Sleep Disturbances Are Associated With Depressive Symptoms and Suicidality Among Adolescents in Pediatric Primary Care

Abstract

Objective

Sleep disturbances increase in adolescence and are associated with depression and suicide risk. However, research on enhancing depression and suicide risk assessment in clinical settings using behavioral markers, such as sleep, remains limited. Pediatric primary care (PPC) provides an opportunity for early risk identification, yet few studies examine how sleep disturbances, which are commonly reported, readily observable, and modifiable, are associated with concurrent depression and suicidality in PPC settings. This study examined the prevalence of adolescent sleep disturbances and concurrent associations with depression and suicidality in a large, sociodemographically diverse PPC sample.

Method

Between November 15, 2017, and February 1, 2020, 70,590 adolescents (ages 12-17) completed the Patient Health Questionnaire-9 modified for teens (PHQ-9-M) at PPC well visits. Logistic regressions examined associations between PHQ-9-M sleep item responses, sociodemographic variables, and depression and suicidality.

Results

Nearly 40% of youth endorsed sleep disturbances at their well visit via the PHQ-9-M sleep item. Youth who were older, female, racial or ethnic minoritized, and Medicaid insured were more likely to endorse sleep disturbances (odds ratios ≥1.03, ps < .001). Youth who endorsed sleep disturbances lasting several days or more were more than 20 times more likely to have a PHQ-9-M total score in the clinical range (odds ratio 23.31, 95% CI [21.64, 25.13]) and 4 times more likely to endorse any suicidality item (odds ratio 4.14, 95% CI [3.89, 4.41]).

Conclusion

Findings underscore the clinical utility of screening for sleep disturbances in PPC. Considering the significant associations between sleep disturbances and concurrent depression and suicidality, PPC providers may consider screening for and managing sleep disturbances.

Clinical guidance

• Sleep disturbances among adolescents should be standardly screened in primary care.

• Risk for depression and suicide may be increased in adolescents endorsing sleep disturbances in primary care.

• Consider implementing and/or referring for evidence-based approaches to manage sleep disturbances.

Diversity & Inclusion Statement

We worked to ensure sex and gender balance in the recruitment of human participants. We worked to ensure race, ethnic, and/or other types of diversity in the recruitment of human participants. We worked to ensure that the study questionnaires were prepared in an inclusive way. We actively worked to promote sex and gender balance in our author group. While citing references scientifically relevant for this work, we also actively worked to promote sex and gender balance in our reference list. While citing references scientifically relevant for this work, we also actively worked to promote inclusion of historically underrepresented racial and/or ethnic groups in science in our reference list. The author list of this paper includes contributors from the location and/or community where the research was conducted who participated in the data collection, design, analysis, and/or interpretation of the work. One or more of the authors of this paper received support from a program designed to increase minority representation in science. One or more of the authors of this paper self-identifies as a member of one or more historically underrepresented sexual and/or gender groups in science.

Plain language summary

This study utilized retrospective data from pediatric primary care well visits occurring between November 2017 and February 2020 from a large healthcare network. The goal was to determine the rate and sociodemographic characteristics of youth reporting sleep disturbances in well visits, as well as how sleep disturbances were associated with concurrent depression and suicidality. Of 70,590 adolescents (12-17 years old) with well-visit data, 37% endorsed at least some difficulty falling or staying asleep or sleeping too much. Youth who endorsed sleep disturbances were more likely to be older, female, from a minoritized racial or ethnic group, and Medicaid insured, and were 23 times more likely to also have elevated depressive symptoms and 4 times more likely to endorse suicidality. Findings highlight the importance of screening for sleep disturbances in pediatric primary care clinics.

Depression typically onsets during adolescence¹ and is characterized by greater symptom severity, longer episode duration, and increased rates of recurrence compared with adult-onset depression.^2,3 Of particular concern is the substantially elevated risk for suicidal thoughts and behaviors (STBs) among adolescents with depression.^2,4 With rates of STBs and depression among high school students increasing over time,⁵ enhanced identification of youth at risk for depression and suicide is of critical public health importance.

