Insomnia, Comorbidity, and Risk of Injury Among Insured Americans: Results from the America Insomnia Survey

Abstract

Study Objectives:

To estimate associations of broadly defined insomnia (i.e., meeting inclusion criteria for International Classification of Diseases, Tenth Revision (ICD-10), Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), or Research Diagnostic Criteria/International Classification of Sleep Disorders, Second Edition (RDC/ICSD-2) diagnosis) with workplace/nonworkplace injuries controlling for comorbid conditions among workers in the America Insomnia Survey (AIS).

Design/Setting:

Cross-sectional telephone survey.

Participants:

National sample of 4,991 employed health plan subscribers (age 18 yr and older).

Interventions:

None.

Measurements and Results:

Broadly defined insomnia with duration of at least 12 mo was assessed with the Brief Insomnia Questionnaire (BIQ). Injuries in the 12 mo before interview were assessed with a standard self-report measure of injuries causing role impairment or requiring medical attention. Eighteen comorbid condition clusters were assessed with medical/pharmacy claims records and self-reports. Insomnia had significant gross associations (odds ratios, ORs) with both workplace and nonworkplace injuries (OR 2.0 and 1.5, respectively) in logistic regression analyses before controlling for comorbid conditions. The significant population attributable risk proportions (PARPs) of total injuries with insomnia was 4.6% after controlling for comorbid conditions. Only 2 other conditions had PARPs exceeding those of insomnia. The associations of insomnia with injuries did not vary significantly with worker age, sex, or education, but did vary significantly with comorbid conditions. Specifically, insomnia was significantly associated with workplace and nonworkplace injuries (OR 1.8 and 1.5, respectively) among workers having no comorbid conditions, with workplace but not nonworkplace injuries (OR 1.8 and 1.2, respectively) among workers having 1 comorbid condition, and with neither workplace nor nonworkplace injuries (OR 0.9 and 1.0, respectively) among workers having 2 or more comorbid conditions.

Conclusions:

The associations of insomnia with injuries vary with comorbid conditions in ways that could have important implications for targeting workplace interventions.

Citation:

Kessler RC; Berglund PA; Coulouvrat C; Fitzgerald T; Hajak G; Roth T; Shahly V; Shillington AC; Stephenson JJ; Walsh JK. Insomnia, comorbidity, and risk of injury among insured Americans: results from the America Insomnia Survey. SLEEP 2012;35(6):825-834.

INTRODUCTION

Insomnia is highly prevalent in the United States, with an estimated 23.6% of non-institutionalized civilians meeting prevailing diagnostic criteria for clinically significant insomnia.¹ Recent experimental and epidemiologic literature associates insomnia with multiple daytime symptoms, including sleepiness and fatigue,^2,3 psychomotor performance deficits,^4,5 cognitive impairments,^6,7 and mood dysregulation;^8,9 leading to broader impairments in daytime role functioning^1,10–13 and increased risk of injuries.^14–16

The annual cost of sleep-related workplace injuries in the US civilian workforce is estimated to exceed $100 billion (2004 US dollars) using conservative assumptions about sleep disorder prevalence.¹⁷ Self-reported sleep problems,^14,16,18,19 short time asleep,²⁰ fatigue,²¹ and daytime sleepiness²² have also been linked to workplace injuries in a number of community epidemiologic surveys, with odds ratios (ORs) typically in the range of 1.5-2.5. We are aware of only 1 large-scale survey, though, that examined the association between insomnia defined in terms of full Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) inclusion criteria and accidents.²³ That study documented an OR of 1.7 between insomnia and serious accidents, but excluded respondents with any comorbid chronic physical or mental disorders or extended work absences, precluding insights into real-world scenarios where insomnia is typically highly comorbid.²⁴

The focus of previous research on injuries in the workplace is understandable in light of the concerns that exist about workplace safety. But this means that little is known about the associations of sleep problems with nonworkplace injuries. In addition, very little is known about the ways in which insomnia might interact with comorbid conditions in leading to injuries. The current report addresses these limitations by presenting data on the associations of rigorously defined insomnia, with both workplace and nonworkplace injuries adjusting for comorbid conditions among workers who participated in the America Insomnia Survey (AIS).¹ The AIS is a national survey of subscribers to a large (more than 34 million members) US health plan that used a validated self-report scale²⁵ to diagnose insomnia, a self-report questionnaire comparable to questions used in previous large-scale epidemiologic surveys to assess injuries,²⁶ and information from both medical-pharmacy claims records and self-reports to assess and adjust for both treated and untreated comorbid conditions.

