The darkest hours: McCarthy et al. (2019) report increased risk for suicide from midnight to 3 am for U.S. veterans and civilians

Suicide is an increasing mental health crisis that not only plagues the United States, but all nations across the globe. The World Health Organization projects that 1.53 million people worldwide will die by suicide in the year 2020, with 10–20 times that number making a suicide attempt that year [1]. In 2017, nearly 50 000 people died by suicide in America alone [2]. Suicide was the second leading cause of death for all Americans ages 10–34, the fourth leading cause of death for Americans ages 35–54 [3], and the death rates for suicide have increased yearly in the United States since 2005 [4]. The statistics are even more disheartening for military Veterans, a population that is at high-risk for suicide. The most recent report from the U.S. Department of Veteran Affairs (VA) indicated that U.S. Veterans are 1.5 times more likely to die by suicide compared to non-Veterans, and this figure increases to 1.8 times more likely when comparing female Veterans to their civilian counterparts [5]. Overall, this translates to approximately 20 Veterans dying by suicide each day [6]. Given this tragic and unceasing loss of life, the VA and other organizations have introduced public health initiatives to combat this crisis, with an emphasis on identifying upstream risk factors as potential targets for treatment development [7]. A compelling and growing body of evidence suggests that sleep loss [8, 9] and nocturnal wakefulness [10], as demonstrated in a recent report in SLEEP from McCarthy et al. [11], are modifiable biobehavioral factors that may be critical to target as a means of preventing suicide in the future. Healthy sleep is essential to our physical health and mental well-being. Meta-analytic evidence has suggested that disturbed sleep is associated with psychiatric disorders (compared to healthy controls) [12] and increased risk for cardiovascular disease [13] and premature mortality [14]. Disturbed sleep has also been shown to be associated with increased risk for suicidal thoughts and behaviors [8, 15, 16]. A meta-analysis of 39 studies including 147 753 participants reported that sleep disturbances (defined as insomnia or nightmares) were associated with increased risk for suicidal ideation, attempts, and decedents, even when controlling for depressive symptoms [8]. A meta-analysis published in SLEEP earlier this year suggests that sleep disturbances (defined as symptoms of insomnia) were cross-sectionally and longitudinally associated with increased risk for suicidal ideation among adolescents [15]. Critically, a recent study from the VA not only replicated the positive relationship between sleep disturbance (i.e. insomnia) and suicide attempts in a Veteran population, but they additionally reported that receiving sleep medicine treatment significantly reduced the likelihood of a suicide attempt during a 6 month period in Veterans with sleep disorders [16]. While causal links between sleep and suicidality remain unknown, some potential mechanisms linking sleep and suicidality include impairments in decision making, abnormalities in serotonin (5-HT) function, and hyperarousal of the sympathetic nervous system [17]. In sum, these reports highlight the potentially life-saving role of sleep medicine as a treatment for individuals at risk for suicide. Beyond the continuity of our sleep, the timing of our sleep–wake patterns is also important for our health. Previous research has suggested that later sleep midpoint is associated with greater body mass index (BMI) [18] and that large changes in sleep midpoint from workdays to free days (i.e. social jet lag) is associated with poor metabolic health (e.g. greater fasting insulin, insulin resistance, BMI [19]). Circadian rhythm disturbances have been robustly associated with increased risk for the development of mood disorders [20, 21]. Additionally, individuals with a preference for going to bed later and waking up later (i.e. evening-type chronotype) have been shown to have greater suicidal thoughts and behaviors [22]. Because the body of research on this link to sleep timing is smaller than for the sleep–suicide link more generally, it remains to be seen whether modifying the timing of our 24 h rhythms may be beneficial for individuals at risk for suicide. Based on this converging evidence from the sleep and behavioral rhythms literature, Perlis et al. [10] examined whether nocturnal wakefulness is associated with increased risk for suicide. Aligning with previous work, their initial analysis replicated that the greatest probability of suicide occurred at noon. This group, however, made an important and novel contribution by then adjusting the denominator to include only the percentage of the population awake at each time, which represents the true population at risk for suicide—in other words, one must be awake in order to make a suicide attempt. In these adjusted analyses, the peak risk for suicide occurred at 2 am, suggesting that being awake in the middle of the night is a risk factor for suicide. Earlier this year