Association of insomnia phenotypes based on polysomnography-measured sleep duration with suicidal ideation and attempts

Abstract

Objective:

To assess the association of insomnia phenotypes, being insomnia with short sleep duration (ISSD) and insomnia with normal sleep duration (INSD), with suicidality in a randomly-selected population-based sample.

Methods:

Data were analyzed from the Penn State Adult Cohort. Participants (N=1741, 52.5y, 57.4% female) were randomly recruited from the general population between January 1990 through March 1999 and mortality data were available through December 2018. Insomnia symptoms were defined as self-reports of moderate-to-severe difficulties initiating or maintaining sleep, early morning awakening and non-restorative sleep, or having chronic insomnia (n=719). Short sleep duration was defined as <6h of in-lab polysomnography-measured sleep duration (n=879). Suicidality (SAI; n=102) was ascertained by a lifetime history of suicidal ideation (SI; n=84), suicide attempts (SA; n=48) or death by suicide (DBS; n =10).

Results:

Compared to normal sleepers who slept ≥ 6h, participants with ISSD and INSD were associated with 1.72-fold and 2.22-fold increased odds of SAI, respectively; these associations were significant for SI, with 2.09-fold and 2.24-fold increased odds, respectively, but not for SA, after adjusting for physical and mental health comorbidities. ISSD and INSD differed in age of onset before 30 years old and hospitalizations after SA.

Conclusions:

The results of this cohort study suggest that both INSD and ISSD phenotypes are associated with increased suicidal ideation, while the INSD phenotype has an earlier age of onset and is more likely to experience hospitalizations after attempting suicide. These results highlight the importance of targeting insomnia symptoms to help prevent suicide.

Introduction

Suicide is a rising cause of death worldwide and a major public health crisis.¹ In the United States alone, suicide rates have increased 33% between 1999 and 2019. In 2019, about 12 million American adults had serious suicidal thoughts, 3.5 million planned a suicide attempt, 1.4 million attempted suicide, and more than 47,500 individuals died by suicide.² Thus, it is essential to identify risk factors that can be modified to reduce suicidality (SAI), including suicidal ideation (SI) and suicide attempts (SA), to ultimately prevent death by suicide (DBS).

Disturbed and insufficient sleep have been identified as modifiable risk factors for SAI; specifically, insomnia symptoms have been shown to predict SI and SA across several cross-sectional and longitudinal studies.^3,4 A recent meta-analysis of longitudinal studies examining the association between sleep and suicide has found a small-to-medium effect for the association between insomnia symptoms and SI (d=0.45), SA (d=0.38), and DBS (d=0.30).⁴ Similarly, another meta-analysis of longitudinal studies found that insomnia symptoms conferred a 2.1-fold increased odds of SI, 1.78-fold odds of SA, and 1.54-fold odds of DBS.³ Further, insomnia symptoms predict SI trajectories⁵ and have been shown to predict SI above the effects of other associated psychological factors, such as hopelessness, anhedonia, depression, trauma, anxiety, or substance use.^6,7

Insomnia, however, is a heterogeneous condition with differences in terms of observed sleep duration, measured by polysomnography (PSG) or actigraphy, regardless of the severity of the self-reported symptoms.⁸ Accumulating evidence has shown that two insomnia phenotypes, namely insomnia with PSG-measured short sleep duration (ISSD) and insomnia with PSG-measured normal sleep duration (INSD), differ in their pathophysiologic mechanisms, natural course, cognitive-emotional characteristics, and associated adverse health outcomes, which may help understand insomnia heterogeneity and inform its subtyping,^9–11 as depicted in the conceptual model in Figure 1. Specifically, there is evidence that the two phenotypes differ in terms of their association with physiologic hyperarousal, cardiovascular, metabolic, and neurocognitive health outcomes. The ISSD phenotype has been associated with hyperactivity of the hypothalamic–pituitary–adrenal (HPA) and sympatho-medullary-adrenal (SMA) axes of the stress system,^12,13 as well as with increased risk of hypertension,^14,15 cardiovascular disease,^16,17 and metabolic alterations including type 2 diabetes¹⁸ relative to the INSD phenotype.

