Longitude-based time zone partitions and rates of suicide

Abstract

Background:

Increasing evidence suggests that conditions with decreased morning and increased evening light exposure, including shift work, daylight-saving time, and eveningness, are associated with elevated mortality and suicide risk. Given that the alignment between the astronomical, biological, and social time varies across a time zone, with later-shifted daylight exposure in the western partition, we hypothesized that western time zone partitions would have higher suicide rates than eastern partitions.

Methods:

United States (U.S.) county-level suicide and demographic data, from 2010–2018, were obtained from a Centers for Disease Control database. Using longitude and latitude, counties were sorted into the western, middle, or eastern partition of their respective time zones, as well as the northern and southern halves of the U.S. Linear regressions were used to estimate the associations between suicide rates and time zone partitions, adjusting for gender, race, ethnicity, age group, and unemployment rates.

Results:

Data were available for 2,872 counties. Across the U.S., western partitions had statistically significantly higher rates of suicide compared to eastern partitions and averaged up to two additional yearly deaths per 100,000 people (p < .001).

Limitations:

Ecological design and limited adjustment for socioeconomic factors.

Conclusions:

To our knowledge, this is the first study of the relationship between longitude-based time zone partitions and suicide. The results were consistent with the hypothesized elevated suicide rates in the western partitions, and concordant with previous reports on cancer mortality and transportation fatalities. The next step is to retest the hypothesis with individual-level data, accounting for latitude, photoperiodic changes, daylight-saving time, geoclimatic variables, physical and mental health indicators, as well socioeconomic adversity and protection.

Introduction

In most organisms living on Earth, even when isolated from time cues such as the light-dark cycle, a highly conserved time-keeping neuroendocrine system drives internally (i.e., biologically) alternating days and nights, separated by brief dawn and dusks, manifesting distinct electrophysiological, metabolic, hormonal, and behavioral characteristics (Wehr et al., 2001). These internal fluctuations, called circadian rhythms (Patke et al., 2020), enable humans (Rusak, 1989) and other organisms (Pittenigh and Daan, 1976) to foresee and, with environmental input, adapt to the species-specific metabolic and social demands (and opportunities) of their nighttime and daytime environments (Bass and Takahashi, 2010; Patke et al., 2020). Zeitgebers (i.e., “time givers”) are those signals that modify the timing of endogenous circadian rhythms, which in humans have periods that are slightly longer than 24 hours (Czeisler et al., 1999), thereby entraining them to a 24-hour day. Visible light, specifically short-wavelength light, which activates melanopsin receptors in the ganglion cells of the retina and is transmitted from the retina to the central circadian pacemaker in the suprachiasmatic nuclei (SCN) via the retinohypothalamic tracts (Berson et al., 2002), is the most important zeitgeber for both nonhuman mammals (Refinetti, 2015) and humans (Duffy and Wright, 2005). Therefore, the daily alternation between the environmental day and night caused by the Earth’s rotation around its axis has a powerful impact on the timing of human circadian rhythms—in other words, the daily light-dark cycle synchronizes the internal, or biological, day and night with the external day and night (Stothard et al., 2017; Wright et al., 2013). However, social activities—such as occupational, academic, familial, and recreational events—can also affect circadian timing, such as by influencing when people are awake (i.e. with their eyes open) and are exposed to visible light, not only from the Sun, but also from artificial sources (Mistlberger and Skene, 2004). As such, any misalignment between a social schedule (e.g., waking up early for work) and preferred biological timing can disrupt circadian functioning (Wittmann et al., 2006). Realignment can be achieved by either shifting internal biological rhythms via exposure to light and dark or by delaying or advancing the timing of social demands.

Disruptions of circadian processes are linked with a host of physical and psychiatric disorders (Reis et al., 2022). For example, circadian disturbances are associated with despair behavior and anhedonia in rodent models (Mendoza, 2019), depressed mood in humans (Carpenter et al., 2021; Mendoza, 2019), and suicidal behaviors in patients with bipolar disorder (Benard et al., 2019). Circadian misalignment can result in disordered sleep, like insomnia (Flynn-Evans et al., 2017) or circadian rhythm sleep-wake disorders (Duffy et al., 2021). Certain occupations also affect health through circadian misalignment. For example, night shift work is associated with elevated risk of depression (Torquati et al., 2019), cardiovascular disease (Wang et al., 2021), and suicidal ideation (Kim et al., 2022).

