Shift Work Disorder Index: initial validation and psychosocial associations in a sample of nurses

Abstract

Study Objectives:

Shift work is common yet does not always result in Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5)–defined shift work sleep disorder (SWD). This study reports on the reliability and validity of the DSM-5 informed Shift Work Disorder Index (SWDI), the presence of probable SWD in nurses, and demographic, sleep, and psychosocial correlates.

Methods:

Nurses (n = 454) completed the SWDI, psychosocial, and demographic questionnaires. Of the sample, n = 400 completed 14 days of sleep diaries, actigraphy, and additional questionnaires.

Results:

The global SWDI demonstrated excellent internal consistency (α = .94), as well as good convergent and divergent validity in the nurse sample. Thirty-one percent of nurses were past-month shift workers, with 14% (ie, 44% of shift workers) having probable SWD based on SWDI. Nurses who worked shift work and/or met SWD criteria were more likely to be younger and unmarried and less likely to have children than day workers and reported greater evening chronotype, insomnia, nightmares, and sleep-related impairment, greater depression, anxiety, posttraumatic stress, and perceived stress symptoms, as well as later and more variable sleep midpoint (actigraphy), shorter sleep duration (actigraphy, diaries), and lower sleep efficiency (diaries).

Conclusions:

The SWDI is an efficient and valid self-report assessment of DSM-5–defined SWD. Shift work and/or SWD were prevalent and associated with worse sleep and psychosocial health, particularly among nurses with probable SWD.

Citation:

Taylor DJ, Dietch JR, Wardle-Pinkston S, et al. Shift Work Disorder Index: initial validation and psychosocial associations in a sample of nurses. J Clin Sleep Med. 2022;18(10):2339–2351.

BRIEF SUMMARY

Current Knowledge/Study Rationale: The Shift Work Disorder Index is a brief, reliable, and valid screening measure for Diagnostic and Statistical Manual of Mental Disorders, fifth edition–defined shift work disorder. In the current study, almost 44% of shift-working nurses met criteria for probable shift work disorder. Shift work disorder, and in some cases shift work broadly, was associated with worse sleep health across numerous dimensions, increased other sleep disorder symptomatology, and elevated psychosocial distress including depression, anxiety, stress, and posttraumatic stress.

Study Impact: The Shift Work Disorder Index can provide screening among essential shift workers, which is urgently needed to improve sleep health and other psychosocial correlates in this sample. The Shift Work Disorder Index should be validated in additional shift-working populations, compared against clinical interview, and used to increase shift work disorder screening in both clinical and research applications.

INTRODUCTION

Shift work is associated with worse sleep,¹ mental and physical health,² and psychosocial outcomes³ and is considered a probable carcinogen by the World Health Organization.⁴ Shift work is particularly problematic for health care workers such as nurses, who are the front-line patient care providers in most hospital settings and during the COVID-19 pandemic. Studies demonstrate 50–89% of nurses work some form of shift work each month^5–9 and report significantly worse sleep (shorter sleep duration,¹⁰ greater insomnia symptoms,^11–15 and worse sleep quality¹³) and daytime functioning (fatigue¹⁴ and sleepiness¹⁵), compared to day-shift-working counterparts. Night shift nurses were also at increased risk of developing type II diabetes,¹⁶ breast cancer,¹⁷ coronary heart disease,¹⁸ and mortality.¹⁰

Working shift work or irregular hours can, but does not necessarily have to, lead to circadian rhythm sleep–wake disorder, shift work type (commonly known as shift work disorder, SWD).¹⁹ The Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5)¹⁹ defines SWD as a pattern of sleep disruption (ie, unconventional work hours) causing insomnia and/or excessive sleepiness, as well as clinically significant distress or impairment. To date, most research has focused on the consequences of shift work in general, but there has been little research comprehensively investigating a broad range of health and psychosocial correlates of SWD specifically.¹ It is likely that shift workers with SWD are accounting for most of the variance in previous studies showing shift workers have worse sleep,¹ mental and physical health,² and psychosocial outcomes.³ The World Health Organization⁴ indicated that understanding the consequences of shift work and SWD have been encumbered by the continued use of imprecise, inconsistent, and sometimes confounding definitions of shift work, as well as lack of specification between shift workers with and without SWD.

The most comprehensive epidemiological study to date comparing shift workers, with and without SWD,¹ utilized a 20-minute interview to determine if workers met criteria for International Classification of Sleep Disorders SWD.²⁰ They found 26.1% of rotating shift workers and 32.1% of night workers meet criteria for SWD.¹ Shift workers with SWD had significantly greater sleep-related impairments (ie, sleepiness, shorter total sleep time) and waking accidents compared to shift workers without SWD.¹ Additionally, shift workers with SWD were at elevated risk for a variety of physiological (eg, ulcers), mental (eg, depression), work-related (eg, absenteeism), and social (eg, missing family/social activities) consequences than shift workers without SWD.

Until 2007,²¹ there was no brief, reliable, and valid assessment instrument,²² which likely led to the paucity of studies investigating well-defined SWD. The first such attempt at developing a SWD measure was the Standard Shiftwork Index,²¹ which was not an independent measure, but instead a battery of self-report questionnaires compiled in response to a lack of standardization among shift work research. The Standard Shiftwork Index provided a comprehensive assessment of the severity and relationships between commonly reported complaints due to shift work in 1,532 nurses and midwives, but utility was limited by substantial participant burden (ie, 24 pages and 162 total items). In addition, the questionnaires were not designed to assess if the daytime issues were directly related to shift work.

