Examining Insomnia Disorder and Stress Generation Among Individuals Who Have Experienced Involuntary Job Loss

Abstract

Objective.

Depressed individuals generate more stressful life events than non-depressed individuals. Like depressive symptoms, the symptoms of insomnia disorder may lead to impaired decision-making, daytime sleepiness, fatigue, and emotion dysregulation, yet the prospective relationship with insomnia disorder and stress generation has not previously been investigated. We hypothesized that insomnia disorder within the first three months of involuntary job loss would lead to an increased number of stressful life events three-months later.

Methods.

This project employed a longitudinal design consisting of two timepoints spaced by approximately 3 months. A sample 136 participants with complete data was sourced from the Assessing Daily Activity Patterns through Occupational Transitions study. Insomnia disorder was diagnosed using the Duke Structured Interview for Sleep Disorders, and the number of stressful life events was assessed using the Life Events and Difficulties Schedule.

Results.

A cross-lagged panel analysis utilizing Poisson and logistic regression techniques indicated that insomnia disorder at study baseline predicted an increased number of all stressful life events at follow-up (RR = 1.36, p = .01); conversely, stressful life events at baseline did not predict insomnia disorder (OR = .98, p = .87).

Conclusion.

These results support a stress-generation hypothesis of insomnia disorder. Findings highlight insomnia disorder as a potential target for intervention in the prevention of additional stress exposure among recently unemployed individuals, who have been shown to be at increased risk for adverse health and health disparities.

Introduction

Stressful life events occur across the lifespan and can pose a threat to one’s social or economic status, identity, relationships, functioning, and physical or psychological well-being [1–3]. Individuals with an increased number or severity of stressful life events are at higher risk for adverse health outcomes, including trauma and stressor-related disorders [4] and other psychopathology, cardiovascular diseases, infectious diseases, and poorer outcomes after cancer treatment [2]. They are also at an increased risk for insomnia disorder [5].

Some theoretical [6] and empirical evidence [5, 7] supports a causal relationship between stressful life events and insomnia disorder. Onset of insomnia disorder symptoms is often preceded by a stressful life event [7], and individuals with insomnia disorder report higher distress after experiencing stressful life events than healthy counterparts [5]. Insomnia disorder is characterized by frequent, consistent dissatisfaction with sleep quantity or quality, difficulty initiating sleep (increased sleep latency), frequent awakenings from sleep, difficulties falling back to sleep after nocturnal awakenings, non-restorative sleep, and/or spontaneous early-morning awakenings [8]. These symptoms can result in increased daytime fatigue and sleepiness [9], cognitive complaints [10], emotion dysregulation [11], and increased risk of accidents at home, work, and while driving [12]. In addition, individuals with more severe insomnia symptoms are more likely to use maladaptive coping strategies, such as venting, behavioral disengagement, self-blame [13]. Sleep loss has also been associated with impulsive decision-making [14] as well as avoidance [15].

While theoretical models commonly incorporate stressful life events as a contributing factor to insomnia disorder [5, 6], very little is known about whether insomnia disorder is a contributing causal factor to stressful life events. In the depression literature, research has shown that depressed individuals are more likely to generate stressful life events [16]. The stress generation hypothesis posits that individuals are more likely to experience stressful life events following a depressive episode [16] due to impediments to normative cognitive processes and increasing the risk of maladaptive coping strategies associated with depression [17]. Decades of research appear to support the stress generation hypothesis, as demonstrated by Liu and Alloy’s systematic review of 57 articles addressing the stress generation hypothesis [18].

It is possible that, like depression and sleep loss, insomnia disorder may also lead to increased number of stressful life events due to symptoms which impair decision-making, concentration, executive functioning [19], and communication skills [20]. For example, an individual experiencing fatigue and difficulty with emotion regulation [21] might inadvertently escalate a tense interpersonal exchange into a major interpersonal conflict. As another example, a person with insomnia might be more likely to be late for work in the morning leading to a poor job evaluation and eventual job loss. To our knowledge, no study has examined the potential causal relationship between insomnia disorder and the number of stressful life events.