METHODS

The Sample

The AIS was carried out between October 2008 and July 2009 in a stratified probability sample of 10,094 adults (ages 18 yr and older). The sample was restricted to fully insured members of the national US commercial health plan that served as the sample frame. Members had to be enrolled for at least 12 mo before interview to allow medical and pharmacy claims data to be used in substantive analyses. Sample eligibility was also limited to members who provided the plan with a telephone number, could speak English, and had no impairment that limited their ability to be interviewed by telephone. The sample was selected with stratification to match the US national census population distribution on the cross-classification of age (18-34 yr, 35-49 yr, 50-64 yr, 65-74 yr, and older than 75 yr), sex, urbanicity (census Standard Metropolitan Statistical Areas [SMSA], non-SMSA urbanized areas, and rural areas), and census region (Northeast, South, Midwest, and West). Information about diagnoses or treatment of sleep disorders was ignored in the sample selection to make the sample representative of all plan subscribers.

A letter was sent to target respondents in advance, explaining that the survey was designed “to better understand how health and health problems affect the daily lives of people,” that respondents were selected randomly, that participation was voluntary, that responses were confidential, that participation would not affect health care benefits, and that a $20 incentive was offered for participation. A toll-free number was included for respondents who wanted more information or to opt out. Once respondents were contacted by telephone, verbal informed consent was obtained before beginning interviews. The Human Subjects Committee of the New England Institutional Review Board (http://www.neirb.com) approved these recruitment, consent, and field procedures. The cooperation rate (the rate of survey completion among target respondents with known working telephone numbers, including respondents who were never reached) was 65.0%. The 10,094 interviews were weighted for residual discrepancies between the joint distribution of the sociodemographic and geographic selection criteria in the sample compared with the census population. A total of 7,428 AIS respondents were either employed or self-employed.

In addition to assessing insomnia, the AIS included many questions about the correlates of insomnia. To reduce respondent burden, some questions were administered only to probability subsamples. One such set concerned physical and mental conditions found in previous research to be comorbid with insomnia. Self-report questions about these conditions were administered to all AIS respondents reporting any sleep problems plus a random sample of 50% of other respondents. The random half-sample of other respondents retained in the full survey was compared with the other half-sample and a weighting adjustment of the former was used to make it representative of both half-samples on the marginal distributions of a wide range of poststratification variables.²⁷ This weighting adjustment was normalized to have a mean of 2.0 (multiplied by the weight described in the previous paragraph) in the comorbidity sample to adjust for the fact that they represent only half of those without sleep problems in the full sample. A total of 4,991 AIS respondents in this comorbidity subsample were either employed or self-employed. This is the subsample of respondents used in the analyses reported here.

Measures

Insomnia

Insomnia in the 30 days before interview was assessed with the Brief Insomnia Questionnaire (BIQ), a 32-question, fully structured interviewer-administered questionnaire developed for the AIS to operationalize insomnia according to the inclusion criteria of the DSM-IV-Text Revision (TR),²⁸ Research Diagnostic Criteria/International Classification of Sleep Disorders, Second Edition (RDC/ICSD-2),²⁹ and International Classification of Diseases, Tenth Revision (ICD-10)³⁰ systems, hereafter referred to as broadly defined insomnia or insomnia. The cases considered here meet full symptom inclusion criteria in at least 1 of these systems, all of which require respondents to report 1 or more nighttime symptom(s) (difficulty initiating sleep, difficulty maintaining sleep, early morning awakening, or nonrestorative sleep) in addition to daytime distress/impairment and other criteria that vary across systems. Daytime distress/impairment was assessed with questions about a broad range of daytime activities and symptoms (e.g., reduced motivation; reduced performance at work, school, or social activities; occupational errors; irritability; mood lability; tension headaches; digestive problems; interference with home management, close relationships, or work) in an effort to address the concern that the unidimensional assessments of daytime impairment often used in previous epidemiologic studies failed to detect patients with insomnia characterized by atypical types of distress or impairment.⁹

The operational definitions of core insomnia diagnostic criteria were harmonized across systems in the BIQ to require that nighttime symptoms occur at least 3 times per week, that the difficulties initiating and maintaining sleep (i.e., difficulties getting to sleep, difficulties awakening and not being able to get back to sleep at night, and waking too early in the morning) lasted for at least 30 min on the nights they occurred, and that the repeated problems initiating sleep, maintaining sleep, and nonrestorative sleep persisted for at least 1 mo. The full text of the BIQ along with diagnostic algorithms is available at http://www.hcp.med.harvard.edu/wmh/affiliated_studies.php. Respondents who met criteria for broadly defined insomnia were asked to report how long they had experienced insomnia. Because the focus of the current report is on injuries that occurred at any time in the 12 mo before interview, we consider here the subset of respondents who reported that their insomnia had persisted for at least 1 yr at the time of interview.

