Sleep Disorders, Health, and Safety in Police Officers FREE

Sleep disorders affect 50 to 70 million US residents.¹ Most are undiagnosed and remain untreated, resulting in adverse health, safety, and performance outcomes that have important economic ramifications.¹ Obstructive sleep apnea (OSA)² is associated with hypertension,^3- 4 cardiovascular disease,^3,5- 6 cognitive impairment,⁷ and increased risk of motor vehicle crashes.⁸ Insomnia is a risk factor for depression⁹ and hypertension¹⁰ and causes daytime functional impairments³ leading to absenteeism and productivity losses.¹¹ Shift work disorder, affecting approximately 10% of night and rotating shift workers,¹² is associated with social disturbances and higher rates of ulcers, unintentional injury, absenteeism, and depression.¹² Untreated sleep disorders and chronic sleep deficiency¹ increase the risk of unintential motor vehicle crashes and injuries.³ These problems are exacerbated in shift workers, who experience circadian rhythm disruption and chronic sleep deficiency due to their work schedules.¹³

Police officers frequently work extended shifts and long work weeks,¹⁴ which in other occupations are associated with increased risk of errors, unintended injuries, and motor vehicle crashes.¹⁵ According to data through the year 2003, more officers are killed by unintended adverse events than during the commission of felonies.¹⁴ It has been hypothesized that fatigue—likely due to reduced duration and quality of sleep¹⁶ and untreated sleep disorders¹⁷—may play an important role in police officer unintentional injuries and fatalities.¹⁴ To date, the effect of sleep disorders on police officer health, safety, and performance has not been systematically investigated.

We examined, among North American police officers, the risk of major sleep disorders, including those described above as well as restless legs syndrome and narcolepsy.² This prospective cohort study examined the association between screening positive for a sleep disorder and self-reported adverse health, safety, and performance outcomes.

A total of 4957 sworn police officers in North America (United States 97%, Canada 3%) volunteered to participate in either the online or on-site study component (Figure 1). The on-site portion of the study included intense investigations of a municipal police department serving 1 of the 10 largest US cities and a state police department serving 1 of the 10 most densely populated states. The participating police departments were not from the same state. The on-site cohort was included to achieve a high cooperation rate within those departments and to compare characteristics of responders and nonresponders (in the municipal police department). The online cohort was included to provide a comparison group of police officers from across North America. Both cohorts included monthly follow-up surveys.

To recruit for the online cohort, we corresponded with large law enforcement agencies across North America to solicit participation and placed advertisements in police magazines and newsletters and on police-focused Web sites. For the on-site cohorts, we solicited participation from several municipal and state police departments. Ultimately a municipal department and a state department were selected and agreed to participate in the study protocol. Of the 3329 potential attendees from the departments participating in the on-site study, 2004 participated in informational sessions. In addition, 4018 registered for the online study (Figure 1). Of these police officers, 4957 completed the baseline survey, which included a sleep disorders screening: 1264 (63.1%, cooperation rate¹⁸) in the on-site cohort and 3693 (91.9%, participation rate¹⁸) in the online cohort, which made them eligible for the cross-sectional analyses.¹⁸ In both cohorts, the 3545 officers who had completed at least 1 of the monthly follow-up surveys were eligible for the prospective analyses: 1264 in the on-site and 2281 in the online cohort. This represents a 63.1% cooperation rate and 56.8% participation rate, respectively; these were deemed the prospective study cohort.

The study protocol was approved by the Partners Human Research Committee and was conducted between July 2005 and December 2007. Participants provided written or electronic informed consent and were not informed about study hypotheses. For reporting purposes to the human research committee and to funding agencies, participants were required to self-report race and ethnicity using prespecified categories (Table 1). Participants selected for polysomnography studies received up to $440, and survey participants were eligible for a prize drawing valued at up to $1000.

