the third quartile (75th percentile) fit best versus other quartiles (AIC = 2940.28) and thus was used for all models. Then, the interaction model fit best (AIC = 2792.20) compared to the baseline (AIC = 2876.25) and main effects models (AIC = 2857.79); further, it fit the data significantly better than the main effects model (F[20, 448] = 3.00, p < .001). Thus, only the coefficients of the interaction model were interpreted (Table 4). Participants who reported working more standard (β = −0.02, p = .018) or overtime hours (β = −0.05, p = .025) also reported significantly less anxiety symptoms. Managers/supervisors reported significantly lower anxiety symptoms versus housekeepers (β = −0.09, p = .019) at a large effect. After several interaction effects demonstrated significance, post hoc simple slopes analyses were run, revealing several significant slopes. First, for bellhops (β = 0.62, p = .001) and managers/supervisors (β = 0.62, p = .002) who worked more overtime hours, their anxiety symptoms were significantly higher. Second, housekeepers who worked outside of the typical work week reported significantly higher anxiety symptoms (β = −1.33, p = .020; Figure 5). All interaction effects were large, while all other effects were small.
Figure 5.
Effects of working outside of typical work hours on anxiety symptoms by job.
Note. Anxiety symptoms were measured by the GAD-7. Relationship was significant for housekeepers. Lines reflect 95% confidence intervals, with narrower lines indicating more certainty of the direction of the relationship.
Abbreviation: GAD-7: General Anxiety Disorder-7; PHQ-9: Patient Health Questionnaire-9.
Of the logistic regressions modeling PTSD, the interaction model fit best (AIC = 353.86) versus the baseline (AIC = 361.36) and main effects models (AIC = 351.85); likewise, the interaction model fit significantly better than the main effects model (F[8, 349] = 15.90, p = .044). Therefore, only the coefficients of the interaction model were interpreted. While assumptions of multicollinearity were met, job types of bellhop, kitchen worker, and laundry worker had to be removed due to violations in the assumption of a fully represented data matrix. Likewise, interactions for part-/full-time work and shift type were also excluded due to limited sample sizes. Only 1 variable was significantly associated with screening positive for PTSD, namely working outside of the typical work week (OR 95% CI 2.65–14.77, SE = 0.43, p < .001). Specifically, participants who worked outside of the typical work week were 5.93 times more likely to screen positive for PTSD versus those who do not.
Of the logistic regression modeling psychiatric hospitalization, the baseline model fit best (AIC = 365.39). Since this is also the more parsimonious model, only coefficients from the baseline model were interpreted.
Discussion
Our goal was to address the significant gap in our knowledge of the impact of shift work on the health of hotel workers in the United States, particularly mental health, sleep, and chronic fatigue. We found that overall, while those who reported working more overtime generally reported greater sleep problems and lower depressive symptoms, these relationships differed by job type. This aligns with existing literature that indicates overtime, on average, is associated with poorer sleep among workers, regardless of shift.19–21 Surprisingly, for some workers more overtime was associated with fewer depressive symptoms. An exception to this was front desk workers who reported more depressive symptoms when working more overtime hours. Front desk staff may be more exposed to social interactions with customers than other hotel workers.22 Frequent interaction with the public has been associated with poorer mental health among service industry workers.23 Similarly, kitchen workers and managers/supervisors endorsed fewer chronic sleep problems when they worked outside of typical work hours, which may be related to less public exposure and thus, lower work-load. Contrastingly, some job categories (bellhops and mangers/supervisors) reported more anxiety when participating in overtime work, yet on average, more paid overtime was associated with decreased anxiety. This is consistent across literature covering multiple worker industries; however, the literature to date does not examine this relationship among service workers.24,25 Some studies found an association with female sex and heightened anxiety and depression, and poorer sleep.26 PTSD has been associated with overtime and poorer sleep. However, these studies were primarily conducted with health care workers or first responders.27,28
Of note, of those who responded that they had chronic sleep problems, most did not receive treatment (21.04% vs 9.46%). Similarly, of those who responded that they had chronic fatigue, most did not receive treatment (16.99% vs 6.37%). A possible reason is due to lack of time. However, given our lack of information on the severity of their sleep problems and chronic fatigue, it is challenging to determine how concerning the lack of active treatment should be. Also, there are other approaches that can be used to remediate sleep issues that do not necessitate medication. Most hospitality workers expressed concerns that they have very little personal time.29 This may lead to a decreased number of primary care provider appointments to follow-up on chronic conditions.