Notably, people who prefer to be awake and active during evening hours, i.e., those with evening chronotypes, are at a higher risk for depression (Levandovski et al., 2011), cardiovascular disease, diabetes, and all-cause mortality (Knutson and von Schantz, 2018). While chronotype is often thought of as an inherent trait-like characteristic, “eveningness” could be in part perpetuated by decreased exposure to morning light and increased exposure to evening light (Skeldon et al., 2017; Wright et al., 2013), which delays the timing of circadian rhythms relative to the light-dark cycle (Casiraghi et al., 2020). Eveningness is associated with suicidal ideation in adolescents, as well as with specific risk factors for suicidal behavior, such as substance use disorders, impulsivity, and aggression (Gau et al., 2007). It is also associated with more severe depressive symptoms, as demonstrated by a recent meta-analysis (Norbury, 2021). Furthermore, while undergoing treatment with selective serotonin reuptake inhibitors (SSRIs, the most prescribed antidepressant), those with evening chronotypes reported trying a higher number of antidepressant medications, experiencing lower efficacy of SSRIs, and having higher levels of suicidality as compared to those with morning chronotypes (McGlashan et al., 2018). These findings may be related to the substantial impact that SSRIs have on the circadian system’s reaction to external light (McGlashan et al., 2018).

Social schedules are set by a complex interplay of factors such as school start times and the timing of dominant work and recreational activities. Furthermore, with the advent of global economic interactions and transmeridian travel, a clock time shared across the globe was conceptualized by dividing the Earth into 24 geographical time zones that are roughly equal in longitude degrees. While time zones have helped to facilitate an increasingly interconnected world, they encompass wide (and often arbitrary) geographical distances. As a result, they can induce a discrete and overlooked local desynchrony between the internal biological clock, which is driven to a great degree by the astrophysical clock, and the external “social clock.” For example, in the United States (U.S.), a resident on the western edge of a time zone can experience sunrise and sunset an hour or more later than a resident on the eastern edge, all while both follow the same social time, resulting in a net shift of natural light exposure towards evening hours. This disparate exposure to light has been associated with differences in circadian timing, meaning that, across a time zone, individuals systematically vary in how their circadian rhythms align with social time (Roenneberg et al., 2007).

Indirect evidence of the circadian-related consequences associated with intra-time zone location is provided by research into the health impacts of daylight-saving time (DST), during which the social clock is abruptly moved forward one hour, resulting in a shift of daylight away from morning and towards evening hours. As such, the transition to DST is akin to moving from the eastern region of a time zone to the western. DST is associated with an increase in medical encounters for cardiovascular diseases, immune disorders, and injuries in the weeks following the transition (Zhang et al., 2020). This transition is also associated with an acute increase in risk of fatal traffic accidents (Fritz et al., 2020) and suicide (Berk et al., 2008; Osborne-Christenson, 2022). Arguing against the idea that change itself, rather than the direction of change, is detrimental, the offset of DST, which shifts daylight back towards the morning and away from the evening, is not associated with an increase in adverse health outcomes (Fritz et al., 2020; Osborne-Christenson, 2022; Zhang et al., 2020).

Several recent studies have more directly evaluated the relationship between intra-time zone location and health outcomes. Giuntella and Mazzonna (2019) found that living in the western region of a U.S. time zone, compared to an eastern region, was associated with decreased total sleep time and a higher prevalence of obesity, diabetes, and cardiovascular diseases. Living in a western time zone region is also associated with elevated risk of traffic fatalities (Gentry et al., 2022). Relatedly, time zone location has implications for oncology, as there is accumulating evidence that chronic circadian rhythm disruption facilitates the initiation and progression of malignancies (Sulli et al., 2019). Long-term exposure to conditions that induce circadian misalignment, such as those experienced by shift workers and flight attendants, is associated with elevated cancer risk (Härmä et al., 2022; McNeely et al., 2018). In line with this, Borisenkov (2011) showed that cancer incidence and mortality increased from the eastern to western border of a time zone. Similarly, using U.S. county-level data, Gu et al. (2017) demonstrated that the risks of multiple types of cancers were highest in western regions of time zones, and Giuntella and Mazzonna (2019) found a higher incidence of breast cancer in western compared to eastern time zone regions. Importantly, such health conditions are also associated with an increased risk for suicide (Heinrich et al., 2022; Liu et al., 2018; Wang et al., 2017), further highlighting the need to evaluate the relationship between time zone location and suicide.