In a cross-sectional baseline survey, we assessed demographics, physical and mental health status, and the risk of sleep disorders. For up to 2 years following the baseline screening, each month we sent all participants an e-mail with a link to a short online survey assessing work-related performance, work hours, and safety. Reminder e-mails were sent to those who did not immediately complete the survey. (The baseline screening questions and follow-up surveys are available from the authors.)

The sleep disorders screening questionnaire used validated, self-report screening tools for OSA (Berlin Questionnaire; sensitivity 0.86, specificity 0.77),¹⁹ moderate to severe insomnia (Athens Insomnia Scale: sensitivity 0.93, specificity 0.85),²⁰ restless legs syndrome (RLS Epidemiology, Symptoms, and Treatment questionnaire: sensitivity 0.82, specificity 0.90),²¹ and narcolepsy with cataplexy (Cataplexy Questionnaire: sensitivity 0.92, specificity 0.95²² and Epworth Sleepiness Scale [ESS]²³). For shift work disorder, we created a screening tool based on the International Classification of Sleep Disorders, Second Edition (ICSD-2) diagnostic criteria² (criteria are provided in the eMethods). The municipal police department allowed that only the OSA risk could be assessed for its 659 participants. Excessive sleepiness was assessed using the ESS (sensitivity 0.94, specificity 1.00)²³ to examine prevalence of this symptom and to compare across positive and negative sleep disorder groups.

In the baseline survey, participants reported current health status (poor to excellent); previous diagnoses of sleep and other medical disorders (eg, diabetes, cardiovascular disease, gastrointestinal tract disorder, depression, anxiety); likelihood of falling asleep while driving after work; and use of sleeping medications (never or nearly never to nearly every day), caffeine (0->8 servings/d), and alcohol (0- >14 servings/wk). The Maslach Burnout Inventory was used to assess 2 subscales of burnout: emotional exhaustion and depersonalization.²⁴

In the monthly surveys, participants were asked about work and sleep hours, likelihood of falling asleep in various situations and outcome measures of work performance, such as administrative errors, injuries, uncontrolled anger toward suspects or citizens, absenteeism, citizen complaints, and safety violations. The work-hours instrument was previously validated against daily work diaries (for monthly work hours, r  = 0.76; for extended shifts, r = 0.94; P < .001),²⁵ which in themselves were validated by direct observation (r = 0.98, P < .001).²⁶

To compare questionnaire OSA screening outcome with polysomnography, full, attended polysomnography studies were performed for 126 participants from the state police department. Of these, 63 had screened positive on the Berlin questionnaire and were the first from the overall study to agree to have sleep studies performed and agree to provide the investigators with access to relevant medical records; 61 had screened negative on the Berlin questionnaire and were selected randomly from the group of participants who screened negative; and 2 had invalid Berlin screening outcomes. Obstructive sleep apnea severity was classified by a sleep specialist, blind to questionnaire outcome, and was made on the basis of both the respiratory disturbance index and the minimal arterial oxyhemoglobin saturation level associated with respiratory disturbances (eMethods).

Multiple logistic regression models were used for cross-sectional analysis of the associations between health outcomes at baseline and sleep disorder screening outcomes (for positive screening for any sleep disorder, or positive OSA screening). For the prospective analysis, the generalized estimating equations (GEE) method²⁷ was used to assess performance and attentional failures in those who screened positive and those who did not screen positive for any of the sleep disorders because these outcomes were measured monthly over the course of this study and hence likely to be correlated. For performance and attentional failures, we used a logit link in the GEE model for dichotomized outcomes (1 for any number of the performance and attentional failures and 0 for no performance and attentional failures). Supplemental analysis was conducted on the continuous version of each outcome, assuming that the actual number of events follow a Poisson distribution and using a log link in the GEE model.