Two other SWD measures were developed in 2012: the 23-item Bergen Shift Work Sleep Questionnaire²³ and the 4-item Shiftwork Disorder Screening Questionnaire.²⁴ The Bergen²³ questionnaire was created and validated in a sample of 760 nurses to comprehensively assess insomnia (ie, 30 minutes to fall asleep, trouble staying asleep, early awakening) and “tiredness/sleepiness” (ie, inadequate rest, sleepiness at work, after work, and on off days) impacted by particular types of work shifts (ie, day, evening, night, rest days). The Bergen questionnaire provided excellent assessment of sleep symptoms in relation to various shift work schedules, which were major components of the major diagnostic categories (ie, DSM-IV-TR,²⁵ International Classification of Sleep Disorders, second edition,²⁶ and International Classification of Diseases, tenth revision²⁷) at the time. However, its utility as a screener is limited by its length. In addition, the Bergen questionnaire does not specifically assess if “the sleep disturbance causes clinically significant distress or impairment in social, occupational, and other important areas of functioning,” which is a major diagnostic criteria of all diagnostic manuals.

The developers of the SWD Screening Questionniare²⁴ used discriminate function analyses to determine the diagnostic utility of 26 individual items (narrowed from 37 items in a previous study of police)²⁸ when compared to a physician diagnosis in 311 general shift workers. These researchers narrowed the number of items to 4 that mapped onto the major insomnia and sleepiness criteria of the International Classification of Sleep Disorders, second edition.²⁶ One drawback of the actuarial approach is that it can lack diagnostic specificity. For instance, the final 4-item measure does not actually assess difficulty falling asleep but instead only wake time after sleep onset and early morning awakening insomnia symptoms. This could be easily remedied by adding an additional fifth item (item 9 of the 26-item battery). However, similar to the Bergen measure,²³ the SWD Screening Questionnaire also does not specifically assess the major DSM criterion “the sleep disturbance causes clinically significant distress or impairment in social, occupational, and other important areas of functioning.” There is no single item (it is spread across multiple) in the original 26 that comprehensively assesses this criterion. Thus, to our knowledge, there is no brief, validated self-report measure to assess all of the criteria for DSM-5 SWD.¹⁹

Finally, it is imperative research be performed to comprehensively examine the physiological and psychosocial correlates of the additive impact of SWD above and beyond shift work. For instance, in the aforementioned study comparing shift workers with and without SWD,¹ several important domains were not assessed, such as perceived stress, anxiety, trauma symptoms, or alcohol use. These domains are all considered essential for epidemiological studies of insomnia according to leaders in this field.²⁹ In addition, the aforementioned shift work and SWD studies typically only used self-report, retrospective estimates of sleep, which could be biased and give relatively incomplete information.³⁰ Leaders in the insomnia field²⁹ also recommended sleep be assessed prospectively with sleep diaries or inferred/objective assessments (eg, actigraphy) in epidemiological and clinical studies. No such recommendations exist for the study of shift work or SWD, but it is apparent that the science would benefit from these more comprehensive and precise measurements of both sleep and daytime impairments.

In summary, lack of a precise screening tool for SWD and lack of a comprehensive correlate assessment has limited our understanding of the sleep and psychosocial health problems associated with SWD. To address the above gaps in the literature, the primary aim of the current study was to report on the development and initial validation of the Shift Work Disorder Index (SWDI). The SWDI was developed as a brief and face-valid measure of all the criteria of the DSM-5 SWD criteria. The secondary aim was to estimate the rates of probable SWD based on the SWDI. The tertiary aim was to replicate and extend previous work,¹ by comparing nurse shift workers with and without probable SWD on a comprehensive retrospective self-report assessment of psychosocial and sleep health and prospective objective and self-reported sleep parameters, in line with recommendations for a standard research assessment of insomnia.²⁹ These final analyses provide benchmarks for future studies of shift work and SWD.

METHODS

Participants

Participants (n = 454) were nurses at two area hospitals in the Dallas–Fort Worth metroplex who participated as part of a parent study investigating predictors of influenza vaccine response, Sleep and Vaccine Response in Nurses (SAV-RN) (R01AI128359-01; Principal Investigators: Taylor and Kelly). Inclusion criteria included being between the ages of 18 and 65 years, actively working a nursing shift at least part-time at one of the two hospitals, and planning to be vaccinated for the influenza virus. Exclusion criteria included being pregnant or having an egg allergy. Table 1 reports demographic characteristics for the entire sample. The majority of participants were female (91.0%, n = 413), White (78.3%, n = 351), non-Hispanic (89.8%, n = 405), currently married (61.2%, n = 278), and had children (63.7%, n = 289), mirroring characteristics of the nurse population of the area.

Table 1.

Participant characteristics by shift work disorder status on the Shift Work Disorder Index.

	Total Sample (n = 454)	Day Shift (n = 313; 68.9%)	Shift Worker without SWD (n = 79; 17.4%)	Shift Worker with SWD (n = 62; 13.7%)	F or Fisher’s exact	P	Part. η2 or V
	n (% of sample) or Mean (SD)	n (% of column) or Mean (SD)	n (% of column) or Mean (SD)	n (% of column) or Mean (SD)
Age*	38.9 (11.1)	40.3 (11.0)a	37.0 (11.3)b	34.2 (9.6)b	9.67	< .001	.041
Sex					1.90	.357	.061
Male	41 (9.0%)	25 (8.0%)	8 (10.1%)	8 (12.9%)
Female	413 (91.0%)	288 (92.0%)	71 (89.9%)	54 (87.1%)
Race†					7.73	.021‡	.131
White	351 (78.3%)	253 (80.8%)	57 (72.2%)	41 (66.1%)
Black/African American	31 (6.9%)	17 (5.4%)	7 (8.9%)	7 (11.3%)
Asian	51 (11.4%)	29 (9.3%)	13 (16.5%)	9 (14.5%)
Multiracial/other	15 (3.3%)	14 (4.5%)	2 (2.5%)	5 (8.1%)
Ethnicity*					5.85	.053	.115
Not Hispanic/Latino	405 (89.8%)	285 (91.9%)	69 (87.3%)	51 (82.3%)
Hispanic/Latino	46 (10.1%)	25 (8.1%)	10 (12.7%)	11 (17.7%)
Marital status†					22.68	< .001	.227
Married	278 (61.2%)	65 (21%)a	35 (44.3%)b	25 (40.3%)b
Single	125 (27.5%)	210 (67.7%)	38 (48.1%)	30 (48.4%)
Divorced/separated/widowed	51 (11.2%)	38 (12.3%)	6 (7.6%)	7 (11.3%)
Children					12.46	.002	.166
No children	165 (36.3%)	97 (31.3%)a	39 (49.4%)b	29 (46.8%)b
Has children	289 (63.7%)	216 (69.7%)	40 (50.6%)	33 (53.2%)
Number of children	1.4 (1.3)	1.5 (1.3)a	1.1 (1.3)b	1.2 (1.3)b	3.09	.046	.014