Because of difficulties involved in distinguishing primary insomnia from insomnia comorbid with physical/mental disorders or substance/medication use, no attempt was made in the BIQ to operationalize the diagnostic hierarchy or organic exclusion rules in DSM-IV-TR Criteria C-E or to distinguish DSM-IV-TR Primary Insomnia, RDC/ICSD Insomnia Disorder, or ICD-10 Non-organic Insomnia from other insomnia phenotypes. This decision is consistent with the most recent recommendations of the task force revising the DSM criteria.²⁴ However, medical and pharmacy claims data for the 12 mo before interview were obtained from the health plan for all AIS respondents to study the effects of diagnosed and treated comorbid conditions on correlates of insomnia. The AIS interview also obtained self-report assessments of chronic conditions known to be associated with insomnia for the same purpose (see next paragraphs). These were introduced as control variables in regression analyses to adjust for the effects of comorbid conditions. This analysis approach is consistent with the recommendations of both the 2005 National Institutes of Health (NIH) State-of-the-Science Conference³¹ and the 2006 Recommendations for Research Assessment of Insomnia.³²

A clinical reappraisal study was carried out with a subsample of AIS respondents that oversampled those with positive screening results in the BIQ. Blinded clinical interviewers who were highly experienced sleep medicine experts carried out semistructured clinical interviews to make diagnoses of insomnia according to the definitions and inclusion criteria of the 3 systems considered here. Psychometric analyses documented good individual-level concordance of diagnoses based on the BIQ with these independent clinical diagnoses.²⁵ Sensitivity of broadly defined BIQ diagnoses of insomnia compared with clinical diagnoses was 72.6%, specificity was 98.9%, and the area under the receiver operating characteristic curve (a measure of classification accuracy insensitive to disorder prevalence) was 0.86. Cohen kappa was 0.77, a value at the upper end of the range conventionally judged to represent substantial agreement with clinical diagnoses.³³ A more detailed description of the BIQ, its validation, and the qualifications of the clinical interviewers is presented elsewhere.²⁵

Other chronic conditions

As noted previously, medical and pharmacy claims data for the 12 mo before interview along with self-report data on untreated conditions were obtained for disorders and syndromes documented in the literature to be associated with elevated rates of insomnia.^34–36 A total of 18 such condition clusters were considered. These included cardiovascular disorders, diabetes, musculoskeletal disorders (arthritis, chronic back/neck pain), respiratory disorders (chronic obstructive pulmonary disease [COPD], seasonal allergies, and other respiratory disorders including chronic bronchitis and emphysema), digestive disorders (gastroesophageal reflux disease [GERD], irritable bowel syndrome), pain conditions (migraine, other frequent or severe headaches, neuropathic pain, and any other chronic pain), mental disorders (major depression, other mental disorder), other sleep disorders (sleep apnea, restless leg syndrome [RLS]), climacteric symptoms common to perimenopausal women (ICD-9 Clinical Modification (CM) Diagnosis Code 627.2), and urinary/bladder problems,.

Diagnoses were obtained for all the previously mentioned conditions from ICD-9 codes in medical claims and inferred from pharmacy claims. Diagnoses based on self-reports were obtained in 2 ways. First, a chronic conditions checklist was used based on the list in the US National Health Interview Survey³⁷ (http://www.hcp.med.harvard.edu/ncs/replication.php). Such checklists have been widely used in epidemiologic studies and yield more complete and accurate reports than estimates derived from responses to open-ended questions.³⁸ Methodologic studies have documented good concordance between such checklists and medical records.^39–41 Second, a series of validated disorder-specific self-report scales was used to detect a number of undiagnosed symptom-based physical conditions that included arthritis,⁴² COPD,⁴³ GERD,⁴⁴ sleep apnea,⁴⁵ and RLS.⁴⁶ Self-report scales were also used to detect 3 undiagnosed mental disorders: major depression,⁴⁷ anxiety,⁴⁸ and alcohol abuse.⁴⁹