Most job categories who reported working extra hours without pay reported increased sleep problems, fatigue, and depression, which is consistent with our findings with overtime; however, to date, few studies have examined overtime work without pay. Unexpectedly, the odds of being told by a health care provider that they had sleep problems decreased with increased hours worked without pay. While it is difficult to parse the exact reason for this, we could hypothesize that this may be due to multiple reasons, including underreporting, access to sleep medicine (and diagnostic services), and/or selection bias (eg, those with fewer sleep problems being more willing to work extra).
Other miscellaneous findings also emerged. Screening positive for PTSD was associated with more nonstandard working hours, and job categories that experienced adverse effects of working standard hours included bellhops, who have more interaction with the public.23 While depressive and anxiety symptoms were low on average for our sample, they were bimodally distributed, suggesting separate experiences for low- and high-severity participants.
To date, the existing literature on shift work has primarily focused on workers in other industries with minimal attention to hotel workers. Studies that focus on hotel workers are often older than 10 years, fail to examine the differentiation of paid versus unpaid overtime (working extra hours without compensation), and stem from an international context.19 Given the structural and regulatory differences between the United States and other countries on occupational health and safety, studies filling the gap in our understanding of the impact of shift work and targeted interventions that are US focus are warranted.
A limitation of this study was that we did not examine physiologic reasons for poor sleep, including conditions such as sleep apnea or the presence of dependents at home, which could affect sleep among workers. Also, we used social media as a method of recruitment. There are concerns in the scientific community about the quality and validity of surveys when participants are recruited through social media platforms. For this project, we established measures to address this concern including (1) creating set screening questions to ensure eligibility and (2) paying attention to duplicate IP addresses. Our sample sizes were small for some job categories, such as bellhops and laundry workers, which precluded them from some of our analyses. Additionally, given the cross-sectional nature of this study, we cannot establish causal inferences between shift work and the health outcomes we explored. Future longitudinal studies and larger sample sizes for some job categories are needed to explore this phenomenon further. While our study had some limitations, we also presented multiple strengths. Lastly, we did not use established validated sleep tools such as the Pittsburgh Sleep Quality Index and Epworth Sleepiness Scale, or chronic fatigue measures such as the fatigue severity scale for this study. However, participants subjectively responded that their health care providers notified that them that they had those health conditions. Future studies exploring shift work and sleep health and chronic fatigue can extend this current project to include those sleep measures. This study expands our understanding of the impact of shiftwork on worker health. To date, our study is the only exploration of overtime work delineations (paid vs unpaid) among hotel workers. Additionally, we investigated the effects of shiftwork on specific job categories of hotel workers compared to most studies that focus on hotel workers as a conglomerate population.
To conclude, examining the multiple psychological effects of shift work can contribute to the development of organizational policies and regulations needed to curtail the negative effects of shift work without jeopardizing the economic or job security of workers. On-site worker wellness programs and on-site health care can facilitate the improvement and prevention of chronic conditions.
Acknowledgments
We would like to thank all the hotel workers who responded to our survey, as well as the community advisory board (CAB), which assisted in recruitment and survey design.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by the Blue Cross Blue Shield of Michigan, grant #2021040025.MG. Julianne Armijo is funded in part by Predoctoral Fellowship Training Grant (T32 NR016914).
Footnotes
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr Marie-Anne Rosemberg is a guest editor for this Special Issue and was recused from the peer review and decision-making process. The authors declare no other potential conflicts of interest concerning the research, authorship, and/or publication of this article.