Figure 1. The insomnia with objective short and normal sleep duration phenotypes.

INSD = Insomnia with normal sleep duration. ISSD = Insomnia with short sleep duration. Taken with permission from Fernandez-Mendoza J. Insomnia with Objective Short Sleep Duration. In: Kushida CA (Ed.): Encyclopedia of Sleep and Circadian Rhythms, Reference Module in Neuroscience and Biobehavioral Psychology. 2nd Edition. Elsevier Academic Press, New York, February 2021, 1–9. doi.org/10.1016/B978-0-12-822963-7.00013-X.

Across insomnia phenotypes, there is an increased risk for psychopathology (Figure 1). Previous studies have shown that the INSD phenotype is associated with anxiety, rumination, depressed mood, dysfunctional beliefs, poor coping resources, and greater sleep misperception;^19–23 whereas the ISSD phenotype is associated with depressive mood, somatic symptoms, fatigue, and health concerns.^19,20,22,23 Additionally, research has shown that although both insomnia phenotypes are at an increased risk of developing depression, poor coping resources mediate the increased risk of incident depression associated with the INSD, but not ISSD, phenotype.²¹ Given the rigor of this prior phenotyping research and the known association of insomnia symptoms with SAI, it is imperative to examine whether the proposed insomnia phenotypes can further expand our understanding of the relative association of insomnia and PSG-measured sleep duration with suicidality. To our knowledge, this is the first study designed to investigate the relations between INSD and ISSD phenotypes and SAI, including SI, SA, and DBS. Therefore, no specific hypotheses were advanced, excepting that it was expected that INSD and ISSD phenotypes would be at greater risk for SAI relative to normal sleepers.

Methods

Participants

Data were collected as part of the Penn State Adult Cohort (PSAC), a randomly-selected, population-based sample of 1,741 adults, which used a two-phase protocol in order to recruit adult participants of various age groups to estimate the prevalence and risk factors for sleep disorders in adults.^24,25 The second phase of the study randomly recruited 741 men and 1,000 women to be studied in the sleep laboratory (response rates of 67.8% and 65.8%, respectively) between January 1990 and March 1999. Vital status was collected until December 2018. Statistical analyses were conducted between October 2021 and January 2022. Detailed descriptions of the sampling procedure and the cohort composition have been described previously.^15,18,24,25 Written informed consent was obtained. The study protocol study complied with the Declaration of Helsinki and was approved by Penn State Institutional Review Board.

Suicidality

SAI was assessed as part of the in-lab clinical history and physical exam via a standardized questionnaire administered by trained research staff. Participants reported on the presence of a lifetime history of suicide attempts (SA; n=48) or suicidal ideation (SI; n=84). Participants also reported on whether they had received treatment for SI or SA, including whether they were hospitalized. In addition, DBS was ascertained from death certificates for deceased individuals in the PSAC as of December 31, 2018, which were retrieved from the US Centers for Disease Control and Prevention. Participants were linked by the US Centers for Disease Control and Prevention to death records from the National Death Index for the years 1992 through 2018.^26–28 Standard guidelines and algorithms were followed to avoid data misclassification.²⁶ See the eMethods for details on DBS. A total of 102 subjects had a lifetime history of SAI, defined as either SI (n=84), SA (n=48) or DBS (n=10).

Polysomnography-Measured Sleep Duration

All participants were evaluated for one night in the sleep laboratory in sound-attenuated and light- and temperature-controlled rooms. During this evaluation, each participant was continuously monitored for 8 hours (fixed-time period) using 16-channel PSG, including electroencephalogram, electrooculogram, and electromyogram. Bedtimes were adjusted to conform to participants’ usual bedtimes, and participants were recorded between 22:00–23:00 and 06:00–07:00. The sleep recordings were subsequently scored independently, according to standardized criteria.²⁹ According to the median PSG-measured sleep efficiency rounded to meaningful numbers, the entire study sample was categorized into two groups: ≥50^th percentile (i.e., ≥75% sleep efficiency equivalent to ≥6 hours of sleep) and <50^th percentile (i.e., <75% sleep efficiency equivalent to <6 hours of sleep). This cutoff to define short sleep duration has been utilized in other studies and previously been shown to be associated with adverse cardiometabolic and neurocognitive outcomes in adults with insomnia symptoms.^{14–16,18,20,24}