Consistent with the cancer mortality data and given that past studies have linked circadian disruption to suicidal ideation (Kim et al., 2022), suicide attempts (Benard et al., 2019), and suicide (Berk et al., 2008; Osborne-Christenson, 2022), we hypothesized that people living on the western side of a time zone are at an increased risk for suicide due to the relative shift of daylight away from the morning and towards the evening. The purpose of this study was to perform a preliminary assessment of the association between intra-time zone location (i.e., time zone partition) and suicide rates using U.S. county-level data and methodologies previously applied to cancer research (Gu et al., 2017).

Methods

Data sources

This was an analysis of public U.S. data compiled from multiple sources. County-level populations and death counts attributed to suicide were obtained for the years 2010–2018 from the Centers for Disease Control and Prevention Wide-ranging ONline Data for Epidemiologic Research (CDC WONDER) database (https://wonder.cdc.gov/). The location of each county seat (i.e., the county administrative center) was determined by geographical data from https://geohack.toolforge.org/, available on Wikipedia. Time zone information for each county was obtained from https://greenwichmeantime.com.

Time Zone Partitions

The time zone of each county seat, based on the standard time system, was used as the time zone for the respective county. The U.S. time zones included in this study were the Eastern Time Zone (EST; Greenwich Mean Time [GMT]-5), the Central Time Zone (CST; GMT-6), and the Mountain Time Zone (MST; GMT-7). The Pacific Time Zone (PST) of the U.S. was excluded (see below for rationale). The geographical location of each county seat was then used to establish its time zone partition (i.e., western, middle, or eastern partition). Each time zone was divided into three partitions based on longitude with respect to the meridian of the time zone (i.e., 75° west for EST, 90° west for CST, and 105° west for MST). Any county seat located to the east of its respective meridian was placed in the eastern partition of the time zone, any county seat located between the meridian and 5° west of the meridian was placed in the middle partition, and any county seat located further than 5° west of the meridian was placed in the western partition (Figure 1). All county seats were placed into the appropriate partition by their visual location upon a map. Multiple independent teams categorized all county seats, with discrepancies resolved by discussion and, when necessary, an independent party. Note that while division of time zones by longitude roughly divided each time zone into thirds, time zones are not geographically symmetrical across the U.S., meaning that the areas of the partitions are not equivalent across time zones. In the event that a county had multiple county seats, placement was decided using the county seat with the greatest population. Finally, counties were also divided based upon their location in the northern and southern halves of the U.S., determined by the latitude reflecting the 2010 mean center of the U.S. population (approximately 37.52° north), for use in exploring if latitude moderates the relationship between time zone partition and suicide.

Figure 1.

Representation of Time Zone Partitioning. The counties in the Mountain, Central, and Eastern time zones were each divided into three partitions (eastern, middle, and western) based on location of the county seat. Areas excluded from the analyses are highlighted in gray. The horizontal break in the map marks where the county was divided into northern and southern halves based on the mean center latitude of the 2010 census.

Variables

County-level population and mortality data from CDC WONDER were used to obtain the two dependent variables for this study: suicide rate (all methods) and suicide rate (violent methods only). A separate subgroup for violent suicide was created because previous research has found that light exposure may have a stronger association with violent suicide than non-violent suicide (e.g., Lambert et al., 2003; Vyssoki et al., 2014). First, the total number of deaths in each county from 2010–2018 were obtained for the following groups of ICD-10 codes: X60-X84, representing suicide (all methods); and X70-X82, representing suicide (violent methods only), which was based on previous research defining violent suicide as all methods other than poisoning (Ludwig and Dwivedi, 2018; Vyssoki et al., 2014). Note that CDC WONDER suppresses total deaths if less than 10 for a given county, which affected 420 (14.6%) of counties in the final dataset. To address this without unnecessarily removing counties, it was originally planned to impute a value of 5 (representing the midpoint between 0 and 10) for suppressed death totals. However, during analyses, it was found that a higher proportion of CDC suicide death data were suppressed for counties in western partitions, compared to middle and eastern partitions, and in the northern U.S., compared to the southern U.S., most likely due to the smaller average county populations in these regions. To avoid bias caused by inappropriate imputation, the decision was made to instead directly model the censored nature of the data (see Analyses section below).