Injuries

Injuries were assessed with questions based on those used in other large-scale community epidemiologic surveys.²⁶ The stem question asked respondents if in the past 12 mo they had an “accident, injury, or poisoning that either put you out of commission for the rest of the day or that required you to get first aid or medical attention.” The minimum severity threshold (i.e. either causing impairment for at least 1 day or requiring medical attention) in this question is consistent with the thresholds used in a series of injury supplements to the US National Health Interview Survey (http://www.cdc.gov/nchs/nhis/injury_poisoning/ip_history.htm). Additional study questions asked about how many times during the past 12 months the respondent had an injury and the nature, location, and cause of this injury (or, for respondents who had multiple injuries, of the most serious injury). Review of open-ended responses about nature shows that the incidents reported were true injuries (e.g., falls, cuts, vehicular accidents) rather than flares of chronic conditions (e.g., migraines). For purposes of the current analyses, locations reported were divided into work-related (including travel to and from work) versus all others.

Although we did not have access to medical claims data on injuries (ICD-9 800-999 Codes) to validate the survey reports, previous studies using roughly similar measures have reported sensitivities in the range 0.46-0.65,^50–53 suggesting that self-reports underestimate true injury prevalence. There is also a suggestion in the literature that failure to recall injuries that are not very serious begins to emerge for recall periods longer than 3 mo, leading to the recommendation that self-report injury questions should have a 3-mo recall period.^54–58 However, other research has shown that the predictors of injuries are quite comparable for shorter and longer recall periods up to 1 yr, arguing that statistical power to study the predictors of injuries can be increased without introducing more than a small amount of bias by increasing the recall period to 12 mo.⁵⁹ Based on that observation, a 12-mo recall period was used in assessing injuries in the AIS, recognizing that this would yield a conservative estimate of injury prevalence but a more precise estimate of the association between insomnia and injury than with a shorter recall period.

Employment and other sociodemographic variables

All AIS respondents were asked if they were employed, self-employed, unemployed and looking for work, a student, homemaker, retired, or something else. All respondents who reported they were either employed or self-employed (henceforth referred to as employed) were included in the current analysis. Information obtained on respondent age (18-29 yr, 30-44 yr, 45-59 yr, 60 yr and older), sex, and educational attainment (high school graduate or less, some college, college graduate or more) was included as control variables in the regression analyses.

Analysis Methods

Cross-tabulations were used to examine comorbidities of insomnia with other chronic conditions. Logistic regression analysis was used to estimate associations of insomnia with 12-mo prevalence of workplace, nonworkplace, and total injuries both with and without variables in the equations to control for chronic conditions. Logistic regression coefficients and their standard errors were exponentiated to create ORs and 95% confidence intervals. Statistical significance was consistently evaluated in the logistic regression equations using 0.05-level 2-sided tests. As the AIS data are weighted, the design-based Taylor series method⁶⁰ implemented in the SAS SURVEY procedures⁶¹ was used to estimate standard errors and evaluate statistical significance.

Population attributable risk proportions (PARP) were calculated for the associations of insomnia with each type of injury, using simulation methods described in previous AIS reports.^1,13 PARP is defined as the incremental (i.e., controlling statistically in the regression equations for all comorbid conditions) proportion of observed injuries that would not have occurred under the logistic regression model if insomnia were not present (either prevented or effectively treated) and the insomnia coefficient in the model was due to a causal effect of insomnia. So, for example, a PARP of 7% would mean that 7% of all the observed injuries in the sample would be predicted not to have occurred if all cases of insomnia were absent.

RESULTS

Prevalence and Sociodemographic Correlates of Insomnia and Injuries

As mentioned in the introduction, the prevalence (standard error) of broadly defined insomnia with a duration of at least 30 days is 23.6% (0.5). This percentage decreases to 23.0 (0.6) when restricting the sample to employed respondents in the AIS comorbidity subsample, whereas prevalence drops to 20.0% (0.5) when we impose a duration requirement of at least 12 mo among employed respondents. We focus on the latter prevalence estimate in the current report. This prevalence estimate is significantly higher (OR) among workers ages 18-29 yr (1.6), 30-44 yr (1.7), and 45-64 yr (1.8) than those 65 yr and older (1.0), higher among women than men (1.4), and not significantly related to education (Table 1).

Table 1.