Insomnia Symptoms

Commensurate with our previous studies, the presence of sleep difficulty was first established based on three levels of severity based on a standardized questionnaire.^{14,15,18,20,21,25} “Chronic insomnia” was defined by a self-reported complaint of insomnia with a duration of at least 1 year. “Poor sleep” was defined as a moderate-to-severe complaint, based on a 4-point Likert scale, of self-reported difficulty falling asleep, difficulty staying asleep, early morning awakening, and/or nonrestorative sleep. “Normal sleep” was defined by the absence of either of these two categories. Given the relatively low prevalence of SAI (n=102), the “chronic insomnia” and “poor sleep” groups, which showed similar frequency of SAI, were combined to represent the presence of “insomnia symptoms” (n=719). Thereafter, four sleep difficulty subgroups were established and used in the primary analyses; namely, normal sleep ≥6h, normal sleep <6h, insomnia symptoms ≥6h, and insomnia symptoms <6h. Please see eTable 1 for a detailed description of the data supporting this approach.

Other Key Measures

Physical health conditions, mental health problems, and substance use disorders were assessed as part of the in-lab clinical history and physical exam, as described elsewhere.^15,18,24,25 A composite score of global health comorbidities that captured both physical and mental health problems, including depression, was created²⁵ and was treated as a covariate. See the eMethods for details on the global health comorbidity index. The presence of substance use disorders (SUDs) was ascertained during the clinical history as a report of having been diagnosed or treated for alcohol use problems (n=51) or drug use problems (n=26) and treated as an additional covariate. Sex, race/ethnicity, age, and years of education were also treated as covariates.

Statistical Analysis

Hierarchical binary logistic regression models were estimated to examine predictors of SAI, SI, and SA. Given the relatively low prevalence of SAI in this sample, 1000 bootstrapped samples were drawn to provide stable estimates of the regression coefficients. In the first step of the regression models, demographic covariates were entered (i.e., sex, race/ethnicity, age) along with the sleep difficulty subgroups. In the second step, the global health comorbidity index reflecting the number of comorbid physical and mental health conditions was entered. In the final step, the presence of SUD was entered. This analytic approach was used to examine the association of sleep difficulty subgroups with SAI, SI, and SA. Finally, exploratory analyses were conducted in participants who endorsed SAI to determine if there were significant differences in terms of age of onset of SAI and if the participant received treatment for SAI. All statistical analyses were conducted in SPSS version 28 (IBM SPSS Corp.).

Results

The demographic characteristics of the sample stratified by the presence of SAI as well as SI, SA, and DBS are presented in Table 1. Participants with a history of SAI were more likely to be younger (OR = 0.97, 95% CI [0.96, 0.98]; p = 0.001), have more health comorbidities (OR = 1.55, 95% CI [1.43, 1.72]; p = 0.001) and SUDs (OR = 8.26, 95% CI [4.38, 14.54]; p = 0.001), whereas neither racial/ethnic minority status (OR = 0.63, 95% CI [0.27, 1.17]; p = 0.19) nor sex (OR = 1.21, 95% CI [0.81, 1.90]; p = 0.36) were significantly related to the presence of SAI.

Table 1.