Next, the cumulative population for each county over the 2010–2018 timespan was extracted. Additionally, the percentages of females, people aged 50 or older, and people from different racial and ethnicity groups were used as covariates in statistical analyses. To calculate these, subgroup-specific cumulative population counts for each county, from 2010–2018, were also extracted from CDC WONDER. The subgroup-specific population counts were then divided by the total population to obtain percentages. Regarding age, a cut point of 50 or older was chosen because this age has been implicated in previous research as the typical onset of age-related physiological changes, including circadian abnormalities, such as the timing and strength of melatonin and body temperature circadian rhythms (Keihani et al., 2022), and menopause (Gold, 2011). Finally, given the consistent evidence that unemployment is associated with elevated suicide risk (Classen and Dunn, 2012), the unemployment rate for each county was extracted from County Health Rankings (https://www.countyhealthrankings.org/) and also used as a covariate.

Exclusions

Various counties were excluded from analyses. Due to geographic constraints, the PST of the U.S. could not be divided into three partitions based on the rules applied to the other time zones. Therefore, all counties residing within the PST were excluded. Non-contiguous regions of the U.S. (e.g., Hawaii, Alaska) were similarly excluded. Counties within Arizona were excluded because the state does not follow DST. Finally, Washington D.C. and all counties with a 2010 census population of less than 1,500 were also excluded.

Analyses

Analyses were performed using R 4.2.1 (R Core Team, 2022). Alpha was set to 0.05. Linear regressions were used to evaluate the associations between variables. Death counts for suicide (all methods) and suicide (violent methods only) were included as dependent variables in separate models, along with county population as an offset, thereby modeling suicide rates. County-level covariates, included in all models, were the percentage of females, the percentage of people that were 50 years of age or older, the percentage of people from different racial and ethnicity groups, and the percentage of unemployment for the county. Depending on the model (see below), time zone partition and north/south location within the U.S. were included as independent variables.

As mentioned previously, suicide death counts were not suppressed evenly across time zone partitions. Therefore, left-censored Poisson regressions, which directly modeled the suicide count data (including the suppressed values), were estimated using the integrated nested Laplace approximation via the R package INLA (Martins et al., 2013; Yu, 2018). Default priors from the INLA package were used for all models. For each dependent variable, two models were estimated to determine if the north/south location variable should be included in analyses. A base model, with only time zone partition and covariates as variables, was evaluated against a full model that also included north/south location as both a main effect and as an interaction with time zone partition. The Deviation information criteria (DIC) metric was used to compare models, with lower DIC values indicating better fit. For both outcomes, DIC was lower for the full model that included the north/south location parameters (ΔDIC = 2,524.9 for suicide, all methods; ΔDIC = 3,013.7 for suicide, violent methods only). As such, full models were used to evaluate dependent variable differences across time zone partitions.

Because the primary goal of the study was to compare death rates across time zone partitions, the regression results of the two full models were used to calculate marginal means, adjusted for covariates, for each time zone partition separately for the northern and southern halves of the U.S. Then, adjusted mean differences for all possible partition comparisons were calculated (i.e., west vs. middle, west vs. east, middle vs. east) from the respective approximated posterior distributions based on 10,000 samples of the fitted models. Note that partitions in the north were not compared against those in the south. This resulted in a total of 6 partition comparisons per model, for a total of 12 partition comparisons. To account for multiple comparisons, the Benjamini-Hochberg false discovery rate procedure was used to adjust the p-values, defined as twice the proportion of the approximated posterior samples that did not cross zero (to equate to a two-tailed approach), and confidence intervals of all 12 partition comparisons (Yekutieli and Benjamini, 1999).

Finally, sensitivity analyses were conducted using the originally planned imputation strategy for suppressed values (i.e., 5), two alternative imputation values (i.e., 0 and 10), as well as removing counties with suppressed data. For sensitivity models, crude death rates per 100,000 people were calculated using the following formula, which reflects the average yearly death rate for a county: (sum of deaths from 2010–2018 / cumulative population from 2010–2018) * 100,000. Linear regressions were then conducted using the full models and non-parametric bootstrapping with 3,000 iterations was used to calculate 95% bias-corrected accelerated confidence intervals for all regression coefficients and comparisons of interest.

Results

A summary of the data is provided in Table 1. Data were available for 2,872 counties. In total, 1,394 counties were in the western partition of their respective time zones, 897 in the middle, and 581 in the east. Counties in eastern partitions tended to have larger populations, although there was substantial variability in population sizes. On average, ~38% of the people in each county were age 50 or older, slightly over 50% were female, ~86% were White, ~92% were not Hispanic or Latino, and unemployment rates were between 6–7%.

Table 1.