We examined whether insomnia disorder at study baseline would be significantly associated with an increased number of stressful life events at three-month follow-up in this sample. Based on past research on the stress generation hypothesis [17], we predicted that the positive association between insomnia disorder and the number of stressful life events would be replicated for events which were rated dependent or interpersonal in nature. Interpersonal events refer to events in which the core of the event involves social relationships (e.g., arguments between family members, extramarital affairs), while dependent events refer to events for which the individual is at least partly responsible, such as experiencing termination due to purposeful professional misconduct or experiencing a car accident while driving under the influence of intoxicants [22]. If insomnia disorder is found to lead to increased generation of stressful life events, then results would support the deployment of sleep health intervention in the aftermath of involuntary job loss as one way to prevent the accumulation of stress and to mitigate any poor social, emotional, and biological health outcomes.

Materials and Methods

Sample

Data were obtained from the Assessing Daily Activity Patterns through Occupational Transitions (ADAPT) study. Details about the ADAPT study protocol and methodology have been published elsewhere [23]. In brief, ADAPT is a longitudinal observational study with an 18-month follow-up period which examined factors related to sleep, mental health, social rhythms, and obesity in individuals who had recently experienced involuntary job loss. Participants were eligible individuals between the ages of 25 and 60 who had involuntarily lost their jobs in the three months preceding the baseline visit [23]. Written informed consent was obtained at the screening visit prior to the full assessment of eligibility [23].The current analysis examined data gathered at baseline and three-month follow-up interviews from all ADAPT study participants who completed the Life Events and Difficulties Schedule [22] at their three-month follow-up visit (N = 136).

Measures

Demographics.

Demographic characteristics, including age, gender, race, ethnicity, and educational attainment, were collected during a structured demographic interview at the baseline visit. To account for sparseness in certain categories and missingness in the race variable in participants who identified as Hispanic (n=5), the race and ethnicity variables were combined into one binary (White/Non-White) variable for the primary analysis described in Table 1.

Table 1.

Descriptive characteristics of Daily Activity Patterns through Occupational Transitions (ADAPT) participants who completed the Life Events and Difficulties Schedule at 3-month follow-up, by insomnia disorder diagnosis at study baseline (N = 136).

Characteristic	Insomnia n=77		No insomnia n=59		All participants N=136
	n	%	n	%	n	%
Sex
Female	50	65	36	61	86	63
Male	27	35	23	39	50	37
Racea
American Indian/Alaska Native	3	4	1	2	4	3
Asian	1	1	0	0.0	1	1
Black/African American	2	3	4	7	6	5
Native Hawaiian/Other Pacific Islander	1	1	0	0	1	1
White	53	74	44	75	97	74
Other	7	10	2	3	9	7
>1 race	5	7	8	14	13	10
Ethnicity
Non-Hispanic	48	81	44	57	92	68
Hispanic	11	19	33	43	44	32
Educational attainment
No college degree	42	55	22	37	64	47
Associate degree or higher	35	46	37	63	72	53

^an=5 missing

Insomnia disorder.

The exposure of interest, insomnia disorder, was considered present if participants met the International Classification of Sleep Disorders (ICSD-3) criteria for a current chronic or acute insomnia disorder on the Duke Structured Interview for Sleep Disorders (DSI) at study baseline. The DSI is a valid and reliable assessor-administered interview that allows for systematic assessment of medical sleep disorder symptoms using the ICSD-2 criteria [24, 25]. The DSI was updated using ICSD-3 criteria for this study [26]. To meet the threshold for insomnia disorder, reported symptoms must have occurred at least three times per week and be associated with daytime fatigue, sleepiness and impaired concentration causing distress or impairment in social, occupational, educational, academic, behavioral, or other domains [27]. If symptoms resolved within three months, acute insomnia disorder was diagnosed. If symptoms persisted for three or more months, chronic insomnia disorder was determined. However, for the purposes of this analysis, the acute and chronic insomnia disorder diagnoses were collapsed into a binary, 0/1 insomnia disorder variable (0 = no insomnia, 1 = acute or chronic insomnia disorder) to enhance generalizability of findings to all individuals presenting with significant insomnia complaints.