For the cross-sectional analysis, we adjusted for the potentially confounding effects of age, sex, body mass index (BMI), hypertension, cigarette smoking, alcohol consumption, primary police activity, second job, usual shift length, night shift work, and shift rotation. For the prospective analysis, we adjusted for age, sex, BMI, primary police activity, second job, night shift work, shift rotation, mean total work hours per week, and monthly sleep time. We tested these confounders individually and only included those that were significant (P < .10) in the initial multivariable models. We then applied a backward elimination method to remove those that were not significant in the final model. Only significant variables of P < .05 were left in the final model. Because there were some missing data for some of the confounders, we repeated the analysis using the missing-indicator method.²⁸ Unadjusted and adjusted odds ratios (ORs) with 95% confidence intervals are reported for the complete-case analysis, and adjusted ORs are reported when the missing-indicator method was applied.

For the cross-sectional analysis, we deemed the following primary outcomes: diabetes, depression, burnout (emotional exhaustion), and falling asleep while driving. For the prospective analysis, we deemed the following as the primary outcomes: falling asleep while driving, making a serious administrative error, making fatigue-related errors or safety violations, and having an occupational injury. P < .0125 was considered statistically significant for the 4 primary outcomes of interest in each analysis with Bonferroni adjustment. P < .05 was considered statistically significant in all secondary analyses. Statistical analyses were conducted using SAS for Windows (version 9.2, SAS Institute Inc).

Sample size and power calculations are based on our previous studies examining the effects of extended duration shifts on performance and attentional failures. A similar magnitude of difference was expected in the present study because both sleep disorders and extended durations of wakefulness have been reported to induce neurobehavioral performance impairments comparable with that of alcohol intoxication.³ For the sample size and power calculation, we considered 3 outcomes: falling asleep while driving, having an occupational injury, and making an error or committing a safety violation attributed to fatigue. Assuming that one-third of our sample would screen positive for a sleep disorder, with 15 000 person-months of data, we estimated 90% power to detect the following differences from those who had screened negative vs those who had screened positive for a sleep disorder: rate of falling asleep while driving, 0.066²⁵ vs 0.083 (OR, 1.28); rate of occupational injury, 0.010²⁹ vs 0.017 (OR, 1.71); and rate of error or safety violation due to fatigue, 0.038³⁰ vs 0.051 (OR, 1.36), with a 2-sided z test at P <.05/3 = .0167.

A total of 4957 police officers (mean [SD] age: 38.5 [8.3] years; years of police service: 12.7 [8.1]) completed the baseline survey. All participants reported being 18 years or older and sworn police officers. A total of 3930 officers (79.3%) were overweight or obese and 1663 (33.5%) were obese (Table 1). In 256 participants who attended a follow-up clinic visit, BMI was measured and found to be highly correlated with self-reported BMI (R = 0.912, P < .001).

As is the case nationally, white males made up the majority of officers. Based on the national average, women were slightly overrepresented in our sample (17.4% [95% CI, 16.3%-18.4%] vs 13.7% nationwide) whereas racial/ethnic minorities were somewhat underrepresented (7.9% [95% CI, 7.2-8.7] vs 12.7% for black officers nationwide; 5.9% [95% CI, 5.2-6.5] vs 9.1% for Hispanic officers nationwide³¹; Table 1). Our sample of onsite participants from the municipal police department (responders) was generally consistent in age, sex, and rank to characteristics of the entire police department, although the responders were slightly younger and had fewer high-ranking officers (eTable 1).

A total of 15 735 monthly surveys were completed during the 2-year follow-up, with a mean (SD) of 684.2 (254.1) completed monthly. Each participant completed a mean of 4.4 (5.2) surveys (median [interquartile range], 2 [1-6]). Officers who participated in the follow-up survey were similar to those who did not in terms of demographics and risk of sleep disorders (eTable 2).