County Characteristics Across Time Zone Partitions

	West		Middle		East
	M	SD	M	SD	M	SD
Number of counties (N, %)	1,394	48.5	897	31.2	581	20.2
Cumulative population, 2010–2018 (1,000s)	722.7	2,100.4	716.6	1,391.4	1,123.20	2,981.5
Proportion 50 or older (%)	38.4	6.7	38.5	6.0	38.5	5.3
Proportion women (%)	49.8	2.3	50.2	2.1	50.3	2.1
Race
American Indian or Alaska Native (%)	1.9	5.2	1.9	6.3	1.6	7.1
Asian or Pacific Islander (%)	1.2	1.4	1.5	2.1	1.6	2.6
Black or African American (%)	8.0	12.4	13.6	17.7	11.2	15.1
White (%)	88.9	13.3	83.1	18.2	85.5	16.4
Ethnicity
Hispanic or Latino (%)	10.4	15.6	6.4	9.9	6.4	9.3
Not Hispanic or Latino (%)	89.6	15.6	93.6	9.9	93.6	9.3
Unemployment rate (%)	6.3	1.9	6.5	1.6	6.7	1.7

M = Mean; SD = Standard Deviation

Results for comparisons across time zone partitions are displayed in Table 2. Across the U.S., suicide rates (all methods and violent methods only) were significantly higher in the western partitions compared to their respective eastern counterparts. In the northern U.S., western partitions had an adjusted average of over 2 additional suicides per 100,000 people each year, both when looking at all suicide methods and when looking at violent methods only. In the southern U.S., western partitions had an adjusted average of nearly 1 additional suicide per 100,000 people each year for all suicide methods and nearly 0.5 additional suicides per 100,000 people each year for violent methods only. Western partitions were also found to have higher rates of suicide compared to middle partitions in the northern, but not southern, U.S. Partition-specific rates for each dependent variable are displayed in Figure 2.

Table 2.

Adjusted Death Rate Comparisons Across Time Zone Partitions in the North and South U.S.

Comparison	Est.	SE	Adj. lCI	Adj. uCI	Adj. p
Outcome: All Suicide
North: West - Mid	0.21	0.09	0.02	0.39	.036*
North: West - East	2.07	0.09	1.88	2.25	<.001***
North: Mid - East	1.86	0.09	1.69	2.04	<.001***
South: West - Mid	0.00	0.13	−0.26	0.24	0.976
South: West - East	0.97	0.13	0.71	1.23	<.001***
South: Mid - East	0.97	0.16	0.66	1.28	<.001***
Outcome: Violent Suicide
North: West - Mid	0.33	0.09	0.15	0.50	.001***
North: West - East	2.06	0.09	1.89	2.23	<.001***
North: Mid - East	1.73	0.09	1.56	1.89	<.001***
South: West - Mid	−0.02	0.12	−0.25	0.21	0.939
South: West - East	0.47	0.13	0.23	0.72	.001***
South: Mid - East	0.49	0.14	0.21	0.78	.002**

Adj. = Adjusted for multiple comparisons using the Benjamini-Hochberg false discovery rate procedure; lCI = adjusted lower limit of 95% credible interval; Est. = Estimated change in average yearly death rate across partitions (per 100,000 people); SE = Standard error; uCI = adjusted upper limit of 95% credible interval.

^*p < .05

^**p < .01

^***p < .001

Figure 2.

Estimated Adjusted Death Rates per 100,000 People Across Time Zone Partitions in the Northern and Southern U.S. Numbers represent the average yearly death rate. Error bars represent the standard error of the estimate obtained from the approximated posterior distribution. Time zone partitions in the northern half of the U.S. were not compared against those in the southern half. Top Panel: Northern U.S.; Bottom Panel: Southern U.S.; Left Side: Suicide (all methods); Right Side: Suicide (violent methods only).

*p < .05, **p < .01, ***p<.001 after adjusting for multiple comparisons using the Benjamini-Hochberg false discovery rate method.

As reported in Tables S1–S4, results of the sensitivity analyses were partly consistent with those of the censored Poisson regressions. In all but one model, rates of suicide using violent methods in the northern U.S. were higher in western vs. eastern partitions. However, results were more variable for all suicide methods and for partitions in the southern U.S.

Discussion

Overall, analyses of U.S. county-level data provided preliminary support for an association between time zone partition and rates of suicide. In both northern and southern halves of the U.S., counties in the western partition of a time zone displayed higher rates of suicide than their eastern counterparts. This was true for all methods of suicide as well as violent methods only. In the northern U.S., suicide rates (both for all methods and violent methods only) were also higher in western compared to middle partitions. To the best of our knowledge, this is the first study to report that western partitions within a time zone are associated with higher rates of suicide.