^a,bNonmatching letter pairs between columns indicate significant differences in proportions of group falling into row category based on post hoc analyses. *For age and ethnicity, n = 451. †Due to small cell sizes, statistical comparison for race was collapsed to White vs other and for marital status was collapsed to married vs other. ‡No significant post hoc between-group differences were found after applying Holm-Bonferroni adjustment. Probable shift work disorder = shift work disorder as defined by the Shift Work Disorder Index using Diagnostic and Statistical Manual of Mental Disorders, fifth edition criteria. SD = standard deviation.

Procedure

Nurse participants were recruited via nursing staff presentations, notification through employee email systems, and flyers that directed them to an initial online consent form as part of a larger parent study investigating the effects of sleep on influenza vaccination response. Following consent, participating nurses (n = 454) completed a baseline assessment questionnaire battery consisting of the questionnaires listed below via online (Qualtrics) survey (except for 3 measures, which were administered later). Participants were not compensated for the baseline assessment (ie, the current study) but were compensated if they chose to participate in the next phase of the study. Four hundred nurses moved on to the next phase of the study and completed 14 days of sleep diaries and actigraphy and then completed the Patient-Reported Outcomes Measurement Information System (PROMIS) sleep-related impairment³¹ questionnaire. All procedures were approved by the Medical City Plano Institutional Review Board.

Measures

Essential constructs measured included demographics, multidimensional sleep health, mental health, and stress. Other than the demographics form and the SWDI, all measures are commonly used and have good reliability and validity. Detailed descriptions of the measures can be found in the supplemental material. To make the results clinically meaningful and comparable to previous epidemiological studies, we created post hoc clinical groups (ie, met clinical cutoff levels on self-report questionnaires) on psychosocial variables.

Demographics

The General Health Questionnaire is an internally developed 14-item self-report measure used to collect data on age, race, ethnicity, height, weight, and health behaviors using both checklists and yes/no/open-ended questions.

Shift work

SWDI

The SWDI is a 6-item self-report screening measure developed specifically for this study because no other brief measure of DSM-5–defined SWD existed at the time. It was intended to evaluate the presence and severity of symptoms of SWD according to DSM-5 diagnostic criteria. Items assess night shift work frequency in the past month (ie, work between 9 pm and 6 am), wake time sleepiness associated with shift work, difficulty falling/staying asleep associated with shift work (ie, insomnia), distress related to shift work, social or occupational impairment associated with shift work, and duration that the individual has had a shift work schedule, with individual ratings ranging from 0 to 4 (anchors vary). The SWDI can be summed to obtain a total score ranging from 0 to 24, with higher scores indicating greater likelihood of and impairment due to shift work. The SWDI can also be used to categorize individuals as having probable SWD by meeting minimum criteria as defined by the DSM-5 (Table 2).¹⁹ The SWDI was developed by our lab in a format similar to the Sleep Condition Indicator (SCI),³² such that a minimum score is required on each item to classify an individual as having probable SWD. In the current study, the SWDI demonstrated excellent internal consistency (α = .94).

Table 2.

DSM-5 diagnostic criteria for circadian rhythm sleep–wake disorder and shift-work type.

Criterion	Text	SWDI Item	Qualifying Responses
A	A persistent or recurrent pattern of sleep disruption that is primarily due to an alteration of the circadian system or to a misalignment between the endogenous circadian rhythm and the sleep–wake schedule required by an individual’s physical environment or social or professional schedule	#1 Frequency	(2) 1–3 nights per week (3) 4–6 nights per week (4) 7 nights per week
B	The sleep disruption leads to excessive sleepiness or insomnia, or both	#2 Sleepiness OR #3 Insomnia	(2) Sometimes (3) Often (4) Always
C	The sleep disturbance causes clinically significant distress or impairment in social, occupational, and other important areas of functioning	#4 Distress OR #5 Impairment	(2) Somewhat (3) Much (4) Very Much
Specifier	Shift-work type: Insomnia during the major sleep period and/or excessive sleepiness (including inadvertent sleep) during the major awake period associated with a shift work schedule (ie, requiring unconventional work hours)	(all)
Specifier	Episodic: Symptoms last at least 1 month but less than 3 months Persistent: Symptoms last 3 months or longer	#6 Duration	(2) 1–3 months (3) 3–12 months (4) > 12 months

DSM-5 = Diagnostic and Statistical Manual of Mental Disorders, fifth edition.

Shift work status

Nurses who endorsed working at least one night shift per week between the hours of 9 pm and 6 am in the past month on the SWDI item #1 were classified as shift workers. All other nurses were classified as day workers.

SWD status

Nurses were classified as having probable SWD if they met minimum threshold for each of the primary DSM-5 criteria on the SWDI. Specifically, probable SWD was assigned if they reported 1 or more nights per week of night shift work, sleepiness OR insomnia associated with the work schedule, distress OR impairment associated with the work schedule, and at least 1 month of night shift work. See Table 2 for DSM-5 SWD criteria.