Associations (odds ratios) of sociodemographic variables with BIQ/broadly defined insomnia of at least 12 mo duration^a and 12-mo injuries among employed AIS respondents in the comorbidity subsample (n = 4,991)^b

The estimated prevalence (standard error) of 12-mo injuries among employed respondents in the AIS comorbidity subsample is 12.3% (0.5), including 3.2% (0.3) workplace injuries and 9.1% (0.4) nonworkplace injuries. Both workplace and nonworkplace injuries are significantly and inversely related to age, with ORs for younger versus oldest workers in the range 1.8-5.5 (Table 1). Workplace injuries have a significantly elevated OR among men compared with women (2.1) and among respondents with less than a college education than those who completed college (1.3-2.0). Nonworkplace injuries, in comparison, are not significantly related to sex or education.

Comorbidity of Insomnia with Other Disorders

Insomnia is positively associated (i.e., an OR greater than 1.0) with all the 18 comorbid conditions considered here and significantly so with 17 of these conditions (Table 2). ORs range from a high of 5.5 with major depression to a low of 1.2 with cardiovascular disorder. The inter-quartile range (i.e., 25^th-75^th percentiles) of the ORs is 1.4-2.5.

Table 2.

Comorbidity (odds ratio) of BIQ/broadly defined insomnia of at least 12 mo duration^a with other physical and mental disorders among employed AIS respondents in the comorbidity subsample (n = 4,991)^b

Associations of Insomnia with Injuries

Respondents with insomnia were significantly more likely than others to report both workplace and nonworkplace injuries, with significantly elevated ORs of 1.6 for total injuries, 1.9 for workplace injuries, and 1.5 for nonworkplace injuries in the model that had no control variables (Table 3). The ORs are virtually unchanged in the model that includes control variables for sociodemographic but not comorbid conditions. The ORs decrease in magnitude in the model that adds control variables for comorbid conditions, although the OR in this model remains statistically significant for all injuries (1.2), but not in separate equations for either workplace) or non-workplace) injuries (1.4 and 1.2, respectively). Estimates of PARP suggest that 5.6% of all workplace injuries are associated with insomnia, with a decrease to 2.3% when control variables are introduced into the equation for both sociodemographic and comorbid conditions. The comparable percentages are 5.4% (without any control variables) and 2.8% (with control variables for sociodemographic and comorbid conditions) for nonworkplace injuries and 7.7% (without these control variables) and 4.6% (with these control variables) for total injuries.

Table 3.

Individual-level (odds ratios) and aggregate (population attributable risk proportions) associations of BIQ/broadly defined insomnia of at least 12 mo duration^a with 12-mo injuries with and without control individuals for sociodemographics and comorbidity among employed AIS respondents in the comorbidity subsample (n = 4,991)^b

Six of the 18 comorbid conditions also are significantly associated with total injuries in the multivariate model: 4 pain conditions (arthritis, frequent or severe headaches, neuropathic pain, any other chronic pain), mental disorders other than depression, and COPD). ORs are in the range 1.3-2.1. Two of these 6 conditions have higher PARPs than insomnia: other chronic pain conditions (20.2%) and arthritis (5.2%). Five conditions significantly predict workplace injuries: frequent or severe headaches, other chronic pain conditions, mental disorders other than depression, arthritis, and diabetes. ORs are in the range 1.6-2.2. Three of these 5 have higher PARPs than insomnia: headaches (3.2%), pain conditions (8.5%), and arthritis (4.3%). Four conditions significantly predict nonworkplace injuries: frequent or severe headaches, neuropathic pain, chronic back/neck pain, and chronic bronchitis/emphysema. ORs are in the range 1.2-2.1. Two of these 4 have higher PARPs than insomnia: neuropathic pain (14.7%) and chronic back/neck pain (4.3%). (More detailed results are available on request.)

Subgroup Variation in the Associations of Insomnia with Injuries

Analyses of statistical interactions found no evidence that the associations of insomnia with injuries vary either with respondent age (χ²₃ = 1.5-3.5, P = 0.32-0.69), sex (χ²₁ = 0.1-1.0, P = 0.32-0.82), or education (χ²₂ = 0.3-0.8, P = 0.67-0.86; noting that respondents with less than a high school education had to be combined with high school graduates because of the small number of respondents in the former category). (Detailed results are available on request.) The association of insomnia with injuries varies significantly, though, with number of significant comorbid conditions (χ²₁ = 5.0, P = 0.026 for total injuries; χ²₁ = 5.2, P = 0.022 for workplace injuries; χ²₁ = 2.5, P = 0.11 for nonworkplace injuries).

Disaggregation of the association between insomnia and injuries in subsamples defined by number of significant comorbid conditions found that the association is significant among respondents with no comorbid conditions (1.5-1.8) and only 1 comorbid condition (1.2-1.8), but insignificant among respondents with 2 or more comorbid conditions (0.9-1.0) (Table 4). Insomnia is consistently a stronger predictor of workplace than nonworkplace injuries (1.8 versus 1.2-1.5) among respondents with no comorbid conditions or 1 such condition.