These findings are broadly consistent with previous studies showing an association between adverse health outcomes, including elevated rates of cancer and traffic accident mortality, and residence within the western regions of time zones (Borisenkov, 2011; Gentry et al., 2022; Giuntella and Mazzonna, 2019; Gu et al., 2017), as well as literature linking evening chronotype (Gau et al., 2007; McGlashan et al., 2018; Norbury, 2021) and DST (Berk et al., 2008; Osborne-Christenson, 2022) to suicide and related risk factors. One possible explanation for these findings is that circadian misalignment, caused by a relative shift in light exposure away from the biological morning and towards the early part of biological night, may contribute to suicide pathogenesis. If the present findings resist scrutiny in studies with individually linked data that also account for alternative physical and socioeconomic variables, it may lead to specific interventions involving timed light exposure and avoidance, melatonin or melatonin agonist administration, and potential schedule modifications to allow activity to match the internal as well as solar rhythms. It remains unknown if decreasing morning or increasing evening light exposure might be most beneficial for reducing the increased risk of suicide due to “westerness” within time zones. Both a decrease in exposure to morning light and an increase to that of evening light will delay the timing of circadian rhythms, as demonstrated by the phase response curve (PRC) to light (Khalsa et al., 2003). However, because human circadian rhythms have internal periods that tend to last slightly longer than 24 hours (Czeisler et al., 1999), the PRC to light suggests that morning light is critically important for entraining these rhythms to a 24-hour solar day. As such, delayed exposure to morning light may result in a loss of the daily physiological phase advance needed to maintain synchrony between the biological and social clocks.

Analyses of suicide data from countries around the world have found that suicide rates follow a seasonal pattern with a primary peak in spring, a time when both photoperiod and overall light exposure are increasing (Yu et al., 2020). This spring peak is strongest among individuals with mood disorders (Postolache et al., 2010) and in countries with greater seasonal variation in sunshine (Petridou et al., 2002). There is evidence that sunlight is consistently positively associated with suicide, and, importantly, that this association remains even when controlling for seasonality (Vyssoki et al., 2014). Note that a lengthening photoperiod from winter to summer increases exposure to light in both the morning and evening. Thus, the literature on seasonality and suicide suggests that the detrimental effects of evening light exposure may outweigh the beneficial effects of morning light.

While interpretation of the present results has been based in part on the circadian effects of light exposure, other unaccounted socioeconomic (e.g., income, education, medical and mental health services, violence) and health-related factors (e.g., cancer, metabolic and cardiovascular morbidity) may also contribute to elevated suicide rates in the western partition of time zones. Some of these effects are potentially a consequence of circadian dysregulation, such as reduced work productivity or propensity for depression and presenteeism/absenteeism, and thus could be mediators of the observed effects. However, the reverse is also possible, with socioeconomic and health-related factors affecting social schedules and the timing of sleep and wake, and thus the timing of eyes being open and the lights being on, which in turn determines when input from the retina reaches the central circadian pacemaker, the SCN. As such, circadian dysregulation may instead mediate (at least in part) the relationship between socioeconomic and health-related risk factors and suicide, or it may even be an unrelated third variable.

Sleep disruption could also be an important contributor to the present findings. Of relevance, individuals residing in western time zone partitions report an average of 19 minutes less sleep per night and are more likely to endorse having received insufficient sleep than their eastern counterparts (Giuntella and Mazzonna, 2019). The increased rates of suicide seen in the western partitions may therefore be attributable to increased suicidal ideation and behavior, which have been previously associated with reduced sleep duration (Baiden et al., 2020; Guo et al., 2017; Kim et al., 2013). Previous meta-analyses have found that insomnia is consistently associated with suicidal ideation and behavior (Liu et al., 2020), even when adjusting for mental illness (Malik et al., 2014), and may more than double the risk of such outcomes (Pigeon et al., 2012). The link between insomnia and suicide may be strongest in the context of short sleep durations (i.e., less than 7 hours per night; Hedström et al., 2021), highlighting the importance of sufficient sleep time in suicide prevention. The relative circadian phase delay experienced by individuals in western time zone partitions (Roenneberg et al., 2007) may also contribute to sleep onset insomnia (Flynn-Evans et al., 2017) and, by extension, elevated risk for suicide.