Convergent measures

To date, self-report measures or sleep diary in combination with clinical interview are required for the diagnosis of SWD.^5,33 Because a structured clinical interview for sleep disorders³⁴ was beyond the scope of the parent study, below we indicate how we investigated the convergences of the individual components of the SWD criteria assessed by the SWDI (ie, #1: frequency of shift work, #2: sleepiness; #3: insomnia; #4: distress, #5: impairment, and #6: duration) with measures of analogous constructs when applicable.

Sleep diary

The Consensus Sleep Diary³⁵ is a self-report prospective assessment of self-reported sleep patterns. Participants completed the Consensus Sleep Diary each morning for 14 days via REDCap on their smartphone device. Convergent validity was assessed by comparing SWDI item #1 (frequency) to an added sleep diary item asking “Approximately, what schedule did you work yesterday (your last wake period)?” Sleep midpoint was investigated for convergence with SWDI item #1 (frequency), and mean sleep efficiency was investigated for convergence with SWDI item #3 (insomnia).

Actigraphy

Actigraphy is an objective prospective measure of activity patterns and sleep used in conjunction with sleep diaries to assess participants’ sleep parameters (eg, duration, midpoint, wake after sleep onset). Participants wore an Actiwatch Spectrum device (Phillips Respironics, Andover, Massachusetts)³⁶ for 14 days. Actigraphy-derived mean and intraindividual variability of the sleep midpoint were investigated for convergence with SWDI item #1 (frequency).

SCI

The SCI³² is an 8-item self-report measure intended to evaluate insomnia symptoms according to DSM-5 diagnostic criteria.¹⁹ Convergent validity was assessed by comparing SWDI item #3 (insomnia) to SCI total score.

Patient-Reported Outcomes Measurement Information System sleep-related impairment

The Patient-Reported Outcomes Measurement Information System sleep-related impairment³¹ is an 8-item short measure of sleep-related complaints that is agnostic to underlying sleep condition and has been validated to assess the impact of sleep problems during wake. Convergent validity was assessed by comparing SWDI items #2 and # 5 (sleepiness and impairment) to the sleep-related impairment.

Discriminant measures

The following sleep measures were used as markers of discriminant validity with the SWDI total score because they are all sleep-related, but none should overlap significantly with SWD.

Nightmare Disorder Index

The Nightmare Disorder Index³⁷ is a 5-item self-report screening measure intended to evaluate symptoms (ie, frequency, characteristics, distress and impairment, and duration) of nightmare disorder according to DSM-5 diagnostic criteria. The Nightmare Disorder Index was used for discriminant validity because it is possible nightmares can cause some of the nocturnal (eg, insomnia) and daytime (eg, sleep-related impairment) symptoms associated with SWD.

Snoring, Tired, Observed, Blood Pressure Sleep Apnea Screen

The Snoring, Tired, Observed, Blood Pressure Sleep Apnea Screen³⁸ is a 4-item self-report screening questionnaire that assesses an individual’s risk for sleep apnea. The Snoring, Tired, Observed, Blood Pressure Sleep Apnea Screen was used for discriminant validity because it is possible apnea can cause some of the nocturnal (eg, insomnia) and daytime (eg, fatigue) symptoms associated with SWD.

Number of awakenings

Actigraphy-derived number of awakenings were used as a discriminant measure, as this is one of the few sleep parameters that would not be expected to correlate substantially with SWD.

Additional psychosocial measures

The following measures were considered primarily as measures of mental health and psychosocial correlates of shift work and SWD, following recommendations from the insomnia literature on standard assessments.²⁹ These measures could also alternatively be considered convergent validity with the SWDI item #5 (impairment) if one assumes they were as a consequence of shift work and SWD.

Patient Health Questionnaire-9

The Patient Health Questionnaire-9³⁹ is a 9-item self-report measure used widely to assess the severity of depressive symptoms. It assesses both affective and somatic symptoms related to depression and depressive disorders and corresponds to the diagnostic criteria for DSM-5 Major Depressive Disorder.¹⁹

Generalized Anxiety Disorder Screener-7

The Generalized Anxiety Disorder Screener-7⁴⁰ is a 7-item self-report measure used widely to screen for generalized anxiety disorder and assess the severity of anxiety symptoms.

Posttraumatic Stress Disorder Checklist for DSM-5

The PTSD Checklist for DSM-5⁴¹ is a 20-item self-report measure used widely to assess the symptoms of PTSD identified by the DSM-5.

Perceived Stress Scale

The Perceived Stress Scale⁴² is a 14-item self-report questionnaire used widely to assess the stress domains of unpredictability, lack of control, burden overload, and stressful life circumstances.

Quick Drinking Screen

The Quick Drinking Screen⁴³ is a 4-item self-report screening measure intended to identify people with hazardous or harmful patterns of alcohol consumption.

Analysis plan

Data were first screened for missingness, outliers, and assumption violations. There was almost no missing data (< .01%) on primary outcomes reported for the 454 participants in these analyses, and no major assumption violations or outliers for primary analyses were detected. Ten participants were excluded from analyses for missing data on the SWDI measure.

Internal consistency of the SWDI measure was assessed using coefficient alpha, assessed only for nurses who were classified as shift workers. Interitem and item-total correlations were calculated to assess the degree to which the SWDI items related to the larger construct (ie, SWD). Convergent validity (ie, association between the SWDI items and total score and measures of related characteristics; see measures section for details) was assessed using Spearman’s rho correlations (due to zero-inflated data on the SWDI) between the different SWDI items, SWDI total score, and measures listed above under “Convergent measures.” Similarly, discriminant validity (ie, association between the SWDI and theoretically unrelated measures) was assessed using Spearman’s rho correlations between the SWDI items, SWDI total score, and measures listed above under “Discriminant measures.”