Most adolescents attend annual well visits with their primary care clinicians.⁶ Therefore, pediatric primary care (PPC) well visits provide an opportunity to identify risk of depression and STBs. The American Academy of Pediatrics Guidelines for Adolescent Depression in Primary Care (GLAD-PC) recommend routine universal screening for depression beginning at age 12,^7,8 listing the Patient Health Questionnaire-9-Item modified for teens (PHQ-9-M)^9,10 as a valid and reliable self-report depression screening measure. Accordingly, many health care systems now administer versions of the PHQ-9-M at pediatric well visits using predetermined cutoff scores to indicate the need for additional diagnostic screening and treatment. Increased screening with the PHQ-9-M in PPC has led to increases in identification and management of youth with depression and at risk for suicide.11, 12, 13 However, considering the transdiagnostic nature of STBs⁴ and the recognition that adolescents with depression and suicidality may not clearly identify low mood as their presenting symptom, GLAD-PC guidelines additionally recommend identifying other risk factors to enhance depression and suicide risk screening processes.⁷

Concurrent with increases in depression and STBs in adolescence are shifts in sleep and circadian patterns, resulting in high rates of insufficient sleep and other circadian disturbances.14, 15, 16 Research demonstrates robust associations between a range of sleep disturbances and depression and suicide risk in youth.^17,18 Sleep disturbances in youth not only precede depressive episode onset,^19,20 but also are associated with greater depressive symptom severity,²¹ recurrence following treatment,²² and poorer treatment response.^21,23 Furthermore, studies indicate an alarming link between sleep disturbances and suicide independent of depressed mood¹⁸; sleep disturbances have been cross-sectionally and prospectively associated with suicidal ideation, plan, and attempt and death by suicide, with studies demonstrating both proximal and distal predictive effects.24, 25, 26, 27

Critically, sleep disturbances represent a potentially less stigmatized, readily observable, and modifiable target for adolescent depression and suicide risk. Symptoms of mental illness in youth are typically more stigmatized than symptoms of physical illness,²⁸ and disclosure concerns are leading barriers to youth seeking help for mental illness.²⁹ In fact, a recent study showed that nearly half of adults with depression in primary care reported only physical symptoms, including sleep disturbances, to their physicians as the reason for consultation.³⁰ Thus, youth may be more likely to disclose sleep disturbances and/or parents may be more able to report on observed sleep disturbances over other internalizing warning signs of depression and STBs. Additionally, although it is well known that sleep disturbances have transdiagnostic effects on mental and physical health, there is evidence to suggest that some of those effects may be specific to increased depression and STB risk, indicating sleep as promising target for intervention.¹⁸ Specifically, mechanistic research shows that sleep disturbances lead to affective and behavioral dysregulation, including impulsivity, stress reactivity, and negative affect, all of which pose plausible pathways for increasing depression and STB risk.¹⁸ While research examining sleep interventions for depression and STBs is scarce, there is evidence that nonpharmacological sleep interventions not only are effective for improving sleep,³¹ but also result in reductions in both depressive symptoms³² and STBs.^33,34

Given the crucial role that sleep disturbances are hypothesized to play in adolescent depression and suicidality, early identification of these disturbances in PPC settings in concert with current screening efforts may offer promise in enhancing risk assessment, prevention, and intervention efforts. While meta-analyses examining sleep disturbances, depression, and STBs in youth demonstrate significant positive linkages overall,^{19,20,26,27,35} considerable heterogeneity exists across study methods, measures, definitions of sleep disturbances, and effect sizes, obscuring research to practice translation. Currently, only 2 studies have examined such associations in PPC settings, with only 1 study examining STBs as an outcome.^36,37 Results of both studies support associations between sleep disturbances, specifically short sleep duration and a sleep disorder diagnosis, and concurrent positive depression screenings. However, the focus of these studies on the impacts of sleep duration³⁷ and diagnosed sleep disorders³⁶ limits generalizations of risk estimates to adolescents who may endorse more general sleep disturbances (eg, difficulty falling asleep, staying asleep) or diagnostically subthreshold yet clinically impactful symptoms. Therefore, it is less clear how sleep disturbances may be related to depression and STBs specifically in PPC samples. Considering the important touchpoint PPC serves in adolescent health care and thus early identification of adolescent depression and suicidality, research elucidating the associations between endorsement of general sleep disturbances and concurrent depression and suicidality in this population is warranted.

The current lack of policies for systematically screening, documenting, and managing sleep disturbances in PPC further contributes to knowledge gaps. Electronic health record (EHR) data suggest that pediatric sleep disturbances are inconsistently screened for and underidentified in PPC, with lower screening rates in adolescents compared with younger children³⁸ despite data indicating high rates of insufficient sleep in adolescents.¹⁶ Without systematic screening efforts in adolescents in PPC, the extent to which sleep disturbances are associated with risk of depression and STBs in PPC samples remains unclear, as does the prevalence and sociodemographic characteristics of youth with sleep disturbances—data critical to further elucidate sleep health inequities and inform approaches to enhance sleep health.