Sample characteristics

	No SAI		SAI		SI		SA		DBS
	(n = 1639)		(n = 102)		(n = 84)		(n = 48)		(n = 10)
	M	SD	M	SD	M	SD	M	SD	M	SD
Age	52.80	13.48	46.99	11.35	46.56	11.42	45.67	10.02	51.00	10.78
Years of education	13.02	2.71	12.60	2.39	12.73	2.48	12.10	2.22	11.70	1.89
Global health comorbidity	2.91	1.86	4.99	2.52	5.15	2.43	5.13	2.64	3.40	2.59
Physical comorbidity	2.66	1.70	3.60	2.11	3.62	2.10	3.65	2.23	2.60	2.32
Mental comorbidity	0.28	0.60	1.65	1.22	1.80	1.20	1.85	1.34	1.30	1.42
	N	%	N	%	N	%	N	%	N	%
Sex
Male	702	42.83	39	38.24	27	32.14	14	29.17	8	80.00
Female	937	57.17	63	61.76	57	67.86	34	70.83	2	20.00
Race/ethnicity
Non-Hispanic white	1422	86.76	93	91.18	77	91.67	44	91.67	10	100.00
Minority	217	13.24	9	8.82	7	8.33	4	8.33	0	0.00
Substance use disorders
No	1592	97.13	82	80.39	68	80.95	34	70.83	6	60.00
Yes	47	2.87	20	19.61	16	19.05	14	29.17	4	40.00
Sleep difficulty
Normal sleep	992	60.52	30	29.41	22	26.19	12	25.00	6	60.00
Insomnia symptoms	647	39.48	72	70.59	62	73.81	36	75.00	4	40.00
PSG sleep duration
≥ 6h	806	49.18	56	54.90	45	53.57	31	64.58	5	50.00
< 6h	833	50.82	46	45.10	39	46.43	17	35.42	5	50.00
Sleep difficulty subgroups
Normal sleep ≥ 6h	512	31.24	15	14.71	11	13.10	8	16.67	3	30.00
Normal sleep < 6h	480	29.29	15	14.71	11	13.10	4	8.33	3	30.00
Insomnia symptoms ≥ 6h	294	17.94	41	40.20	34	40.48	23	47.92	2	20.00
Insomnia symptoms < 6h	353	21.54	31	30.39	28	33.33	13	27.08	2	20.00

Note. SAI = Participants who endorsed current or past suicide attempts or ideation or who died by suicide or due to non-medical causes with unknown intent. SA = Participants who endorsed a past suicide attempt. SI = Participants who endorsed current or past suicidal ideation. DBS = Participants who died by suicide or due to non-medical causes with unknown intent. PSG = Polysomnography.

Table 1 also shows the univariate association between sleep difficulty subgroups and SAI. Relative to normal sleepers who slept ≥ 6h, participants who reported insomnia symptoms and slept ≥ 6h (OR = 4.76, 95% CI [2.71, 9.23]; p = 0.001) and those who reported insomnia symptoms and slept < 6h (OR = 3.00, 95% CI [1.55, 5.64]; p = 0.001) were associated with significantly increased odds of SAI, whereas normal sleepers who slept < 6h were not associated with increased odds of SAI (OR = 1.07, 95% CI [0.50, 2.27]; p = 0.87). Figure 2 presents the prevalence of SAI, SI and SA across sleep difficulty subgroups.

Figure 2.

Suicidality prevalence across sleep difficulty subgroups based on insomnia symptoms and polysomnography-measured sleep duration.

Multivariable-adjusted hierarchical binary logistic regression results are shown in Table 2. In demographics-adjusted models (Step 1), participants with insomnia symptoms who slept < 6h and those who slept ≥ 6h were associated with 3.8-fold and 4.5-fold significantly increased odds of SAI, respectively, when compared to normal sleepers who slept ≥ 6h (Table 2). These associations were diminished after adjusting for physical and mental health comorbidities (Step 2) and SUD (Step 3). Compared to normal sleepers who slept ≥ 6h, participants with insomnia symptoms who slept < 6h and those who slept ≥ 6h were associated with 1.72-fold and 2.22-fold increased odds of SAI, respectively (Table 2). These fully-adjusted associations (Step 3) were significant for SI, with 2.09-fold and 2.24-fold increased odds for participants with insomnia symptoms who slept < 6h and those who slept ≥ 6h, respectively, but not for SA. Finally, normal sleepers who slept < 6h were not associated with significantly increased odds of SAI (Table 2).

Table 2.