We then separated nurses into groups based on their status on the SWDI: day shift, shift work with no/subthreshold SWD symptoms, and probable SWD. We next contrasted these three categories on demographics, sleep, and psychosocial health measures using one-way analysis of variance with Tukey post hocs or Fisher’s exact test with Holm-Bonferroni–adjusted Fisher’s exact post hocs (as appropriate). Holm-Bonferroni sequential corrections were applied to P values to adjust for multiple comparisons within measure group (ie, within psychosocial measures group, within sleep parameters group).⁵¹ For clinical relevance and comparability with past studies, scores for all applicable measures were categorized by clinical cutoffs for probable sleep and psychiatric disorders (see the measures section for cutoffs) and again compared by SWDI status. All tables report unadjusted rates and values unless noted. All analyses were conducted in SPSS, version 25.0 (IBM, Armonk, New York). Across all analyses, effect sizes were reported using either partial η², Cramer’s V, or odds ratios. Odds ratios were computed with day shift as the reference group.

RESULTS

Demographics

As seen in Table 1, 68.9% (n = 313) of the sample worked exclusively day shift and 31.6% (n = 141) of the sample worked at least one night shift per week during the past month. Of the full sample, 13.7% (43.9% of shift workers) met criteria for probable SWD on the SWDI. More than half of the shift-working sample worked 3 night shifts (n = 77; 54.6%) or 4 night shifts (n = 39; 27.7%) per week, with 9.9% working fewer (n = 14; 2 night shifts per week) and 7.8% working more (n = 11; 5–7 night shifts per week). Shift workers with probable SWD did not work more shifts per week on average (mean = 3.53, standard deviation = 1.04) compared to shift workers with no/subthreshold SWD (mean = 3.34, standard deviation = 1.02), t(139) = −1.09, P = .277.

Shift workers were significantly younger, less likely to be married, and less likely to have children compared to day workers. Shift workers had fewer children on average than day workers. Shift workers were less likely to be White than day workers, though post hoc differences were not statistically significant after adjusting for multiple comparisons. There were no significant differences between groups for sex or ethnicity.

Psychometrics of the SWDI

Reliability

The SWDI total score mean was 10.51 (standard deviation = 2.49) for those with shift work but no/subthreshold SWD symptoms and 16.16 (standard deviation = 2.64) for those who met full criteria for probable SWD based on the SWDI. The mean was not calculated for individuals who endorsed no shift work, or shift work less than one time per week. Reported correlations are just for individuals who endorsed shift work. Interitem and item-total correlations are shown in Table 3. Interitem correlations ranged from r = .77–.92, with an average interitem correlation of r = .86. The item-total correlation with the total SWDI was highest for the frequency item (r = .97) and lowest for the distress item (r = .86). Average item-total correlation was r = .92.

Table 3.

Interitem, item-total, and convergent correlations for the Shift Work Disorder Index.

	1	2	3	4	5	6	7	8	9	10	11	12	13
Interitem; Item-Total
1. SWDI frequency	—
2. SWDI sleepiness	.907**	—
3. SWDI insomnia	.833**	.882**	—
4. SWDI distress	.769**	.842**	.833**	—
5. SWDI impairment	.856**	.872**	.848**	.882**	—
6. SWDI duration	.916**	.837**	.789**	.716**	.807**	—
7. SWDI total	.965**	.955**	.909**	.864**	.918**	.913**	—
Convergent
8. Diary night shifts	.663**	.596**	.551**	.487**	.570**	.682**	.642**	—
9. Diary SE	−.244**	−.245**	−.290**	−.278**	−.298**	−.250**	−.270**	−.249**	—
10. Actigraphy SM (mean)	.614**	.557**	.521**	.473**	.562**	.646**	.610**	.720**	−.249**	—
11. Actigraphy SM (IIV)	.598**	.556**	.511**	.448**	.533**	.603**	.593**	.692**	−.263**	.710**	—
12. Insomnia (SCI)	−.180**	−.254**	−.351**	−.341**	−.295**	−.198**	−.266**	−.106*	.411**	−.186**	−.157**	—
13. PROMIS SRI	.236**	.303**	.312**	.305**	.315**	.242**	.292**	.225**	−.283**	.267**	.249**	−.520**	—

**P < .01; *P < .05; (n = 454). Underlined items indicate analogous constructs at the item level. IIV = intraindividual variability, PROMIS = Patient-Reported Outcomes Measurement Information System, SRI = Sleep Related Impairment, SCI = Sleep Condition Indicator, SE = sleep efficiency, SM = sleep midpoint, SWDI = Shift Work Disorder Index.

Convergent validity

Correlations to establish convergent validity are shown in Table 3. The SWDI total score correlations with convergent sleep and circadian measures ranged from small (r = −.27) to large (r = −.64). The largest correlation for the total SWDI score was with the reported days of shift work on 14-day sleep diary (r = .64) and the smallest correlation was for the SCI measure of insomnia (r = −.27 [lower scores indicate greater insomnia symptom severity]). The SWDI frequency item had a moderate correlation (r = .66) with summed 14-day sleep diary number of night shifts worked. The SWDI insomnia item demonstrated a moderate (r = −.35) correlation with the SCI measure of insomnia and a small (r = −.29) correlation with sleep diary sleep efficiency. The SWDI sleepy, distress, and impairment items correlated moderately with the Patient-Reported Outcomes Measurement Information System measure of sleep-related impairment (rs = .30, .31, and .32, respectively).

Discriminant validity

Correlations to establish discriminant validity are shown in Table 4. The SWDI total score demonstrated small to moderate correlations with sleep related discriminant measures of sleep constructs (rs = .06–.32).

Table 4.

Discriminant correlations for the SWDI.

	1. SWDI Frequency	2. SWDI Sleepiness	3. SWDI Insomnia	4. SWDI Distress	5. SWDI Impairment	6. SWDI Duration	7. SWDI Total
Discriminant sleep and circadian
NDI	.013	.081	.092	.125**	.091	.039	.069
STOP	.081	.107*	.112*	.177**	.132**	.070	.111*
Tired	−.002	−.025	−.036	−.009	−.016	−.017	−.012
Actigraphy number of awakenings	−.074	−.032	−.043	−.009	−.011	−.093	−.063

**P < .01; *P < .05. NDI = Nightmare Disorder Index, STOP = Snoring, Tired, Observed, Blood Pressure Sleep Apnea Screen, SWDI = Shift Work Disorder Index.