The current study aimed to better understand the rate and sociodemographic characteristics of youth reporting sleep disturbances in PPC well visits, as well as how sleep disturbances are associated with concurrent depression and suicidality among a large, diverse PPC sample. The Children's Hospital of Philadelphia (CHOP) primary care network conducts universal adolescent depression screenings using the PHQ-9-M at adolescent well visits. Notably, the PHQ-9-M assesses symptoms of both insomnia (“trouble falling asleep or staying asleep”) and hypersomnia (“sleeping too much”), offering an existing method to systematically examine adolescent sleep disturbances in youth presenting to PPC. We therefore aimed to examine prevalence rates of sleep disturbances using the PHQ-9-M sleep item, identify sociodemographic differences in youth with vs without sleep disturbances, and determine associations between levels of sleep disturbance endorsement and concurrent depression and suicidality. Our results will enhance understanding of the extent to which sleep disturbances reported in PPC settings may serve as a useful clinical indicator of depressive symptoms and suicidality, meaningfully informing follow-up assessment and intervention strategies.

Method

Participants and Procedure

Participant data were from the EHR for adolescents ages 12 to 17 who were seen for PPC well visits between November 15, 2017, and February 1, 2020 at 31 CHOP PPC practices throughout urban (large, metropolitan) and suburban settings in Pennsylvania and New Jersey.³⁹ PPC well visits are a type of preventive care visit and as such are typically scheduled on an annual basis in accordance with American Academy of Pediatrics preventive care/periodicity schedule for providers to check in with families about their child’s health, development, and well-being and intervene accordingly.⁴⁰ CHOP primary care network began universal depression screenings using the PHQ-9-M at well visits for patients ages 12 and older in November 2017 to routinely assess for depression and suicidality. During the study period, there were 122,682 adolescent well visits for 82,531 unique adolescents. Adolescents completed the PHQ-9-M electronically on a personal device, tablet, or waiting room kiosk or during rooming for their well visit. For adolescents with repeat well visits during the study time frame, data from their first well visit were used. Adolescents who provided responses to all 9 core PHQ-9-M items were included in the current analyses. There were no exclusion criteria for EHR data extraction. For the current study, EHR data extraction was limited to the adolescent’s PHQ-9-M scores and sociodemographic information; other data regarding patient medical and diagnostic history, including visit diagnosis, major medical concern, or intake of prescription medication, if any, were not extracted. Detailed study procedures are published elsewhere.⁴¹

All study procedures were approved by the institutional review boards at CHOP and the University of Pittsburgh. A waiver of Health Insurance Portability and Accountability Act authorization and consent was applied.

Measures

Sociodemographic Information

Sociodemographic information including the adolescent’s age, sex assigned at birth, race, ethnicity, insurance type (Medicaid or private insurer), and practice location (urban, meaning a large central metropolitan area, or suburban, meaning sites located outside of the larger metropolitan area) were extracted from the EHR.

Patient Health Questionnaire-9 Item Modified for Teens

The PHQ-9-M is a modified version of the PHQ-9^9,10 that was developed to include wording relevant to youth depression (eg, irritability) and 4 supplemental items not included in the total score about severity/impairment (2 items) and suicidal ideation and attempt (2 items).

Depression

Consistent with previous work showing PHQ-9-M total score of 11 to be optimal for detecting major depression in youth in the United States,⁴² we categorized adolescents with total scores 11 or greater as having threshold depressive symptoms. Internal reliability for the 9 core items in the final analytic sample was good (Cronbach α = .80).

Suicidality

Adolescents who endorsed any PHQ-9-M item indicating past or current suicidality were considered at risk for suicide. Specifically, adolescents who endorsed “yes” on one or more of the supplemental PHQ-9-M STBs items (“Has there been a time in the past month when you had serious thoughts about ending your life?”; “Have you ever, in your whole life, tried to kill yourself or made a suicide attempt?”) and/or endorsed frequency of 1 (“several days”) or higher on item 9 (“Thoughts that you would be better off dead, or of hurting yourself”) were categorized as adolescents with suicidality.