Binary logistic regression models associating insomnia phenotypes with suicidality, suicidal ideation, and suicide attempts

Outcomes:	SAI	SI	SA
	(n = 102)	(n = 84)	(n = 48)
Step 1	OR (95%CI)	OR (95%CI)	OR (95%CI)
Sleep difficulty subgroups
Normal sleep ≥ 6h	1.00	1.00	1.00
Normal sleep < 6h	1.48 (0.64–3.19)	1.55 (0.60–4.25)	0.75 (0.11–2.83)
Insomnia symptoms ≥ 6h	4.52 (2.53–8.92)*	4.85 (2.43–12.68)*	4.30 (1.98–13.52)*
Insomnia symptoms < 6h	3.76 (2.02–7.50)*	4.68 (2.35–11.25)*	2.80 (1.15–8.70)*
Race/ethnicity	0.51 (0.20–0.94)*	0.45 (0.16–0.94)*	0.44 (0.08–1.11)
Sex	1.21 (0.75–1.98)	1.64 (1.01–2.80)*	1.83 (0.95–4.13)
Age	0.97 (0.95–0.98)*	0.96 (0.95–0.98)*	0.97 (0.94–0.98)*
Step 2
Sleep difficulty subgroups
Normal sleep ≥ 6h	1.00	1.00	1.00
Normal sleep < 6h	1.32 (0.56–2.94)	1.35 (0.53–3.67)	0.66 (0.11–2.55)
Insomnia symptoms ≥ 6h	2.46 (1.30–5.04)*	2.50 (1.22–6.37)*	2.24 (0.95–7.24)
Insomnia symptoms < 6h	1.96 (1.04–4.00)*	2.38 (1.18–5.61)*	1.44 (0.55–4.50)
Race/ethnicity	0.43 (0.16–0.81)*	0.37 (0.12–0.82)*	0.38 (0.06–1.00)
Sex	1.03 (0.63–1.66)	1.43 (0.85–2.54)	1.63 (0.82–3.87)
Age	0.94 (0.92–0.96)*	0.94 (0.91–0.96)*	0.94 (0.91–0.96)*
Global health comorbidity	1.66 (1.47–1.90)*	1.69 (1.51–1.97)*	1.61 (1.39–1.93)*
Step 3
Sleep difficulty subgroups
Normal sleep ≥ 6h	1.00	1.00	1.00
Normal sleep < 6h	1.32 (0.56–2.99)	1.35 (0.51–3.70)	0.66 (0.11–2.66)
Insomnia symptoms ≥ 6h	2.22 (1.15–4.60)*	2.24 (1.04–5.75)*	1.85 (0.69–6.30)
Insomnia symptoms < 6h	1.72 (0.90–3.58)	2.09 (1.02–4.94)*	1.10 (0.37–3.70)
Race/ethnicity	0.41 (0.14–0.82)*	0.36 (0.11–0.83)*	0.35 (0.04–1.03)
Sex	1.25 (0.75–2.11)	1.75 (1.02–3.19)*	2.44 (1.17–6.78)*
Age	0.95 (0.92–0.96)*	0.94 (0.92–0.96)*	0.94 (0.92–0.97)*
Global health comorbidity	1.62 (1.44–1.87)*	1.65 (1.48–1.93)*	1.55 (1.35–1.89)*
Substance use disorders	5.33 (2.55–10.65)*	4.65 (2.05–10.16)*	9.68 (4.07–27.28)*

Note. SAI = Participants who endorsed current or past suicide attempts or ideation or who died by suicide or due to non-medical causes with unknown intent. SA = Participants who endorsed a past suicide attempt. SI = Participants who endorsed current or past suicidal ideation. DBS = Participants who died by suicide or due to non-medical causes with unknown intent. Given the low prevalence of SAI, SI, and SA, 1000 bootstrapped samples were drawn to provide stable estimates for each outcome.