A total of 2003 of 4957 participants (40.4%) screened positive for at least 1 sleep disorder (eTable 3 for demographic comparisons). One thousand six hundred sixty-six participants (33.6%) screened positive for OSA, the most common disorder, followed by 281 (6.5%) with moderate to severe insomnia; 269 (5.4%) with shift work disorder—those reporting excessive wake time sleepiness and insomnia associated with night work—representing 14.5% of those who work night shifts in the cohort; 70 (1.6%) with restless legs syndrome; and 16 (0.4%) with narcolepsy with cataplexy (Table 2). Applying the ICSD-2² criteria for shift work disorder, (ie, excessive wake time sleepiness or insomnia), 1004 (53.9%) of the police officers who work night shifts screened positive. Considering only those who showed both insomnia and night work–associated drowsiness (defined as a moderate to high chance of falling asleep while driving after working nights compared with never or a slight chance during days off, and moderate to high chance of falling asleep during night shift compared with never or slight chance during day shift), 47 (2.5%) on the night shift screened positive. To examine associations between positive sleep disorder screening and health and safety outcomes, the last definition of shift work disorder was used.

Higher proportions of participants in the online and in the on-site municipal police department cohorts screened positive for OSA than in the onsite state police department cohort: 36.0% (95% CI, 34.5%-37.6%) for the online cohort, 32.2% (95% CI, 28.6%-35.7%) for the municipal police cohort vs 20.3% (95% CI, 17.1%-23.5%) for the state police cohort. The state police cohort also reported a significantly lower mean (SD) BMI of 27.7 (3.6) than the municipal police cohort of 28.9 (4.6; P <.001) and the online cohort of 28.8 (4.8; P <.001). (BMI is calculated as weight in kilograms divided by height in meters squared.) Most participants who screened positive did not report a current diagnosis of the disorder (Table 2).

Positive screening for any sleep disorder was associated with increased risk of reported health- and safety-related outcomes: 203 (10.7%) of those who tested positive for a sleep disorder reported having depression vs 37 (4.4%) of those who did not screen positive (adjusted OR, 2.20; 95% CI [1.52-3.19]); 399 (34.1%) of the positive-screen group reported burnout (emotional exhaustion) vs 89 (17.9%) in the negative-screen group (adjusted OR, 2.85 [95% CI, 2.16-3.77]), and 388 (20.0%) in the positive-screen group reported falling asleep while driving vs 66 (7.7%) in the negative-screen group (adjusted OR, 3.79 [95% CI, 2.79-5.14]). Several of the secondary outcomes were also significantly associated with positive screening for sleep disorder, specifically gastrointestinal tract disorder, anxiety disorder, pharmacotherapy for insomnia, health status, and burnout (depersonalization) (Table 3 and eTable 4).

Because OSA was the most prevalent disorder reported, we examined associations between positive OSA screening results and risk of reported health- and safety-related outcomes. In addition to the risks observed with positive screening results for any sleep disorder, positive OSA screening was also associated with a diagnosis of diabetes, 84 (5.3%) of the positive-screen group vs 46 (1.5%) in the negative-screen group (adjusted OR, 1.61 [95% CI, 1.05-2.47]; of cardiovascular disease: 54 participants (3.4%) vs 30 (1.0%; adjusted OR, 1.95 [95% CI, 1.20-3.18]); and high caffeine consumption: 141 (12.6%) vs 164 (7.3%; adjusted OR, 1.94 [95% CI, 1.32-2.85]) (Table 4). All of these outcomes were statistically significant when adjusted for potential confounders; however, when adjusted for missing values using the missing indicator method, the significance level for diabetes (P = .03) fell short of statistical significance when accounting for multiple comparisons (P =.0125).

At baseline, 1312 of 4608 participants (28.5%) who completed the ESS had scores of 11 or higher, which means that they experienced excessive sleepiness.²³ Of the total 4957 baseline survey respondents, 2276 (45.9%) reported having nodded off or fallen asleep while driving; 1294 of these (56.9%, 26.1% of the total cohort) reported falling asleep while driving at least 1 to 2 times a month; and 307 (13.5%, representing 6.2% of the total cohort) reported falling asleep while driving at least 1 to 2 times a week.