All DSI interviews were video or audio recorded. All assessors held at least a bachelor’s level degree, and assessor training followed a standardized protocol. To be considered competent and reliable, assessors were required to rate a minimum of five recorded interviews and achieve a kappa of .75 [28, 29] for each major diagnosis with a trained and reliable masters-level clinician. For the diagnosis of acute and chronic insomnia, each interviewer achieved excellent inter-rater reliability on the most recent of five training videos (kappa = 1.0, area under the curve = 1.0). The high kappa and AUC for these diagnoses is likely attributable to the highly structured nature of the interview.

Stressful life events.

The primary outcome was the number of rated, stressful life events reported at three-month follow-up using the Life Events and Difficulties Schedule (LEDS) [1]. LEDS is a semi-structured contextualized life events interview that collects information about the nature, date, and duration of life events. The focus of the LEDS interview and subsequent event ratings is on objective elements such as the frequency, duration, and severity of events, and not on the participant’s subjective evaluation of the event. The LEDS is considered the gold standard in life events assessment because it assesses each event’s impact on a typical person under similar circumstances, avoiding the systematic overreporting of negative items found in high-distress samples [30, 31].

The LEDS interview was administered by study staff who completed a standardized training protocol. Participants were asked about all events that occurred in the last three months (baseline) or since the participant’s last visit (follow-up). A three month interval was chosen because longer intervals are associated with fewer recalled events [32]. Following the interview, interviewers prepared vignettes for each event based on the information collected to present at a consensus panel meeting of at least three raters. Raters were required to rate a minimum of five recorded interviews and achieve a kappa of .75 [28, 29] with a trained and reliable senior rater.

An event was rated as a stressful life event if it met the domain-specific inclusion threshold as a stressful event in the LEDS manual [22] and was an event of sufficient severity (rated at least 4 = none or little long-term threat) that happened to the participant, close tie, or a friend/wider kin that was directly involved in the event. Examples of rated, stressful life events range in severity. They may include events with significant long-term threat, such as job loss, being evicted, losing a loved one to death or divorce, being the victim of an assault or burglary, or being diagnosed with chronic illness. Events were also included if they resolved within 14 days or were positive or milestone events; examples include being picked-up by police without charges, starting a new job, or getting married.

Dependent events were defined as events that occurred due to the possible influence, an intentional act, or probable negligence by the participant [22]. Dependent events also included arguments or purposeful termination of contact with a close other [22]. Interpersonal events were rated as such only if the primary characteristic of the event was its interpersonal nature [22]. For example, while an argument with a spouse would be rated as interpersonal, an argument at work would not be interpersonal if the conflict was not about a personal insult but rather because of a disagreement on a professional task.

Statistical Analyses

All data were analyzed using the SAS 9.4 statistical analysis program [33]. A cross-lagged panel analysis was employed to examine the relationship between insomnia disorder at study baseline and number of stressful life events reported at three-month follow-up, as well as to assess causal predominance of the association. Cross-lagged panel analysis utilizes a type of discrete time path analysis model to analyze repeated measures data in which two or more variables are each measured at two or more different time points [34]. The technique facilitates estimation of the directional influence of one variable over another at different timepoints [34] and has been used in a number of disciplines, including economics, communication, and the health sciences [34, 35]. An advantage of this design is that it estimates both inertial effects (i.e., the association between a variable at time 1 and the same variable at time 2) and possible reciprocal effects (i.e., the influence of a variable at time 1 on another variable at time 2 and vice versa). This technique was chosen with consideration of sample size and non-continuous dependent variables.

Preliminary analyses.