To further complicate the possible mechanisms behind the observed results, light has additional major biological effects that are not driven by circadian rhythms. Experimental animal studies have demonstrated that the intrinsically photosensitive retinal ganglion cells (ipRGCs) in the eyes are the common origin of photic information for circadian, learning, and mood pathways; however, the mood regulating pathways are distinct from the other two and bypass the SCN by connecting the ipRGCs with a recently described perihabenular nucleus (Fernandez et al., 2018). Furthermore, rodent studies have demonstrated that aberrant light exposure (e.g., during biological night) increases average corticosterone concentrations, exacerbates depressive-like behavior, and impairs hippocampal long-term potentiation and learning, even in the absence of circadian or sleep disruption (LeGates et al., 2012). Through direct brain stimulation, light has immediate energizing and antidepressant effects (Postolache and Oren, 2005). Thus, it is possible that, in addition to circadian shifting, the connection between light, particularly evening light, and elevated suicide risk could be due to mechanisms similar to those attributed to antidepressant medications, such as akathisia (i.e., restlessness), or, in the context of depression with pre-existing suicidal ideation, increased motivation and energy, and thereby increased capability to attempt suicide (Reeves and Ladner, 2010).

While it is unclear why the differences across western and eastern partitions appeared somewhat attenuated in the southern U.S., other studies have documented lower geo-temporal variation of suicide at lower latitudes. For example, a study in Chile, a country with a prolonged north-south orientation, found that the seasonal variation in suicide increased with growing latitude, with no seasonal variation detected at lower latitudes (Heerlein et al., 2006). Furthermore, larger overall (Davis and Lowell, 2002) and seasonal variation of suicide (Björkstén et al., 2009; Heerlein et al., 2006) has been reported at northern latitudes. Due to the Earth’s axial tilt, higher latitudes see larger fluctuations in the duration of daylength (i.e., photoperiod) over the course of a year, with considerably shorter days during winter and longer days during summer in comparison to lower latitudes. Importantly, seasonal variation in suicide rates has been positively associated with variation in sunshine (a composite of photoperiod and solar radiation; Petridou et al., 2002). Thus, it is possible that the amplitude of seasonal changes in photoperiod, which is most pronounced in the northern U.S., potentiates the effect of time zone partition-induced circadian misalignment on risk for suicide. Additional research is needed to investigate the interaction between latitude, longitude-based time zone partitions, seasons, and suicide.

Public Health and Clinical Implications

The above findings, if replicated in individually linked studies accounting for the many potentially confounding, moderating, and mediating variables, have important implications for public health. First and foremost, these results further emphasize the importance of establishing social schedules that follow the Sun, which requires consideration of longitudinal position within a time zone. Local initiatives that promote exposure to sunshine early after awakening (and reduce exposure in the evening) could have substantial health benefits. For example, communities, particularly those in the western partitions of time zones, might consider delaying the start of social activities, such as work or school. Not only might such a change improve health outcomes, but for children and adolescents it would likely improve academic performance as well (Wheaton et al., 2016). Such social adjustment can be seen in cities in the Central European Time Zone, like those in Spain, which may have adapted, at least somewhat, to discrepancies between social and solar time, with western cities evincing later average work-start times (Roenneberg et al., 2019). This compensatory collective behavior may help mitigate circadian disruption by aligning social schedules with preferred biological timing, although schedule-shifting alone may not be able to fully compensate for discrepancies in solar time (Roenneberg et al., 2019). Previous research in the U.S. has not found discrepancies in work start times across time zone borders, suggesting that such social adaptations are not taking place (Giuntella and Mazzonna, 2019).

Nationwide, maintaining standard time all year, as opposed to shifting to DST each spring, should also be considered. From a public health standpoint, current scientific evidence indicates that permanent standard time best supports physical and mental health due to the shift towards morning and away from evening light (Malow, 2022). This is relevant given the increase in suicide risk seen following the onset of DST (Berk et al., 2008; Osborne-Christenson, 2022). Another possible health-aligned strategy is to dynamically shift the spectra of light within buildings (e.g., schools, workplaces, inpatient psychiatric units) over the course of the day to support circadian health. Since the circadian system is most sensitive to short-wavelength light (i.e., blue light; Berson et al., 2002), light sources could shift from blue-enriched in the morning to blue-depleted (e.g., amber) in the evening, thus facilitating circadian alignment with the environment for most people.