in SLEEP, McCarthy et al. [11] replicated and extended the findings of Perlis et al. [10] by exploring civilian- and Veteran-specific sleep and suicide patterns. The purpose of this study was to determine the 24 h distributions of suicide and sleep patterns for Veterans and compare them to age-matched civilian populations. Given that Veterans have a higher incidence of suicide [5] and experience more nocturnal wakefulness compared to civilians [23], due in large part to greater prevalence of nightmares, post-traumatic stress disorder (PTSD), and insomnia symptoms [24–26], McCarthy et al. hypothesized that this increased nocturnal wakefulness would be associated with a greater proportion of nighttime suicides in U.S. Veterans compared to civilians. To test these hypotheses, McCarthy et al. [11] replicated procedures reported in [10] by using data from the American Time Use Survey (ATUS) [27] and the National Violent Death Reporting System (NVDRS) [28] as the basis for their analyses. The ATUS is a national telephone survey wherein participants report their activities over the previous 24 h period, and these data were used to calculate the proportion of the population expected to be awake during each hour of the day. The NVDRS is a reporting system of violent deaths based in 16–18 states during the period analyzed by McCarthy et al. [11], and was used in these analyses to determine temporal suicide patterns based upon the time of fatal injury (notably, 48% of data were unusable due to insufficient information). Together, these datasets were used to calculate standardized incidence ratios (SIR), which is the ratio of observed relative to expected suicides (assumes a constant rate of suicides across 24 h period) for the proportion of people awake during each 1 h period. The authors reported that results for U.S. Veterans closely resembled those of civilians and the general population reported by Perlis et al. [10]. According to the ATUS, 27% of Veterans were awake for at least one clock hour during a fixed nighttime interval of 2300 to 0500. This fixed interval does not allow for variability in desired sleep–wake patterns, and therefore may overestimate wakefulness for some and underestimate wakefulness for others. The raw proportion of observed suicides peaked between the hours of 1000 and 1200, but when accounting for the proportion of U.S. Veterans awake at each hour, peak incidence of suicides shifted to between 0000 and 0300 (p < .00001), with midnight conferring the greatest risk with incidence of suicide at 8 times the expected rate (see Figure 1). Critically, the time interval in which the proportion of Veteran deaths by suicide was significantly greater than expected encompassed all nighttime hours (2200–0600), while deaths by suicide during the early morning (0700–0900) and midday (1100–1200) hours were actually less than anticipated given the number of Veterans expected to be awake at those times. Results from [11]. The top graph shows Veteran sleep distribution across a 24 h period per ATUS. Center graph shows observed and scaled suicide percent across the 24 h period per NVDRS. The bottom graph shows suicide incidence ratios by clock hour, with the greatest incidence from 0000 to 0300. Reproduced from McCarthy et al. [11]. The implications of these findings are not only remarkable, but a matter of life and death. While gross numbers of deaths by suicide may peak at midday, the data reported here suggest that this may only be due to the sheer volume of civilians and Veterans awake and active at that time. Instead, it appears that it is during periods of nocturnal wakefulness that both groups may be most at risk, suggesting that sleep behaviors and suicidality may be inextricably linked. Even with the large effect sizes reported here, this relationship between sleep and suicide may be underestimated for two reasons. First, within the ATUS, participants were coded as ‘sleeping’ if they indicated that they were asleep at any point during a given clock hour, making the sleep disruption and nocturnal wakefulness statistics reported here a conservative estimate. Second, the relationship between sleep disturbance and midday suicides, when the sheer volume of deaths by suicide reaches its zenith, remains equivocal. Suicide decedents at this time of day may have obtained little or no sleep the night before, making sleep loss a potentially critical upstream risk factor for these suicides as well. McCarthy et al. [11] also reported some exploratory findings comparing age-matched civilian and Veteran groups. They hint at some nuanced results that the peak risk of suicide across 24 h may differ between the groups when they stratify by age groups. Even the authors, however, note that these analyses are “crude” at best and must be interpreted with caution because the majority of the sample (91%) were civilians and there were substantial differences in age, sex, employment status, education, socioeconomic status, and geographic region between groups. Additionally, the U.S. Veteran group used here was heavily skewed male (93%), served pre-9/11 (92%), and not a single active duty military member was