Those who screened positive for a sleep disorder also had a higher mean (SD) ESS score than those who did not (9.65 [4.25] vs 7.63 [3.80]; P < .001, Figure 2). Furthermore, 38.6% (95% CI, 36.4%-40.8%) of those who screened positive for any sleep disorder had an ESS score of 11 or higher vs 21.5% (95% CI, 19.9%-23.0%; P < .001) of those who did not.

During the 2-year follow-up, 15 735 monthly surveys were collected, 6587 person-months with positive screens for sleep disorders and 9148 with negative screens for a sleep disorder. Of the participants who screened positive for any sleep disorder and responded to the question, 17.9% reported making important administrative errors vs 12.7% of those who screened negative (adjusted OR, 1.43; 95% CI, 1.23-1.67), 14.4% vs 9.2% reported falling asleep while driving (adjusted OR, 1.51; 95% CI, 1.20-1.90), and 23.7% vs 15.5% reported making errors or committing safety violations due to fatigue (adjusted OR, 1.63; 95% CI, 1.43-1.85; Table 5 and eTable 5). Similarly, 34.1% vs 28.5% reported having uncontrolled anger toward a citizen or suspect (adjusted OR, 1.25; 95% CI, 1.09-1.43), 11.2% vs 9.4% incurring citizen complaints (adjusted OR 1.35; 95% CI, 1.13-1.61), 26.0% vs 20.9% absenteeism (adjusted OR, 1.23; 95% CI, 1.08-1.40), and 14.1% vs 7.0% falling asleep during meetings (OR, 1.95; 95% CI, 1.52-2.52). When these outcomes were analyzed as continuous variables rather than as dichotomized outcomes), the results were similar (eTable 6 and eTable 7).

To address possible reporting bias, we calculated these ratios for the 459 police officers who completed at least 1 year of monthly surveys (5508 person-months; eTable 8). The results were comparable, although falling asleep while driving and occupational injury did not reach statistical significance in the smaller sample.

Two hundred eighty-seven participants reported being in a motor vehicle crash during the follow-up period. Motor vehicle crashes were more likely to be reported by those who reported falling asleep while driving (29 of 180 person-months) than by those who did not (1132 of 9980 person-months; risk ratio [RR], 1.49 [95% CI, 1.01-2.20]; P < .05) or by those who reported falling asleep while stopped in traffic (25 of 177 person-months) than by those who did not (872 of 9898 person-months; RR, 1.68 [95% CI, 1.11-2.55], P = .01).

Of the 126 completed polysomnograph studies, 116 had complete data available. Forty-four of 54 participants who underwent polysomnography assessment and who screened positive for OSA on the survey (positive predictive value, 81.5%) were classified by the assessment as having mild to moderate, moderate to severe, or severe OSA compared with 28 of 60 (46.7%) of those who screened negative. Eleven of 60 (18.3%) of those who screened negative were found to have moderate to severe or severe OSA (eTable 9). We found that 35 of 44 (positive predictive value, 79.5%) of participants with a BMI of 30 or higher and 11 of 12 (positive predictive value, 91.6%) of participants with BMI of 35 or higher had mild to moderate, moderate to severe, or severe OSA (eTable 9).

Sleep disorders are common and are largely undiagnosed and untreated in North American police officers. Our comprehensive sleep disorders screening program found that 40.4% of police officers reported symptoms consistent with at least 1 sleep disorder. The majority reported not having been diagnosed in the past or not taking regular treatment. Obstructive sleep apnea was the most prevalent disorder, with one-third of officers screening positive. Given that obesity is a major risk factor for OSA³ and that one-third of our sample had a BMI of 30 or higher, the high prevalence of OSA could be anticipated but is nevertheless a concern. There are several factors that might account for the lower prevalence of OSA in the state police department cohort, including their lower reported BMI and the department's physical fitness program, which provides fitness facilities and the opportunity to exercise at all stations during paid work time and job performance standards associated with physical ability.