Psychosocial health

Correlations with additional psychosocial health variables are shown in Table 5. The SWDI total score demonstrated small, statistically significant correlations (rs = .17–.29) with measures of mental health (ie, depression, anxiety, PTSD, stress) and small nonsignificant correlations with number of alcoholic drinks per week (r = −.03).

Table 5.

Psychosocial Spearman’s rho correlations with the Shift Work Disorder Index domains and total score.

	1. SWDI Frequency	2. SWDI Sleepiness	3. SWDI Insomnia	4. SWDI Distress	5. SWDI Impairment	6. SWDI Duration	7. SWDI Total
Depression (PHQ-9)	.214**	.300**	.341**	.368**	.329**	.221**	.293**
Anxiety (GAD-7)	.104*	.166**	.199**	.264**	.230**	.112*	.172**
Posttraumatic stress (PCL-5)	.105*	.162**	.184**	.221**	.211**	.112*	.165**
Perceived stress (PSS)	.151**	.215**	.242**	.301**	.286**	.130**	.213**
QDS, d/wk	−.054	−.029	.006	.029	.006	−.032	−.027

**P < .01; *P < .05. (n = 454). GAD-7 = Generalized Anxiety Disorder Screener, PCL-5 = PTSD Checklist for Diagnostic and Statistical Manual of Mental Disorders, fifth edition, PHQ-9 = Patient Health Questionnaire, PSS = Perceived Stress Scale, QDS = Quick Drinking Screen.

Group comparisons

Table 6 and Table 7 display analyses comparing day shift workers, night shift workers without SWD, and night shift workers with SWD. As seen in Table 6, shift workers had significantly lower sleep efficiency and later sleep midpoint on the sleep diary and significantly shorter total sleep time on both sleep diary and actigraphy. Actigraphy-derived sleep midpoint was around 1 hour later for shift workers with probable SWD, whose sleep midpoint was in turn over 3.5 hours later than day workers. Actigraphy sleep efficiency did not differ between groups. Insomnia symptoms, nightmare disorder symptoms, and sleep-related impairment were significantly greater for shift workers with probable SWD compared to those with no/subthreshold SWD and day workers. Models presented in Table 6 are unadjusted for covariates. All models were rerun adjusted for statistically significant covariates (age, marital status, have children) and did not impact significance of the results, except for the Snoring, Tired, Observed, Blood Pressure Sleep Apnea Screen (which became significant upon addition of covariates, F = 3.81, P = .002, part η² = .041).

Table 6.

Sleep and psychosocial characteristics by shift work disorder status on the Shift Work Disorder Index.

	Total Sample (n = 454)	Day Shift (n = 313)	Shift Worker without SWD (n = 79)	Shift Worker with SWD (n = 62)	F (df)	P	Part. η2
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Prospective Sleep Variables*
Sleep diary SM (time)	4.13 (2.40)	3.04 (1.28)a	6.38 (3.00)b	7.29 (2.38)b	165.98 (2, 393)	< .001	.458
Sleep diary SE (%)	90.92 (5.18)	91.72 (4.75)a	89.51 (5.79)b	88.30 (5.44)b	12.95 (2, 393)	< .001	.062
Sleep diary TST (m)	431.63 (49.78)	440.68 (43.46)a	408.38 (51.56)b	411.68 (63.99)b	17.13 (2, 393)	< .001	.080
Actigraphy SM (time)	4.16 (2.44)	3.05 (1.30)a	6.43 (2.57)b	7.47 (2.46)c	172.93 (2, 389)	< .001	.471
Actigraphy SE (%)	86.94 (4.91)	87.17 (4.88)	86.61 (4.51)	86.08 (5.50)	1.19 (2, 389)	.305	.006
Actigraphy TST (m)	401.78 (50.98)	408.82 (48.10)a	383.86 (53.49)b	385.11 (54.94)b	9.65 (2, 389)	< .001	.047
Retrospective Sleep/Circadian Measures
Insomnia (SCI)	21.69 (7.03)	22.62 (6.61)a	22.43 (6.81)a	16.03 (6.85)b	25.80 (2, 450)	< .001	.103
PROMIS sleep-related impairment T-score*	48.38 (8.05)	47.2 (7.80)a	49.02 (7.22)a	54.57 (7.78)b	18.27 (2, 381)	< .001	.087
Nightmares (NDI)	2.50 (3.27)	2.33 (3.02)a	1.99 (3.19)a	4.00 (4.09)b	8.14 (2, 451)	< .001	.035
Sleep apnea risk (STOP)	0.95 (0.81)	0.92 (0.81)	0.95 (0.89)	1.13 (0.69)	1.77 (2, 451)	.171	.008
Psychosocial Measures
Depression (PHQ-9)	3.65 (3.95)	3.05 (3.37)a	3.49 (3.86)a	6.88 (5.12)b	27.29 (2, 451)	< .001	.108
Anxiety (GAD-7)	2.79 (3.45)	2.53 (3.26)a	2.1 (2.89)a	5.00 (4.18)b	16.22 (2, 451)	< .001	.067
Posttraumatic stress (PCL-5)	19.06 (10.55)	17.96 (9.18)a	16.9 (8.47)a	27.34 (14.87)b	24.78 (2, 450)	< .001	.099
Perceived stress (PSS)	12.32 (6.31)	11.64 (5.73)a	11.43 (6.55)a	16.89 (6.96)b	20.45 (2, 451)	< .001	.083
Quick Drinking Screen, d/wk	1.62 (1.65)	1.68 (1.69)	1.35 (1.51)	1.58 (1.61)	1.23 (2, 450)	.293	.005
Quick Drinking Screen, drinks per week	2.95 (3.90)	2.92 (3.69)	2.91 (4.00)	2.98 (4.66)	0.08 (2, 450)	.992	.000