Ethical Considerations
The study was deemed exempt by the Institutional Review Board (IRB) at the University of Michigan (HUM00204532).
Consent to Participate
Respondents gave online consent online before starting the survey.
Data Availability Statement
The de-identified data can be made available upon reasonable request to the project PI.
References
- 1.Rudolph RR, Colligan M. National Institute for Occupational Safety and Health; Division of Biomedical and Behavioral Science; National Institute for Occupational Safety and Health; Education and Information Division; series: DHHS publication no. (NIOSH) 97–145. Published July 1997. Accessed September 10, 2024. https://stacks.cdc.gov/view/cdc/5177
- 2.CDC—NIOSH—NORA Healthy Work Design and Well-Being Cross-Sector Council. Updated May 31, 2023. Accessed September 10, 2024. https://www.cdc.gov/nora/councils/hwd/default.html
- 3.Reinganum MI, Thomas J. Shift work hazards. In: StatPearls. StatPearls Publishing; 2024. Accessed September 12, 2024. http://www.ncbi.nlm.nih.gov/books/NBK589670/ [PubMed] [Google Scholar]
- 4.Lee JJ, Moon HJ, Lee KJ, Kim JJ. Fatigue and related factors among hotel workers: the effects of emotional labor and non-standard working hours. Ann Occup Environ Med 2014; 26(1):51. doi: 10.1186/s40557-014-0051-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Boivin DB, Boudreau P. Interindividual variability in coherence between self-reported alertness and performance in shift workers. Sleep Health. 2024;10(1S):S63–S66. doi: 10.1016/j.sleh.2023.09.005 [DOI] [PubMed] [Google Scholar]
- 6.Wang XS, Armstrong MEG, Cairns BJ, Key TJ, Travis RC. Shift work and chronic disease: the epidemiological evidence. Occup Med 2011;61(2):78–89. doi: 10.1093/occmed/kqr001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Soltanzadeh A, Eyvazlou M, Mohammad-ghasemi M, et al. Investigating the relationship between shift work schedule and blood and metabolic parameters: a 10-years retrospective cohort study. Sci Rep 2024;14(1):17297. doi: 10.1038/s41598-024-68378-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Brown JP, Martin D, Nagaria Z, Verceles AC, Jobe SL, Wickwire EM. Mental health consequences of shift work: an updated review. Curr Psychiatry Rep 2020;22(2):7. doi: 10.1007/s11920-020-1131-z [DOI] [PubMed] [Google Scholar]
- 9.Ervasti J, Pentti J, Nyberg ST, et al. Long working hours and risk of 50 health conditions and mortality outcomes: a multi-cohort study in four European countries. Lancet Reg Health Eur 2021;11:100212. doi: 10.1016/j.lanepe.2021.100212 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Chen Y, Deng K, Hughes IM, et al. More than sleep problems? Testing five key health behaviors as reasons for quality of life issues among shift workers. Health Qual Life Outcomes. 2024; 22(1):52. doi: 10.1186/s12955-024-02269-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Batat H, Baniamer AZ, Hamasha AM, et al. The relationship between night shift work, sleep patterns, psychological wellbeing, and mental health among Jordanian healthcare workers. Arch Environ Occup Health. 2024;79(3–4):131–141. doi: 10.1080/19338244.2024.2379826 [DOI] [PubMed] [Google Scholar]
- 12.Lee JW, Lee JJ, Mun HJ, Lee KJ, Kim JJ. The relationship between musculoskeletal symptoms and work-related risk factors in hotel workers. Ann Occup Environ Med 2013;25(1):20. doi: 10.1186/2052-4374-25-20 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Seibt R, Süße T, Spitzer S, Hunger B, Rudolf M. Nutrition and health in hotel staff on different shift patterns. Occup Med 2015;65(6):477–484. doi: 10.1093/occmed/kqv068 [DOI] [PubMed] [Google Scholar]
- 14.Moon HJ, Lee SH, Lee HS, Lee KJ, Kim JJ. The association between shift work and depression in hotel workers. Ann Occup Environ Med 2015;27:29. doi: 10.1186/s40557-015-0081-0 [DOI] [Google Scholar]