For individuals, the findings of the present study highlight the potential importance of timed light exposure (and avoidance), as well as individual rest-activity schedules, for health, well-being, and performance. Bright light therapy could be used to enhance morning light exposure for people who are required to wake up prior to sunrise, particularly during winter months, provided they are not at risk for side effects such as headaches, intensified cycling such as in rapid cycling bipolar disorder (Benedetti, 2018), or retinal disease (Brouwer et al., 2017). Similarly, reducing evening exposure to both intense light or blue light-emitting screens (e.g., tv, phone, tablets, laptops), such as by wearing amber goggles (Kayumov et al., 2005), or blue-blocking glasses, could also address circadian misalignment related to time zone partition, particularly during summer months. Adequately timed low-dose melatonin, which has been found to reduce the risk of self-harm in children and adolescents (Leone et al., 2023), might counteract the effect of western partitions, particularly for those with demonstrated elevated suicide and circadian rhythm vulnerabilities, such as patients with bipolar I disorder (Benard et al., 2019) and those with delayed sleep phase (Sivertsen et al., 2021) or extreme eveningness (Gau et al., 2007). Clinicians may recommend individualized academic and occupational schedules in those most affected. In extreme cases where timing of social demands cannot be modified and light/melatonin interventions are ineffective, contraindicated, or resulting in unacceptable side effects, individuals could even consider relocating from the western edge of a time zone to the adjacent eastern edge of the bordering time zone, although this option would likely not be feasible for most individuals.

Limitations and Future Directions

The primary limitation of this study is its ecological design. Because ecological studies use aggregated, rather than individual-level, data, there is a risk that any patterns identified at the population level are spurious and do not exist within the individuals of said population, a phenomenon known as the ecological fallacy (Roumeliotis et al., 2021). While the present study provides a preliminary investigation into the possible relationship between time zone partitions and suicide, individual-level studies are needed to ensure that results are not biased. The true strength of ecological studies lies in their ability to generate and explore hypotheses that can then be evaluated with more robust research methods. Furthermore, our methods were cross-sectional and did not include interactions with temporal factors. Therefore, the presented results should be treated as supportive and hypothesis generating for future research, rather than conclusive.

Another limitation is that the present analyses only controlled for age, gender, race, ethnicity, and unemployment, not taking into account many other geoclimatic and socioeconomic variables. The high likelihood of multicollinearity among such ecological variables precludes their use in stepwise regression approaches (Graham, 2003). Furthermore, many clinical variables of relevance for suicidal behavior—such as mental illness, substance abuse, medical illness comorbidity, pharmacological treatments, or individual traits of aggression and impulsivity—were not analyzed. An additional limitation is how age was dichotomized using a cut point of 50 years or older, which may be perceived as somewhat arbitrary. However, this cut point was based on previous research into the onset of age-related circadian abnormalities (Keihani et al., 2022) and the typical age of menopause in the U.S. (Gold, 2011). For example, age-related circadian abnormalities include phase advanced melatonin rhythm, decrease in melatonin levels, strongly advanced core body temperature, and a decreased in the amplitude of the circadian rhythms (Keihani et al., 2022). Alternative approaches based on cut point optimization were considered but dismissed in favor of a physiologically meaningful cut point as attempted optimization can increase the risk of type I error (Altman et al., 1994).

Finally, sensitivity analyses indicated that the choice of imputation (or removal) of censored data affected estimation. This suggests that simple imputation and/or listwise deletion are insufficient for censored public health data. While the present study made use of statistical methods recommended to address this issue by directly modeling censoring (Yu, 2018), such findings further emphasize the need test the hypothesized association between time zone partitions and suicide rates using individual-level data.

Conclusions

The results of this ecological study supported the hypothesized association between the western time zone partition and rates of suicide. Additional research with individual-level and longitudinal data and larger variable sets is necessary to retest this ecologically confirmed association. There is also a need to investigate if and how individual exposure to other factors that directly influence circadian rhythms—e.g., , daylight-saving time and photoperiodic changes—interact with time zone-circadian vulnerabilities, such as evening chronotype, pre-existing mood or sleep disorders, socioeconomic adversity, and indicators of health. If a relationship is confirmed in individuals, innovative public health policies geared towards better aligning social schedules with the solar day and internal circadian rhythms, could reduce the increase in suicide mortality linked with residing within western longitudinal partitions of time zones.

Highlights.

• Social-solar time alignment varies within each time-zone, based on longitude

• Cancer mortality and fatal traffic accidents are higher in west vs. east time zone partitions

• We therefore set out to compare suicide mortality across time zone partitions

• In the U.S., suicide rates were higher in west vs. east time zone partitions

• Misalignments between circadian, light-dark, sleep-wake, and rest-activity rhythms are modifiable suicide risk factors