^a,b,cNonmatching letter pairs between columns indicate significant differences in proportions of group falling into row category based on post hoc analyses. Models presented here are unadjusted for covariates. Models were run adjusted for covariates (ethnicity, marital status) but these adjustments did not impact significance of the results; therefore, they are not reported here. *These measures were only collected on n = 400. df = degrees of freedom, GAD-7 = Generalized Anxiety Disorder Screener, NDI = Nightmare Disorder Index, PHQ-9 = Patient Health Questionnaire, PCL-5 = Posttraumatic Stress Disorder Checklist for Diagnostic and Statistical Manual of Mental Disorders, fifth edition, PROMIS = Patient-Reported Outcomes Measurement Information System, PSS = Perceived Stress Scale, SCI = Sleep Condition Indicator, SD = standard deviation, SE = sleep efficiency, SM = sleep midpoint, STOP = Snoring, Tired, Observed, Blood Pressure Sleep Apnea Screen, TST = total sleep time.

Table 7.

Percentage of nurses above clinical cutoffs for probable sleep and psychiatric disorders.

	Day Shift (n = 313)	Shift Worker with No/ Subthreshold Shift Work Disorder (n = 79)	OR vs Day Shift (95% CI)	Probable Shift Work Disorder (n = 62)	OR vs Day Shift (95% CI)	Fisher’s exact	P	V
	% (n)	% (n)		% (n)
Sleep/Circadian
Insomnia disorder (SCI)	20.1% (63)a	20.3% (16)a	1.01 (0.55–1.86)	59.7% (37)b	5.87 (3.3–10.47)	38.29	< .001	0.311
Nightmare disorder (NDI)	5.4% (17)a	7.6% (6)ab	1.43 (0.55–3.76)	17.7% (11)b	3.76 (1.66–8.48)	9.60	.006	0.158
Sleep apnea high risk (STOP)	16.3% (51)	17.7% (14)	1.11 (0.58–2.12)	21.0% (13)	1.36 (0.69–2.69)	0.95	.609	0.042
Psychosocial Measures
Depression (PHQ-9)	6.1% (19)a	8.9% (7)a	1.50 (0.61–3.72)	27.4% (17)b	5.85 (2.83–12.08)	21.48	< .001	0.246
Anxiety (GAD-7)	6.4% (20)a	5.1% (4)a	0.78 (0.26–2.35)	19.4% (12)b	3.52 (1.62–7.64)	10.46	.005	0.169
Posttraumatic stress (PCL-5)	8.9% (28)a	3.8% (3)a	0.40 (0.12–1.36)	30.6% (19)b	4.50 (2.31–8.75)	23.99	< .001	0.257
Severe stress (PSS)	1.3% (4)a	0.0% (0)a	—	12.9% (8)b	11.44 (3.33–39.33)	18.63	< .001	0.256

^a,bNonmatching letter pairs between columns indicate significant differences in proportions of group falling into row category based on post hoc analyses. OR = Odds Ratio, CI = confidence interval, GAD-7 = Generalized Anxiety Disorder Screener, NDI = Nightmare Disorder Index, PCL-5 = PTSD Checklist for Diagnostic and Statistical Manual of Mental Disorders, fifth edition, PHQ-9 = Patient Health Questionnaire, PSS = perceived stress, SCI = Sleep Condition Indicator, STOP = Snoring, Tired, Observed, Blood Pressure Sleep Apnea Screen.

Odds of probable insomnia disorder (Table 7) based on the SCI among shift workers with probable SWD were almost 6-fold greater than day workers; differences between shift workers with no/subthreshold SWD symptoms and day workers did not differ. Odds of probable nightmare disorder among shift workers with probable SWD were almost 4-fold greater than day workers but did not differ from shift workers with no/subthreshold SWD symptoms. There were no significant differences in sleep apnea symptoms between groups.

Shift workers with probable SWD had greater symptoms of depression, anxiety, posttraumatic stress, and perceived stress compared to shift workers with no/subthreshold SWD and day workers, who did not differ from one another on these measures (Table 6). A similar pattern was seen for odds of being above clinical cutoffs on each of these measures (Table 7). Odds for shift workers with probable SWD compared to day workers were almost 6-fold greater for depressive symptoms, over 3-fold greater for anxiety symptoms, over 4-fold greater for posttraumatic stress, and over 11-fold greater for severe stress.

DISCUSSION

The SWDI is a brief, reliable, and valid assessment of DSM-5–defined SWD, particularly among nurses. The SWDI had excellent internal consistency (α = .94) and interitem (mean r = .86) and mean item-total (r = .92) correlations, especially for a brief diagnostic index measure with distinct facets. Convergent correlations between the SWDI and measures of similar domains of SWD (eg, daytime sleepiness, insomnia) were moderate to high. Divergent correlations between the SWDI with theoretically different sleep domains including nightmares, apnea, and number of awakenings were small and mostly nonsignificant. Taken as a whole, these results provide preliminary evidence that the SWDI is a reliable and valid measure of the distinct construct of SWD.

Probable DSM-5 SWD (44.0%) among shift workers in the current sample was higher than the rate of 26.5% identified in a recent meta-analysis⁴⁴ of shift workers across professions, though the range of included studies was 2.3–84%. This inconsistency is likely due to the previously noted “inconsistent definitions of shiftwork”⁴ in work schedules, variations in the samples assessed (eg, occupation, country), and lack of daytime complaint items in existing measures.