The OSA prevalence in our sample is comparable with recent reports,^32- 33 but higher than reported 2 decades ago.³⁴ Although only a subset of patients with OSA report excessive sleepiness, many describe related symptoms (eg, fatigue, nonrestorative sleep, inattention). The prevalence of OSA without a complaint of excessive sleepiness, even in 1993, was 24% in men and 9% in women.³⁴ Today, it is likely even higher, given that the prevalence of some of the major risk factors for OSA (eg, aging and obesity) is increasing. In 2009, self-reported obesity prevalence in the United States was 26.7%, up from 19.8% in 2000.³⁵ In addition, recent improvements in diagnostic technology would likely yield higher apnea prevalence estimates.

Although in-the-line-of-duty death rates in police have decreased by almost half since 1972, the proportion of deaths due to unintentional injury have shown little change and in 2003 were greater than the rate of felonious deaths.¹⁴ Across 2009-2010, more than one-third of in-the-line-of-duty deaths were due to motor vehicle crashes.³⁶ Driver sleepiness is a major cause of motor vehicle crashes,³ and excessively sleepy individuals have an increased risk of having more crashes and more serious crashes.³⁷ Obstructive sleep apnea exposes individuals to increased sleepiness and a 2- to 3-fold higher risk of motor vehicle crashes.⁸ We found that excessive sleepiness is common in police officers, with almost half reporting having fallen asleep while driving and about one-quarter reporting that this occurs 1 to 2 times per month. This is despite police officers apparently recognizing the dangers associated with drowsy driving; in a survey of North American police officers, almost 90% regarded drowsy driving to be as dangerous as drunk driving.³⁸

Police officers who screened positive for a sleep disorder were likely to report more actual and near-miss administrative errors and safety violations. The loss of even 2 hours of nightly sleep for 1 week is associated with decrements in performance comparable with those seen after 24 hours of continuous wakefulness.³ Sleep disorders resulting in chronic sleep deficiency may therefore adversely affect on-the-job performance. Because long work hours are also associated with decrements in performance and attentional failures,^13- 14 we adjusted for mean work hours in our analysis (eTable 5).

There may be a biological basis to our finding that those who screened positive for a sleep disorder were significantly more likely to report displaying uncontrolled anger toward a citizen or suspect. Yoo et al³⁹ studied changes in the amygdala with functional magnetic resonance imaging and reported that those in a sleep deprived state were unable to appropriately govern behavioral responses to negative emotional stimuli. This may also explain the self-reported increased number of citizen complaints filed against those officers who screened positive for a sleep disorder, although further studies are required to test this potential mechanism.

Cardiovascular disease–related morbidity and mortality and vascular markers associated with these are increased in police officers, and these increases are not fully explained by traditional risk factors.^40- 41 We found significantly increased risk of diagnosed cardiovascular disease and diabetes in those who screened positive for OSA, which is recognized as a major risk factor for cardiovascular disease, diabetes, hypertension, and stroke, independent of obesity.^3,6,42- 43 Untreated OSA may therefore contribute to the increased prevalence of cardiovascular disease and diabetes in police. Furthermore, sleep loss is associated with metabolic abnormalities and may be the pathway contributing to sleep-related increases in the risk of obesity and diabetes.³ Many police officers are at an even greater risk of these outcomes because they are often required to work overnight, on rotating shifts, or both. Impaired cardiometabolic responses are observed in healthy volunteers scheduled to eat and sleep out of phase from their habitual times,^44- 45 and night work greatly increases the risk of progression to diabetes.⁴⁶ These findings may at least in part explain the increased risk of cardiovascular disease and diabetes in shift workers,⁴⁷ in particular police officers.⁴⁸ We note, however, that although we adjusted our analysis for many of the possible variables that influence cardiovascular disease risk, behavioral factors such as physical activity and diet were not assessed and may also account for the observed increased risk.