The covariates for this study were selected a priori and included age (years), sex (male/female) [36], race/ethnicity (White/Non-White) [36], and educational attainment as a proxy for socioeconomic status (at least associate’s degree/no postsecondary degree) [36]. Covariates were selected based on their empirically established association with stressful life events and insomnia disorder [26, 37–39]. The disjunctive cause criterion by VanderWeele (2011) was utilized to determine which covariates would be included in the final model [40]. Chi-squared and ANOVA analyses were used to assess patterns of missingness in the data.

Primary Analysis.

The primary analysis examined the relationship between insomnia disorder at study baseline and number of stressful life events reported at three-month follow-up. Due to the count nature of the outcome variable, as well as its skewed distribution, Poisson regression was employed. This analysis controlled for relevant covariates and number of stressful life events at study baseline.

In the cross-lagged analysis, logistic regression was used to examine the effect of stressful life events at study baseline on insomnia diagnosis at three-month follow-up, controlling for baseline insomnia and covariates.

Secondary Analyses.

The secondary analyses also employed Poisson and logistic regression techniques to examine the cross-lagged associations between insomnia disorder at study baseline and number of dependent and interpersonal stressful life events, respectively, at three-month follow-up.

Results

Demographics

Demographic characteristics of the study sample are given in Table 1. The sample had a mean age of 42 (SD = 11), with no significant difference in mean age between the insomnia disorder and no insomnia disorder groups. A total of 55 participants from the parent study were excluded from the present analysis due to missing follow-up data. Participants included in the current analyses (n=136) were significantly older than participants who did not complete follow-up (n=55). Age was thus included as a covariate in all regression models. Findings from the larger parent study indicated no effects on attrition and reemployment at this 3-month timepoint [41], indicating that reemployment was an unlikely a reason for attrition.

Preliminary Analyses

Preliminary chi-square and ANOVA analyses indicated that, of all demographic factors, only educational attainment was significantly associated with insomnia disorder at study baseline. Individuals who had not attained at least a 2-year degree were at higher risk for insomnia disorder than individuals who held an associate degree or higher, χ² (1, N = 136) = 3.99, p = .05. Also, at baseline, participants reported 2.70 (SD = 1.39) stressful life events on average. There was no difference in the number of stressful life events between individuals with and without insomnia disorder, F (1, 135) = .27, p = .60. The full sample had a mean of 0.33 interpersonal events (SD = .57) and 1.21 dependent events (SD = 1.19) at study baseline. The frequency of interpersonal and dependent events did not differ significantly by insomnia disorder diagnosis. Of demographic variables, only sex was significantly associated with baseline number of stressful life events, with female participants reporting more stressful life events than male participants, F (1, 135) = 9.06, p = .003.

At the three-month follow-up, 27% of participants (n = 36) were classified as having insomnia disorder. Approximately 50% of the sample (n = 55 remitted, n = 14 new case at follow-up) experienced a change in the presence of an insomnia disorder over the three-months period; only 16% of the sample (n = 22) had insomnia disorder at both time-points. A total of 60 participants (n = 44%) were reemployed and working at least 30 hours per week at follow-up. Reemployment was not associated with insomnia, χ² (1, N = 136) = 1.27, p = .26, nor the number of stressful life events, F (1, 135) = 3.68, p = .06, at follow-up.

Primary Analysis

Figure 1 and Table 2 show the results of Poisson regression analysis of the relationship between insomnia disorder at study baseline and number of stressful life events at three-month follow-up. Controlling for age, gender, education, and number of stressful life events at baseline, insomnia disorder was associated with significantly more stressful life events at three-month follow-up (RR = 1.36, p = .01). At the trend level, insomnia disorder predicted increased dependent stressful life events at follow-up (RR = 1.34, p = .07), but not increased interpersonal stressful life events (RR = 1.27, p = .48).

Figure 1. Results from a cross-lagged panel analysis examining the relationship between insomnia disorder and total number of stressful life events (N=136).

Note. Estimates presented in β (Standard Error). Bolded arrow indicates primary relationship of interest. All models controlled for age to account for significantly lower age in excluded versus included ADAPT participants. Models controlled for gender and educational attainment where relevant as determined through application of the disjunctive cause criteria.