Sleep Disturbances

Sleep disturbance was estimated using responses to the PHQ-9-M sleep item. The PHQ-9-M sleep item asks the adolescent how often they have been bothered by “trouble falling or staying asleep or sleeping too much” in the past 2 weeks on a scale of 0 (not at all) to 3 (nearly every day).

Analytic Plan

All analyses were conducted in R version 3.5.3.⁴³ Due to the skewed distribution of the PHQ-9-M sleep item responses (skew = 1.5, kurtosis = 1.3), we dichotomized responses to reflect different base frequencies of sleep disturbances in the prior 2 weeks. We present rates of endorsement of sleep disturbances (PHQ-9-M item 3) as cumulative percentages at the following levels: several days or more (sleep item score ≥1), half the days or more (sleep item score ≥2), and nearly every day (sleep item score 3). Sociodemographic characteristics of youth are presented at both the several days or more (≥1) and half the days or more (≥2) levels. Sleep item response levels were additive, meaning youth were counted in all categories that they qualified for (eg, someone who scored a 2 on the sleep item would be included at both the several days or more and half the days or more levels). To test for sociodemographic differences in youth who did and did not endorse sleep disturbances, we conducted simple binary logistic regressions predicting sleep item endorsement levels from sociodemographic characteristics.

To determine levels of sleep disturbances meaningfully associated with depression and suicidality, we first employed receiver operating characteristic curve analyses using the 'cutpointr' package⁴⁴ to test the predictive accuracy of the PHQ-9-M sleep item for concurrent threshold depressive symptoms and suicidality. Receiver operating characteristic curve analysis can be used to determine the diagnostic and predictive accuracy of a continuous variable, wherein one can calculate the area under the curve (AUC) and determine an optimal threshold that maximizes sensitivity without loss of specificity with respect to the characteristic of interest.⁴⁵ The levels that maximized sensitivity without loss of specificity were then used as cut points to dichotomize the sleep item for subsequent analyses in which we conducted multiple binary logistic regressions predicting threshold depressive symptoms and suicidality from PHQ-9-M sleep item responses, covarying for age, sex, race, ethnicity, insurance provider, and practice location. We report model results at 2 thresholds in tables; results for the optimal cut point per dependent variable are provided in the text. All effect sizes are reported as odds ratios (ORs). To be conservative, we additionally ran models examining associations using a prorated threshold depressive symptom score based on a scaled PHQ-9-M total score calculated without the sleep item. For models with suicidality as the outcome, we conducted sensitivity analyses additionally covarying for threshold and prorated threshold depressive symptoms. Variance inflation factors for all predictor variables in all models were less than 2 (all variance inflation factors ≤1.69).

Results

Sociodemographic and Clinical Characteristics

During the study period, 82,531 unique adolescents attended well visits, of whom 70,988 were screened. Compared with youth not screened, youth who were screened were more likely to be assigned female sex at birth; younger (12-14 vs 15-17 years old); White than Black, Asian, or another race; Hispanic/Latinx; and insured by Medicaid (Table S1, available online).⁴¹

A total of 70,590 (99.4%) screened adolescents completed all 9 core PHQ-9-M items and were included in the present analyses. The final sample was 50.3% male (n = 35,542) with a median age of 14 years (range 12-17 years). Most adolescents identified as non-Hispanic/Latinx (93%; n = 65,684) and White (56%; n = 39,735), followed by Black (27%; n = 18,990), another race not specified (11%; n = 8,079), Asian (4.%; n = 2,817), multiple races (1%; n = 777), American Indian or Alaska Native (0.1%; n = 68), and Native Hawaiian or other Pacific Islander (0.04%; n = 28). Nearly three-quarters of participants had private insurance (74%; n = 51,942) and attended practices in suburban locations (73%; n = 51,363).

Rates of Reported Sleep Disturbances, Depression, and Suicidality

Of the 70,590 youths with PHQ-9-M data, 37% (n = 25,875) endorsed difficulty falling/staying asleep or sleeping too much (PHQ-9-M item 3) for several days or more (≥1); 14% (9,891), for half the days or more (≥2); and 6.9% (4,840), nearly every day (3) over the 2 weeks before their well visit. Sleep disturbance was the most commonly endorsed PHQ-9-M item at the “half the days or more” (≥2) and “nearly every day” (3) thresholds (36.7%), second only to fatigue at the “several days or more” (≥1) threshold (fatigue = 40.2%).