^*p < 0.05

Exploratory analyses showed that the age of onset for SAI significantly differed [F(1,66) = 4.76, p = .03] between participants with insomnia symptoms who slept ≥ 6h (M = 30.05, SD = 10.98) and participants with insomnia symptoms who slept < 6h (M = 36.20, SD = 12.05). A significantly greater proportion of participants with insomnia symptoms who slept ≥ 6h reported an age of onset before age 30 compared to participants with insomnia symptoms who slept < 6h (57.5% vs. 28.6%, respectively, Pearson Chi-square = 5.56, p = 0.02). Finally, a significantly greater proportion of participants with insomnia symptoms who slept ≥ 6h reported having received treatment for SA, including hospitalizations, compared to participants with insomnia symptoms who slept < 6h (63.9% vs. 36.1%, respectively, Pearson Chi-square = 4.56, p = 0.03), whereas no significant differences were found in terms of receiving treatment for SI (54.8% vs. 45.2%, respectively, Pearson Chi-square = 1.85, p = 0.17).

Discussion

This is the first study to examine the association of insomnia phenotypes with SAI in adults from the general population using self-reported measures of insomnia symptoms and PSG-measured sleep duration. After controlling for demographic characteristics as well as physical and mental health comorbidities, the ISSD and INSD phenotypes remained significantly associated with increased odds of SI, with 2.09-fold and 2.24-fold increased odds, respectively, but not with significantly increased odds of SA. Although both INSD and ISSD were associated with increased odds of SI, the INSD phenotype showed a 2.2-fold independent association with overall SAI. Together these findings suggest that psychiatric risk is a shared vulnerability across insomnia phenotypes, and that the comprehensive medical and psychiatric comorbidities examined herein account for some, but not all, of the association between insomnia phenotypes and overall SAI. This is particularly important for SA, as insomnia symptoms are highly comorbid with other physical and mental health conditions, including SUDs, that combined may multiply risk of completing suicide. Furthermore, our novel exploratory finding that the INSD phenotype was more likely to have had an onset of SAI in young adulthood and to have received hospitalization for SA, suggests that specific cognitive-emotional mechanisms may be at play in this insomnia phenotype that puts these individuals at risk of severe psychopathological outcomes earlier in lifespan. Overall, these results indicate that targeting insomnia symptoms may improve SAI prevention, and that INSD is a clinically relevant phenotype from a psychiatric standpoint, in similar manner that the ISSD is for cardiometabolic and neurocognitive morbidity. In addition, these data support the need for more precise therapeutic approaches targeting the underlying mechanisms and risk factors for suicidality, as they may differ across these insomnia phenotypes, particularly in the transition to young adulthood.

As shown in Figure 1, both the INSD and ISSD insomnia phenotypes have been shown to be associated with psychopathology, including maladaptive cognitive-emotional profiles and increased risk of depression. However, the shared vulnerability for psychopathology in the two phenotypes is thought to occur via different mechanisms. For example, the association between INSD and incident depression appears to be dependent upon poorer coping resources, whereas the relation between ISSD and depression is independent of such cognitive-emotional factors and may be due to HPA axis hyperactivity among other potential mechanisms.²¹ Although some studies have linked HPA hyperactivity with SAI,³⁰ a substantial body of literature has linked HPA hypoactivity to the presence of suicidality^31,32 and prospective attempts.^30,33 Thus, prior evidence suggests that phenotypes associated with normal or hypoactivity of the HPA axis, such as INSD, are indeed associated with greatest risk of SAI. These data suggest that INSD may incur in more severe mental health outcomes relative to ISSD. Further, these results highlight the importance of considering both PSG-measured and subjective sleep markers when stratifying suicide risk in vulnerable populations.

Although this study is the first to examine the relation between insomnia phenotypes and SAI using PSG-measured sleep duration, several studies have been conducted that examined the relation between self-reported sleep duration and suicidality. In extant work, self-reported short sleep duration has been associated with increased suicide risk,^34–36 with some investigators observing that insomnia symptoms are only associated with suicidality in individuals who self-report being short sleepers.³⁴ However, in the current study, phenotypes characterized by the presence of insomnia symptoms were similarly related to SAI regardless of PSG-measured sleep duration. Discrepant findings between self-reported vs. PSG-measured sleep duration are common in individuals with insomnia,^16,37,38 likely due to the so-called sleep misperception found in these individuals (i.e., underestimation of sleep duration),²⁰ which may indeed serve as marker of increased risk for psychopathology. Given the diverging findings based on operational definitions of sleep duration, it cannot be assumed that similar relations regarding sleep duration and suicide will be found between self-reported and laboratory-assessed measures.