*significant at p ≤ .05

^aPoisson regression analysis controlling for number of events at baseline, age, gender, and educational attainment (as a proxy for socioeconomic status); ^bLogistic regression analysis controlling for number of events at baseline, age, and gender.

Table 2.

Results of Poisson regression analysis examining associations between baseline insomnia disorder and number of stressful life events at three-month follow-up (N=136).

Outcome at 3-month follow-up	RR [95% CI]
All eventsa	1.36 [1.08, 1.73]*
Interpersonal eventsb	1.27 [0.68, 2.34]
Dependent eventsc	1.34 [.98, 1.82]t

^*Significant at p ≤ .05

^tTrend-level association, p ≤ .10.

^aControlling for number of events at baseline, age, gender, and educational attainment (as a proxy for socioeconomic status);

^bControlling for number of interpersonal events at baseline, age, and educational attainment;

^cControlling for number of dependent events at baseline, age, and educational attainment

Cross-lagged analysis

Controlling for covariates and insomnia disorder at baseline, number of total stressful life events at baseline was not associated with insomnia disorder at follow-up (OR = .98, p = .87). Neither number of interpersonal nor dependent events significantly predicted increased risk for insomnia disorder at follow-up.

Discussion

The present cross-lagged panel analysis in participants who had recently experienced involuntary job loss demonstrated that insomnia disorder at study baseline predicted more stressful life events at three-month follow-up. Findings suggest that, like depression [16], insomnia disorder may also increase the risk of experiencing stressful life events. This is an important expansion of the stress generation hypothesis. Continuity of stress exposure can lead to an accumulation of stress burden that strains the physical, social, and psychological resources available to individuals and families [2]. Insomnia disorder may be an important avenue for intervention in populations at high risk for stressful life events, such as in individuals who have recently experienced involuntary job loss and those living in underserved communities.

An increased number of stressful life events at baseline did not predict insomnia disorder at follow-up. In other words, while the path from baseline insomnia disorder to follow-up number of stressful life events was significant, the path from number of stressful life events at baseline to insomnia disorder at follow-up was not. There was also no association between the number of stressful life events and insomnia disorder in the cross-sectional analysis of the baseline data. This lack of association is surprising given such an association would be consistent with theory relevant to both insomnia disorder [6] and stress coping [2]. One plausible interpretation of results is that participants who have experienced recent, involuntary job loss may experience a ceiling effect in terms of stress exposure, wherein the relationship between stressful life events and insomnia disorder may not hold at higher stress burden thresholds. Involuntary job loss is not only a major stressful life event, but also an important risk factor for both insomnia disorder and additional stressful life events [42–44]. The potential for job loss to lead to additional stressful life events often stems from common sequelae to job loss such as reduced economic resources, family conflict, and drastic changes to routine [43]. Findings from this sample may thus not be generalizable to populations that have not recently experienced involuntary job loss or a similar, recent upheaval. The specificity of the sample may also explain the relatively low number of interpersonal events at baseline.

Insomnia disorder at baseline did not predict insomnia disorder at follow-up. About half of the sample experienced a change in insomnia disorder status over the course of the study. This rate is consistent with prior work demonstrating significant fluctuations [45] of insomnia over time. Taken together, these results suggest that insomnia disorder occurring after an identified stressor is often subject to spontaneous recovery but nonetheless places individuals at risk for the later generation of stressful life events. Future research may benefit from the long-term examination of stress - insomnia - stress cycle over time including the cumulative effect of these interactions on health.

Strengths and Limitations

Strengths of the present analysis included a prospective study design with multiple measurements over a brief (three-month) interval, longitudinal analyses, and an interview-based evaluation of stressful life events using the LEDS. Use of an interview-based assessment is considered best practice in stress research as the contextual information provided allows for independent judges to make objective and reliable ratings of the impact of the event [46] rather than relying on subjective ratings.