Table 4.

Individual-level (odds ratios) associations of BIQ/broadly defined insomnia of at least 12-mo duration^a with 12-mo injuries stratified by number of comorbid conditions among employed AIS respondents in the comorbidity subsample^b

DISCUSSION

The results reported here are broadly consistent with previous epidemiologic studies in documenting significant associations of sleep difficulties, fatigue, and daytime sleepiness with workplace injuries.^{14,16,18–22} The significant ORs between insomnia and workplace injuries in the AIS – 2.0 without adjusting for comorbidity and 1.4-1.8 with adjustment – are in the range found with sleep problems in these previous studies. The PARPs of insomnia associated with workplace injuries in the AIS (6.1% without adjustment, 2.3 with adjustment) are also broadly consistent with those in the small number of previous studies that calculated PARPs.^14,17,62

We also showed that insomnia has a significant association with nonworkplace injuries, although weaker than with workplace injuries. One possible reason for this difference might be the tighter constraints imposed on situational injury risk factors in workplaces than in other settings. Future research could investigate this possibility by disaggregating associations of insomnia with injuries using a standardized cause-of-injury classification system (e.g., http://www.who.int/classifications/icd/adaptations/iceci/en/). This was not possible in the AIS because of the relatively small sample size and low statistical power to study injury types. Sample size restrictions also account for the fact that we made no attempt to replicate previous studies of workplace injuries in specific occupations.^63–68 This kind of disaggregation using measures of sleep problems rather than insomnia would be possible in several epidemiologic surveys of injuries that are 10 or more times as large as the AIS.^{14,16,20,36,69} Another disaggregation of interest would be to examine insomnia by treatment status, as it is of interest to determine if treated insomnia is as strongly related to injury as untreated insomnia, although caution would be needed in interpreting results due to nonrandom assignment to treatment. This kind of disaggregation was not possible in the AIS, as we did not collect data on treatment status in the 12 mo before interview. Future studies might also profitably examine disaggregation by type of treatment, as some kinds of treatment might be risk factors for injuries.⁷⁰

It is becoming increasingly clear that one-fourth or more of injuries are related to pre-existing chronic conditions.^71–73 However, no previous study of these effects included insomnia among the chronic conditions examined. The AIS finding that only a small number of the comorbid conditions considered here have higher PARPs than insomnia is consequently important. This is especially true in light of the fact that some of these other conditions might have been consequences of injuries (such as back/neck pain, headaches, and other pain conditions) rather than causes of injuries, leading to a possible overestimation of their PARPs. This kind of confounding was minimized in estimating the PARP of insomnia due to the focus on insomnia that had persisted for more than 12 mo in predicting injuries that occurred within the past 12 mo. No comparable dating information was collected to allow duration of comorbid conditions to be set in the same way. Just as the PARPs of comorbid conditions might have been overestimated because of this lack of dating information, the PARP of insomnia was underestimated to the extent that injuries were caused by insomnia with duration of less than 12 mo.

We also investigated interactions of insomnia with sociodemographic and comorbid conditions. No evidence was found of interactions with sociodemographic conditions. Previous studies have yielded inconsistent results in this regard.^14,16,62,64 Interactions with comorbid conditions, in comparison, were significant, with associations of insomnia with injuries confined to respondents with either no or 1 significant comorbid conditions. Although no previous studies investigates similar interactions, a large prospective epidemiologic study of municipal workers in Finland found that the significant association between severe self-reported sleep disturbance and subsequent injury-related work disability was strongest among respondents defined as being healthy at baseline (i.e., no baseline evidence of any chronic condition).¹⁶ It would be useful for this same sort of specification to be examined in other large, general-population epidemiologic studies that have relevant data.^14,20,36,69 If replicated, this finding would strengthen the case that the associations of pure insomnia and sleep problems with injuries are due to insomnia and sleep problems per se rather than to comorbid conditions.