- 15.Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med 2001; 16(9):606–613. doi: 10.1046/j.1525-1497.2001.016009606.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Spitzer RL, Kroenke K, Williams JBW, Löwe B. A brief measure for assessing generalized anxiety disorder: the GAD-7. Arch Intern Med 2006;166(10):1092–1097. doi: 10.1001/archinte.166.10.1092 [DOI] [PubMed] [Google Scholar]
- 17.Prins A, Bovin MJ, Smolenski DJ, et al. The primary care PTSD screen for DSM-5 (PC-PTSD-5): development and evaluation within a veteran primary care sample. J Gen Intern Med 2016;31(10):1206–1211. doi: 10.1007/s11606-016-3703-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Prins A, Ouimette P, Kimerling R, et al. The primary care PTSD screen (PC-PTSD): development and operating characteristics. Prim Care Psychiatry. 2003;9(1):9–14. doi: 10.1185/135525703125002360 [DOI] [Google Scholar]
- 19.Sekizuka H, Miyake H. Overtime work is related to nonrestorative sleep independently of short sleep time among a Japanese occupational population. Int Arch Occup Environ Health. 2024;97(1):75–80. doi: 10.1007/s00420-023-02027-x [DOI] [PubMed] [Google Scholar]
- 20.Bell T, Sprajcer M, Flenady T, Sahay A. Fatigue in nurses and medication administration errors: a scoping review. J Clin Nurs 2023;32(17–18):5445–5460. doi: 10.1111/jocn.16620 [DOI] [PubMed] [Google Scholar]
- 21.McNamara KA, Robbins WA. Shift work and sleep disturbance in the oil industry. Workplace Health Saf 2023;71(3):118–129. doi: 10.1177/21650799221139990 [DOI] [PubMed] [Google Scholar]
- 22.Alhassan S, Bariham I, Olu EL. Effects of front desk staff service quality on customer satisfaction in some selected hotels in Tamale Metropolis. Adv Soc Sci Res J 2024;11(2):471–485. doi: 10.14738/assrj.112.16255 [DOI] [Google Scholar]
- 23.Hilton MF, Whiteford HA. Interacting with the public as a risk factor for employee psychological distress. BMC Public Health. 2010;10(1):435. doi: 10.1186/1471-2458-10-435 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Marck CH, Ayton D, Steward T, et al. The workplace culture, mental health and wellbeing of early- and mid-career health academics: a cross-sectional analysis. BMC Public Health. 2024;24(1):1122. doi: 10.1186/s12889-024-18556-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Marković S, Kostić O, Terzić-Supic Z, et al. Exposure to stress and burnout syndrome in healthcare workers, expert workers, professional associates, and associates in social service institutions. Med Kaunas Lith 2024;60(3):499. doi: 10.3390/medicina60030499 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kefelew E, Hailu A, Kote M, Teshome A, Dawite F, Abebe M. Prevalence and associated factors of stress and anxiety among female employees of Hawassa Industrial Park in Sidama Regional State, Ethiopia. BMC Psychiatry. 2023;23(1):103. doi: 10.1186/s12888-023-04575-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Nagamine M, Giltay EJ, Shigemura J, et al. Assessment of factors associated with long-term posttraumatic stress symptoms among 56,388 first responders after the 2011 Great East Japan earthquake. JAMA Netw Open. 2020;3(9):e2018339. doi: 10.1001/jamanetworkopen.2020.18339 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.James KG. Returning to work after experiencing mental health problems. Ment Health Pract 2015;18(5):36–38. doi: 10.7748/mhp.18.5.36.e1011 [DOI] [Google Scholar]
- 29.Zhang TC, Torres E, Jahromi MF. Well on the way: an exploratory study on occupational health in hospitality. Int J Hosp Manag 2020;87:102382. doi: 10.1016/j.ijhm.2019.102382 [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The de-identified data can be made available upon reasonable request to the project PI.