included in the ATUS reports, hindering the generalizability within the group and extension to active military members. Other sources of bias result from the fact that the ATUS only captures activity during a single 24 h period, and therefore is only a “snapshot” in time that may not reflect habitual sleep behavior. Additionally, it is unknown whether the time awake during “nighttime hours” was due to preferred later bedtimes, difficulty initiating sleep, difficulty maintaining sleep, and/or work (in the case of shift workers), and these reasons for nocturnal wakefulness are critical to evaluate because they may be differentially associated with outcomes and require different treatment targets. These considerations highlight the primary limitation of this study, which was that these datasets were not specifically designed to explore the question at hand. While both refs. [10, 11] use the innovative approach of extrapolating across two large, publicly available datasets to shed light on a critically important, otherwise unexplored topic, the datasets sampled two distinct populations (one national, the other in ~18 states). As such, the relationship between sleeping behavior and temporal patterns of suicide could not be determined at the individual level, and therefore the associations of sleep duration, desired sleep–wake schedules, and symptoms of insomnia on suicidality could not be tested. While this limitation requires that these results be viewed as primarily hypotheses-generating, the creativity of the authors demonstrates the power of publicly available datasets to generate preliminary findings that, if mined responsibly, not only extend our knowledge but have tantalizing implications for the safety of millions. Despite reports such as this one demonstrating the importance of sleep behavior and disturbance on symptoms and incidence of suicidality, the relationship remains grossly underexplored beyond the sleep field. Fifty-four percent of suicide decedents had no known mental illness at the time of death [28], and addressing sleep issues may be a critical factor for saving these lives. Fortunately, treatment for poor sleep, such as cognitive behavioral therapy for insomnia (CBT-I), is both highly effective [29, 30] and potentially accessible via the internet [31], however it can be difficult to find trustworthy programs backed with empirical evidence (see [32] for review of current online programs, including indications of those with and without peer-reviewed evidence). Given its effectiveness, further efforts should be made in tracking and publicizing CBT-I treatment availability for both online [32] and in-person options (e.g. CBT-I provider directories maintained by the Society for Behavioral Medicine [33] and the University of Pennsylvania [34]). Additionally, because the sleep–suicide link has been sufficiently well-established, continuing to move research of potential behavioral and pharmacological sleep interventions through both clinical research [35] and randomized clinical trials (RCTs) [36, 37] is of paramount importance, as including sleep as a possible treatment target may improve both prevention and intervention strategies for suicidality. For example, CBT-I for U.S. Veterans with insomnia was shown to reduce both symptoms of insomnia and suicidal ideation at the end of treatment in an uncontrolled evaluation of the US Department of Veterans Affairs’ CBT-I training programs [35]. Importantly, this work laid the foundation for an ongoing RCT evaluating whether CBT-I is better than active control (psychoeducation about sleep) for treating insomnia and suicidal symptoms among US Veterans, and there is a six-month follow-up to determine if these results are maintained (Clinicaltrials.gov Identifier: NCT03603717). In addition to these efforts, there may be numerous other intervention strategies to combat the risk of disturbed sleep and nocturnal wakefulness, such as monitoring sleep patterns in real-time (i.e. on a nightly basis) among high-risk patients to provide an avenue for just-in-time adaptive interventions when a patient’s sleep patterns worsen [38]. Safety plans can also be implemented to provide social support and self-care techniques during these high-risk nocturnal hours. Unfortunately, many of these ideas remain mere speculation for now. As such, we implore clinical researchers to turn their attention to the sleep–suicide relationship and find ways to ease the burden of nocturnal wakefulness in both U.S. Veterans and the general public in an effort to prevent suicides during these darkest hours. The authors would like to thank their funding sources for supporting their work and ongoing training. T.J.C. is currently funded by the Research Training Program in Sleep, Circadian and Respiratory Neurobiology (NIH T32 HL007901) through the Division of Sleep Medicine at Harvard Medical School and Brigham & Women’s Hospital. M.A.B. is currently funded by the Cardiovascular Behavioral Medicine Research Training Grant (NIH T32 HL07560) through the National Heart, Lung, and Blood Institute. Conflict of interest statement. The authors have no conflicts of interest to report.