But what of the AIS finding that the incremental association of insomnia with injuries is insignificant among respondents with 2 or more comorbid conditions? A methodologic interpretation of this finding would be that the cross-sectional AIS study design made it impossible to exclude comorbid conditions that were consequences rather than causes of injuries (most plausibly the pain conditions, which were found to have the strongest and most consistent associations with injuries), leading to statistical overcontrolling for comorbid consequences of injuries and underestimating the effects of insomnia. In considering this interpretation, it is noteworthy that subadditive interactions of a similar sort have been reported in previous studies of a wide range of comorbidities predicting disability, role functioning,⁵ health utilities, and onset of secondary disorders.^74–77 In each of these cases, the associations of focal disorders with outcomes of interest became smaller as comorbidity increased. A plausible substantive interpretation of this general pattern is that the proportion of people with any likelihood of developing the adverse outcome (in our case, becoming injured) who do, in fact, develop the outcome increases to the maximum as comorbidity increases. This means that the remaining at-risk people represent an increasingly resilient segment of the population. The incremental effect of any additional condition, such as insomnia, is likely to be small in this remainder of the population because of their resilience. Although to our knowledge not previously discussed in the comorbidity literature, such unmeasured differences in resilience are well known in the outcomes research literature, where they are referred to as the source of survivorship bias in estimates of treatment effects in the presence of nonrandom selection into treatment.⁷⁸ A related possibility is that the prevalence of insomnia becomes so high among people with multiple comorbid conditions that it becomes impossible to distinguish the effects of insomnia from comorbidity in this subsample without more fine-grained specification. This possibility could not be examined in the AIS because of sample size restrictions, but could be considered in reanalysis of some of the larger epidemiologic surveys of injury.

The AIS results are limited by 2 sampling issues: that the low survey cooperation rate (65.0%) could have distorted estimates of prevalence and correlates of insomnia; and that AIS results do not generalize to the approximately 15% of the US population that lacks health insurance or to segments of the population with insurance not provided by commercial health plans. The AIS measures also have limitations. Insomnia was assessed with a fully structured scale rather than with clinical interviews, which could introduce imprecision into the measurement of insomnia, although this concern is somewhat lessened by the good concordance between diagnoses based on the BIQ and independent diagnoses based on blinded clinical reappraisal interviews.²⁵ A related issue is that diagnostic hierarchy and organic exclusion rules were not applied to insomnia diagnoses. However, we adjusted for the effects of comorbid conditions to correct for any inflation in the estimated prevalence of insomnia. As noted previously, this use of comorbid conditions as statistical control variables in regression analysis rather than as diagnostic exclusions for insomnia was recommended by the 2005 NIH State-of-the-Science Conference,³¹ the 2006 Recommendations for Research Assessment of Insomnia,³² and the task force revising the DSM criteria.²⁴ Another measurement limitation is that injuries were assessed with self-report rather than objective Occupational Safety and Health Administration records or medical records. As noted in the section on measures, this is likely to have led to underestimation of injury prevalence. A final noteworthy measurement limitation is that we did not ask unemployed respondents if they became disabled in the past 12 mo due to an injury. This might have resulted in underestimation of the effects of injuries on workplace outcomes.

These limitations notwithstanding, the AIS findings are useful in several respects. First, they extend previous studies by showing that it is not only self-reported sleep problems but also rigorously diagnosed insomnia that is significantly associated with workplace injuries. Second, they document that insomnia is associated not only with workplace injuries but also non-workplace injuries. Third, they show that insomnia has stronger associations with injuries than most chronic conditions. Fourth, they show that aggregate associations of insomnia with injuries are comparable among workers who differ in key sociodemographic areas. Fifth, they show that associations of insomnia with injuries are strongest among workers with low comorbidity, suggesting that workplace interventions aimed at detecting and treating insomnia might focus on cases with low comorbidity. This suggestion is relevant given that insomnia workplace interventions are known to reduce workplace injuries.²² Targeting such programs, however, should await a more thorough understanding of insomnia-related associations with other human capital outcomes that might also be the focus of workplace interventions, such as sickness absence and work performance, because these could have different relationships with comorbidity.^11,79–81