The results of this investigation are limited by a few factors. Multiple covariates known to be associated with stressful life events and insomnia disorder were included to reduce potential confounding, but unobserved confounders may still be present. In addition, future studies are necessary to test mechanistic questions related to insomnia disorder and stress generation. For instance, depression was not included as a covariate in this analysis largely due to the high comorbidity of depression and insomnia disorder [47] with significant symptom overlap between the two conditions [48] (i.e., insomnia is a symptom of depression; reduction in mood is a symptom of insomnia). While depression cannot be ruled-out as a mechanism of stress generation in the current analysis, insomnia disorder similarly cannot be ruled-out as a mechanism for stress generation in depression since previous stress generation studies have not included insomnia disorder as a control variable. More research on the combination of these two conditions in relation to stressful life events is necessary. From a public health perspective, the separation of the two effects has the benefit of enhanced generalizability. The current results generalize to all individuals who have lost their jobs and have insomnia disorder.

Both cross-sectional and prospective results indicated that stressful life events did not predict insomnia disorder. However, a replication of prospective findings is warranted, given the significant reduction in the prevalence of insomnia disorder at the three month timepoint that may limit the statistical conclusion validity. Further, a larger dataset would also enable the use of advanced statistical techniques, such as structural equation modeling to examine change over time. Although the sample is unique for the representation of the U.S. Southwest, including nearly a third of participants identifying as Hispanic, the sample did not have enough racial variation to assess the effects of race on the relationship of interest. Strictness of criteria related to rating of interpersonal events resulted in very low numbers of interpersonal events, limiting the secondary analyses. Finally, participants were all between 25 – 60 years of age and had all experienced involuntary job loss in the three months preceding the baseline interview. Thus, results cannot be generalized beyond this population.

Implications for Future Research

Future research should focus on clarifying the longitudinal association between insomnia disorder and stressful life events in larger samples and different populations. An especially important area of study is the psychological mechanisms that link insomnia to stressful life events. Several common symptoms of insomnia disorder may be implicated in increasing risk for stressful life events following insomnia disorder. For example, insomnia disorder is associated with fatigue and impairments in cognitive performance [10]. These symptoms of insomnia disorder could lead to poor complex decision making, reduced inhibitory control, and difficulty with working memory that could lead to difficulties with work performance [49]. In addition, mood disturbances such as irritability are common in individuals with insomnia disorder, increasing the risk of interpersonal conflict [50]. Mixed methods research might be particularly valuable in examining this relationship as qualitative techniques could provide insights related to how diurnal insomnia symptoms could directly affect interpersonal relationships and functioning across various realms.

In addition, future research may shed light on how the mechanisms of that relationship between insomnia disorder and stressful life events may differ from the mechanisms of relationships between stressful life events and other disorders, such as depression. Such findings may help explain the relationship between insomnia disorder and interpersonal and dependent events, respectively [17, 18].

Overall, these results provide strong proof-of-concept for the potential of early insomnia intervention after job loss to reduce stress accumulation. Insomnia disorder is highly amenable to intervention via multiple modalities [51], but the accessibility of evidence based treatments for insomnia is limited in individuals without jobs and health insurance [52]. Future research is necessary to examine whether insomnia intervention prevents the accumulation of stressful life events as well as chronic disease in this at-risk population. If so, then public health programs might benefit from the adoption of sleep health campaigns in partnership with unemployment insurance agencies.

Conclusion

Insomnia disorder predicted increased stressful life events in adults who had recently experienced involuntary job loss. As previous theory and research indicate that a number of adverse health outcomes are often preceded by stressful life events [1, 2], the current analysis suggests that insomnia disorder may be a key, modifiable factor in the prevention of stress accumulation. More research is needed to verify these findings and to examine mechanisms in the causal association between insomnia disorder and stressful life events.

Highlights.

• Life Events and Difficulties Schedule used to quantify the number of stressful life events

• Stressful life events and insomnia disorder were measured at baseline (with 90 days of job loss) and again 3 months later

• Insomnia at baseline was associated with more stressful life events three months later

• More stressful life events at baseline was not associated with insomnia three months later

• Stress-generation hypothesis of depression may extend to insomnia