Our finding of increased risk of adverse mental health outcomes (diagnosed depression and risk of burnout) with positive screening for sleep disorders has significant implications, given the reported increased risk for suicide in police officers.⁴⁹ Obstructive sleep apnea, in particular the excessive sleepiness symptoms, and insomnia are associated with depression.⁹ Research has also demonstrated relationships between sleep loss and mental health problems. Resident physicians, who work protracted hours as do police in many municipalities,¹⁴ have been found to have rates of depression and burnout roughly double those in the general population.⁵⁰ Depression and burnout typically develop over the first several months of residency as sleep loss increases⁵¹ and are associated with making more errors.⁵⁰ By analogy, officers who are burned out may be at increased risk of making mistakes on the job, which could compromise public safety; however, further studies will be needed to determine whether this is in fact the case.

This study had several limitations. By collecting data on a monthly basis, we attempted to reduce, but could not eliminate, the effect of recall bias. Performance outcomes and attentional failures were self-reported and therefore may be overreported or underreported. Although participants were assured of confidentiality, underreporting of work-related outcomes may have occurred, possibly due to the stigma associated with these. Additionally, reporting bias could have confounded the results if participants preferentially completed monthly surveys after having had a negative health or safety outcome. However, analysis of a subgroup that completed a full year of monthly surveys yielded a similar pattern of results. Because response rate could not be determined for this study, it is possible that those who participated may not be representative of all North American police officers, although demographic characteristics in municipal police department responders were similar to the entire department. The magnitude of difference in absolute risk of some of the outcomes we assessed was relatively small, and the clinical significance of such differences is unknown. Although we observed a positive relationship between sleep disorders screening result and health and safety outcomes (Table 3), the cross-sectional analysis cannot determine causality. Hypertension and BMI are used in the Berlin questionnaire to identify those at high risk of OSA, which could have contributed to the observed association between OSA risk and cardiovascular disease, even though we adjusted for these variables in our analyses. Finally, the shift work disorder questionnaire remains to be validated.

Questionnaire screening instruments are inherently less precise than objective tests. Although the Berlin questionnaire had a positive predictive value of 81.5% for detecting mild to moderate, moderate to severe, or severe OSA, and negative predictive value of 81.7% for those with moderate to severe or severe OSA, its performance in predicting OSA in our study did not match that reported initially.¹⁹ Several factors likely account for this difference. First, the initial validation of the Berlin quesionnaire was conducted among individuals preselected with high pretest probability for OSA, such as sleep clinic populations.¹⁹ Second, the techniques we used to record breathing abnormalities are much more sensitive than those used in many previous studies on the prevalence of sleep apnea,^34,52 including the initial studies used to validate the Berlin questionnaire in a sleep clinic population.¹⁹ Third, different criteria are now used for the definition of hypopnea,⁵³ influencing respiratory disturbance index values.⁵⁴ As a result of these increases in sensitivity, sleep apnea would now be detected in many individuals who were classified as not having sleep apnea in the initial validation studies for the Berlin questionnaire.¹⁹ This may explain why in our study the positive predictive value remained high for all levels of sleep apnea severity, whereas the negative predictive value was high only for those with more severe illness. Moreover, excessive sleepiness is a criterion for positive screening on the Berlin questionnaire, and given that our sample had high incidences of night shift work and chronic sleep loss, it is possible that excessive sleepiness associated with behaviorally induced insufficient sleep syndrome,⁵⁵ shift work disorder, or OSA were being detected. The utility of the Berlin questionnaire in predicting outcomes in our study may thus have been due in part to the clinical consequences of all 3 of these sleep disorders; for this reason, we extended our prospective analysis to all sleep disorders, rather than to OSA alone. Research to develop more specific diagnostic questionnaires that distinguish OSA from behaviorally induced insufficient sleep syndrome and shift work disorder would be useful because the therapeutic interventions differ.

In conclusion, a large proportion of police officers in our sample showed a positive sleep disorder screening result, which was associated with adverse health, safety, and performance outcomes. Further research is needed to determine whether sleep disorder prevention, screening, and treatment programs in occupational settings will reduce these risks.