DISCLOSURE STATEMENT

This study was funded by Merck & Co. Inc. Dr. Kessler has been a consultant for AstraZeneca, Analysis Group, Bristol-Myers Squibb, Cerner-Galt Associates, Eli Lilly & Company, GlaxoSmithKline Inc., HealthCore Inc., Health Dialog, Integrated Benefits Institute, John Snow Inc., Kaiser Permanente, Matria Inc., Mensante, Merck & Co, Inc., Ortho-McNeil Janssen Scientific Affairs, Pfizer Inc., Primary Care Network, Research Triangle Institute, Sanofi-Aventis Groupe, Shire US Inc., SRA International, Inc., Takeda Global Research & Development, Transcept Pharmaceuticals Inc., and Wyeth-Ayerst; has served on advisory boards for Appliance Computing II, Eli Lilly & Company, Mindsite, Ortho-McNeil Janssen Scientific Affairs, Plus One Health Management and Wyeth-Ayerst; and has had research support for his epidemiological studies from Analysis Group Inc., Bristol-Myers Squibb, Eli Lilly & Company, EPI-Q, GlaxoSmithKline, Johnson & Johnson Pharmaceuticals, Ortho-McNeil Janssen Scientific Affairs., Pfizer Inc., Sanofi-Aventis Groupe, and Shire US, Inc. Ms. Berglund has participated in research activities supported by Dr. Kessler's grants. Dr. Kessler's program at the Department of Health Care Policy at Harvard Medical School has received research funding from Pfizer, Sanofi-Aventis, Shire Development, Inc., and Janssen Pharmceutica, N.V. Dr. Coulouvrat is a full-time employee of Sanofi-Aventis, the industry that sponsored this study. Dr. Fitzgerald is a full-time employee of Merck & Co., Inc. Dr. Hajak has been on speakers board, consultant or a member of an advisory board or research funding by Actelion, Affectis, Astellas, Astra-Zeneca, Bayer Vital, BrainLab, Bristol-Meyers Squibb, Boehringer Ingelheim, Cephalon, Daimler Benz, EuMeCom, Essex, Gerson Lerman Group Council of Healthcare Advisors, GlaxoSmithKline, Janssen-Cilag, Lilly, Lundbeck, McKinsey, MedaCorp, Merck, Merz, Mundipharm, Network of Advisors, Neurim, Neurocrine, Novartis, Organon, Orphan, Pfizer, Pharmacia, Proctor & Gamble, Purdue, Sanofi-Aventis, Schering-Plough, Sepracor, Servier, Takeda, Transcept, Wyeth. Dr. Roth has served as a consultant for Abbott, Accadia, Acogolix, Acorda, Actelion, Addrenex, Alchemers, Alza, Ancel, Arena, AstraZeneca, Aventis, AVER, Bayer, BMS, BTG, Cephalon, Cypress, Dove, Eisai, Elan, Eli Lilly, Evotec, Forest, GlaxoSmithKline, Hypnion, Impax, Intec, Intra-Cellular, Jazz, Johnson and Johnson, King, Lundbeck, McNeil, MediciNova, Merck, Neurim, Neurocrine, Neurogen, Novadel, Novartis, Ocera, Orexo, Organon, Otsuka, Prestwick, Proctor and Gamble, Pfizer, Purdue, Resteva, Roche, Sanofi-Aventis, Schering-Plough, Sepracor, Servier, Shire, Somaxon, Somnus, Steady Sleep Rx, Syrex, Takeda, Transcept, Vanda, Ventus, Vivometrics, Wyeth, Yamanuchi, and Xenoport. He has served on speakers bureau for Cephalon, Sanofi-Aventis, and Sepracor. He has received research support from Aventis, Cephalon, GlaxoSmithKline, Merck, Neurocrine, Pfizer, Sanofi-Aventis, Schering-Plough, Sepracor, Somaxon, Somnus, Syrex, Takeda, TransOral, Ventus, Wyeth, and Xenoport. Dr. Shahly is an employee of the Department of Health Care Policy at Harvard Medical School. That program has received research funding from Pfizer, Sanofi-Aventis, Shire Development, Inc., and Janssen Pharmceutica, N.V. Dr. Shahly has no financial interest in any of these organizations. Dr. Shillington is employee and shareholder in the company Epi-Q, Inc. Epi-Q provides project management service and has received grant and consulting support from the following companies: Merck, Cephalon, Sanofi-Aventis, Pfizer, Biogen-IDEC, Onconova, AstraZeneca, GlaxoSmithKline, Novartis, Roche, Ortho McNeil, Takada, Transcept, Lundbeck Genentech, Bayer, Baxter, and Abbott. Dr. Shillington receives no direct compensation as a result of grants or contracts, other than her salary from Epi-Q. Dr. Stephenson is an employee of HealthCore, Inc., a research and consulting organization. All of her research activities are industry-sponsored. Dr. Walsh has been a consultant for Pfizer, Sanofi-Aventis, Respironics, Transcept, Neurogen, GlaxoSmithKline, Eli Lilly, Merck, Kingsdown, Vanda, Ventus, and Somnus. Research support has been provided to his institution by the following companies: Pfizer, Merck & Co., Somnus, Vanda, Neurogen, Sanofi-Aventis, Ventus, Respironics, and Jazz Pharmaceuticals.