Our analytic strategy also added to the body of literature linking other risk factors, beyond insomnia symptoms, to SAI risk. In particular, substance use was robustly associated with SAI, with individuals with SUDs at 5.3-fold increased odds of SAI even after controlling for demographic factors, health comorbidities, and insomnia phenotypes. It is, thus, important to consider the comorbidity between insomnia and mental health problems, particularly SUDs, when informing comprehensive models of suicide risk assessment and prevention. Continued monitoring of insomnia symptoms may capture individuals at risk for SAI, whereas the introduction of PSG-measured sleep measures may further identify those with the greatest risk, albeit not apparent, for adverse mental health outcomes, particularly in young adulthood. Interventions designed to reduce the impact of insomnia and related risk factors could be specifically targeted among individuals with elevated suicide risk. Together, these measures may reduce the personal and public health burden of suicide.

There are several limitations that should be considered when interpreting these results. First, SAI was defined based on a documented history of ideation (n=84) or attempts (n=48) or if a participant died by suicide or from non-medical causes with unspecified intent as per death certificates (n=10). This definition of SAI may not capture untreated or unreported cases and may explain the relatively low prevalence of SAI in this sample. Nevertheless, the 5.9% frequency of SAI in the sample was consistent with population-based estimates of SAI at the time of assessment in the US, with an estimated 5.6% reporting ideation and 0.7% attempting suicide.³⁹ Suicide rates have been steadily increasing since the 1990’s¹ and different risk factors may be more salient in predicting SAI in the 2020’s (e.g., social media exposure). Using clinical samples at elevated risk for suicidality could also elucidate the longitudinal relations between insomnia phenotypes and SAI. Nevertheless, our estimates of 2-fold increased odds of SI associated with insomnia symptoms replicate those of prior meta-analyses,³ supporting the rigor and reproducibility of study findings. Second, insomnia symptoms were classified solely based on self-reported nighttime sleep disturbances during a clinical history and physical exam, rather than a structured clinical interview including frequency or daytime impairment diagnostic criteria. Our previous studies have supported the face, construct, and predictive validity of the poor sleep and chronic insomnia definitions used in this population-based cohort.^{16,17,20,24,25,29} Third, the objective sleep duration in this study was based on one-night, 8-hour PSG, which may be influenced by the first-night effect⁴⁰ and not be representative of the participants’ habitual sleep duration in the home environment. Finally, the PSAC is a predominantly non-Hispanic white cohort with an average of about 13 years of education, thus, future population-based studies should incorporate longitudinal designs with multiple time points and multiple night recordings and replicate the findings herein in more racially/ethnically diverse cohorts, thereby extending the generalizability of our findings and informing the causal direction of the associations found.

In summary, this study adds to the literature supporting a significant relation between insomnia and suicidality. Indeed, the INSD and ISSD phenotypes were associated with 2-fold odds of SI independent of demographic variables and physical and mental health comorbidities. To determine the temporal relations between insomnia phenotypes and SAI, future studies would benefit from evaluating the association of these insomnia phenotypes using prospective designs as well as in more severe clinical samples where SI and SA are more prevalent. Finally, given the differential association of ISSD and INSD with distinct adverse health outcomes, the current data may also aid in developing more personalized treatment approaches among individuals with these insomnia phenotypes that target the underlying mechanisms responsible for their shared psychopathologic risk. Given the observed relations between insomnia phenotypes and SAI, clinical trials employing targeted sleep interventions in the INSD and ISSD phenotypes have the potential to elucidate differentials effect on suicidality and/or its associated risk factors or mechanisms (e.g., HPA axis, substance use, hopelessness, perceived burdensomeness, thwarted belongingness).