Shift-working nurses with probable SWD were more likely to be non-White compared to those with no/subthreshold SWD, who were in turn more likely to be non-White compared to day workers. These post hoc results were nonsignificant after statistical correction, which may be attributable to unequal cell sizes and thus insufficient power. This is similar to previous findings from the Bureau of Labor Statistics⁴⁵ indicating Black individuals were more likely to work shift work than any other racial/ethnic group. One likely source of racial disparities among shift workers, and especially those with more severe SWD symptoms, is minority stress associated with racism and discrimination. This is an important domain for consideration in future studies.

Shift workers demonstrated worse sleep health across almost every measured dimension including lesser sleep efficiency on sleep diary and shorter total sleep time and later sleep midpoint on sleep diary and actigraphy. Actigraphy-derived sleep midpoint was significantly later in nurses with probable SWD than nurses with no/subthreshold SWD, with a similar effect size for sleep diary–derived sleep midpoint though this difference was not statistically significant. Notably, however, shift workers with probable SWD did not work significantly more night shifts than shift workers with no/subthreshold symptoms and thus did not differ on other prospective measures of sleep. As indicated below, despite few prospective sleep differences, the probable SWD group reported significantly more daytime impairment complaints (eg, “clinically significant distress or impairment in social, occupational, and other important areas of functioning.”) than the no/subthreshold SWD, which is a critical SWD criteria missing from existing screening instruments. This requires further investigation, as there is likely a reciprocal relationship between later sleep timing and severity of sleep disturbance and daytime impairment regardless of night shifts worked.

Consistent with previous studies,^11,15 shift workers, and particularly those with probable SWD, demonstrated worse sleep health compared to day workers on questionnaires assessing sleep disorders symptoms (eg, insomnia, sleep-related impairment, nightmares). Shift workers with probable SWD had 6-fold increased odds of meeting criteria for probable DSM-5 insomnia disorder compared to day-shift workers and shift workers with no/subthreshold SWD. One previous study found that current shift workers were approximately 1.5 times more likely to be categorized with insomnia than their peers with no shift work experience.¹² Further, nurses who had previous, not current, night shift experience were also approximately 1.4 times more likely have insomnia symptoms than their peers with no shift work experience.¹² This is a complex relationship, as shift work can cause insomnia symptoms and insomnia is one criterion of SWD (although it is not mandatory; excessive sleepiness can also meet this criterion). Unfortunately, self-report questionnaires cannot reliably differentiate between insomnia and SWD, thus future studies including clinical interviews are necessary.

Nurses with probable SWD were at 4-fold increased odds for probable nightmare disorder compared to day workers. The rate of probable nightmare disorder among the total sample of nurses (7.6%) and specifically nurses with probable SWD (17.7%) was elevated compared to the general population (4.0%).⁴⁶ Possible sources of this elevation include increased sleep fragmentation and PTSD symptomatology.

Correlations between the SWDI and measures of psychosocial distress were large and significant for all domains but the Quick Drinking Screen. These results were unsurprising given prior findings that shift work and SWD are associated with poor mental health and distress.^1,3 Shift workers with probable SWD had higher depression, anxiety, PTSD, and perceived stress symptoms. Other studies have found that night shift work is associated with greater risk for depression, anxiety, and higher stress, but a recent review among nurses specifically demonstrated mixed findings.⁴⁷

Limitations and future directions

Our study does not offer a true prevalence estimate due to sampling method and should not be considered a definitive rate of probable SWD among nurses. The data were collected from only two hospitals in a single geographical region (northern Texas), which limits generalizability. Future larger-scale studies using random sampling methods are needed to determine accurate prevalence estimates. Future studies should also aim to recruit equal numbers of and night- and day-shift workers across ages to facilitate comparison and strive to correct for the effects of selection bias. Future studies should examine the psychometrics of the SWDI in varied shift working populations and compare the measure to a gold standard for diagnosis of DSM-5 SWD (ie, clinical interview) and other established screening measures (eg, Bergen Questionnaire, SWD Screening Questionnaire).

Our overall sample demographics generally matched national demographics of nurses in terms of sex and race/ethnicity,⁴⁸ but could be affected by selection bias, as much research on night work is influenced by a ‘‘healthy worker’’ effect, where employees struggling with night work are likely to stop working nights early in their career. Consequently, long-term night workers represent a relatively healthy part of the population.^12,49

Although we collected extensive information on mental health using validated questionnaires, we collected limited information regarding medical conditions, medication use, and nurses’ day-to-day roles (eg, administrative, shift-lead, worker) and specialties. These may be important moderators of the associations we observed. Investigators should also continue to examine whether shift work, sleep disturbances, and their combination have direct, measurable consequences on physical health outcomes (eg, inflammation).

CONCLUSIONS

This work is a crucial step that contributes to understanding the unique challenges faced by nurses, a key population who serves as the front line of patient care. The employment of registered nurses is projected to grow 7% between 2019 and 2029, which is faster than the average for all occupations.⁵⁰ As such, the burden and prevalence of night-shift work and SWD among nurses is also likely to increase. It is essential to better understand the causes and consequences of shift work in nurses to design more effective intervention efforts. Shift workers, and particularly those with SWD, are more likely to have various types of sleep disturbances as well as higher levels of depression, anxiety, PTSD symptoms, and perceived stress. Given the unique burden of shift work on health, it is critical both institutional and individual level actions are taken to mitigate its negative effects.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Work for this study was performed at the Department of Psychology, University of North Texas, Denton, TX. This study was funded by the National Institute of Allergy and Infectious Diseases, 1R01AI128359-01 (Principal Investigators: Taylor and Kelly). The sponsor had no role in the collection, analysis, or interpretation of data. Dr. Dietch was supported by 2T32MH019938-26A1 (Principal Investigator: Schatzberg). The authors report no conflicts of interest.

ABBREVIATIONS

DSM

Diagnostic and Statistical Manual of Mental Disorders

PTSD

posttraumatic stress disorder

SCI

Sleep Condition Indicator

SWDI

Shift Work Disorder Index

SWD

shift work sleep disorder