Of all youth, 5.92% (n = 4,182) screened positive for threshold depressive symptoms, 7.2% (n = 5,075) screened positive for suicidality, and 3.07% (n = 2,166) screened positive for both. Of adolescents who screened positive for suicidality, 42.68% also screened positive for threshold depressive symptoms; of adolescents who screened positive for threshold depressive symptoms, 51.79% also screened positive for suicidality.

Table 1 shows the distribution of PHQ-9-M sleep item scores by depression and suicidality thresholds. Most adolescents who screened positive for threshold depressive symptoms (91.9%; n = 3,844), suicidality (69.7%, n = 3,537), or both (90.95%, n = 1,970) endorsed sleep disturbances at least several days or more in the past 2 weeks (≥1). Of adolescents who screened positive for suicidality but not threshold depressive symptoms (n = 2,909, 57% of adolescents with suicidality), 53.9% (n = 1,567) scored ≥1 on the sleep item. At the half the days or more threshold (≥2), 74% (n = 3,096) with threshold depressive symptoms, 42.7% (n = 2,166) with suicidality, 77.5% (n = 1,970) at risk for both, and 22% (n = 639) with suicidality but not threshold depressive symptoms endorsed sleep disturbances.

Table 1.

Distribution of Sleep Disturbances Item Scores by Depression and Suicidality Thresholds

PHQ-9-M sleep itema score	Threshold depressive symptoms				Suicidality
	Threshold depressive symptomsb (n = 4,182)		No or borderline depressive symptomsc (n = 66,408)		Suicidalityd (n = 5,075)		No suicidality (n = 65,515)
	n	(%)	n	(%)	n	(%)	n	(%)
0—“Not at all”	338	(8.1)	44,377	(66.8)	1,538	(30.3)	43,177	(65.9)
1—“Several days”	748	(17.9)	15,236	(22.9)	1,371	(27.0)	14,613	(22.3)
2—“Half the days”	1,058	(25.3)	3,393	(5.1)	943	(18.6)	4,108	(6.3)
3—“Nearly every day”	2,038	(48.7)	2,038	(3.1)	1,223	(24.1)	3,617	(5.5)

Note: PHQ-9-M = Patient Health Questionnaire-9 modified for teens.

^a“Trouble falling or staying asleep, or sleeping too much” over the past 2 weeks.

^bPHQ-9-M score ≥11.

^cPHQ-9-M score <11.

^dPositive response to any 1 (or more) of the following: PHQ-9-M item 9, supplemental suicidality item assessing “serious thoughts,” supplemental suicidality item “suicide attempt.”

Sleep and Sociodemographics

Adolescents who endorsed ≥1 on the PHQ-9-M sleep disturbances item were more likely to be older (OR 1.03, 95% CI [1.02, 1.04]) and assigned female sex at birth (OR 1.42, 95% CI [1.38, 1.47]). Compared with non-Hispanic/Latinx or White youth, those of Hispanic/Latinx (OR 1.22, 95% CI [1.15, 1.30]), Black (OR 1.19, 95% CI [1.15, 1.23]), or another racial and ethnic background (eg, American Indian or Alaska Native, multiple, or other races) (OR 1.12, 95% CI [1.07, 1.18]) were more likely to endorse sleep disturbances item ≥1. Youth who were insured through Medicaid compared with a private insurer (OR 1.26, 95% CI [1.22, 1.30]) and were seen at urban practice locations compared with suburban (OR 1.19, 95% CI [1.15, 1.23]) were also more likely to endorse sleep disturbances at the several days or more threshold. Sociodemographic differences between endorsers and nonendorsers remained significant when using a score ≥2 on the sleep item as a cutoff, with larger effect sizes via 95% CI comparisons for race (Black vs White: OR 1.71, 95% CI [1.63, 1.79]), insurance provider (Medicaid vs private: OR 1.71, 95% CI [1.63, 1.79]), and practice location (urban vs suburban: OR 1.52, 95% CI [1.45, 1.59]) (Table 2).

TABLE 2.

Sociodemographic Characteristics by Sleep Disturbances Item Score at Several Days or More Threshold and Half the Days or More Threshold

Sleep disturbances item score ≥1 (several days or more)
Characteristic	Sleep item score ≥1 (n = 25,875)		Score = 0 (n = 44,715)		OR	95% CI	z
	n	(%)	n	(%)
Sex at birth
Male	11,589	(44.79)	23,953	(53.57)	Reference
Female	14,289	(55.22)	20,761	(46.43)	1.422∗∗∗	1.379, 1.467	22.45
Ethnicity
Hispanic	1,914	(7.40)	2,747	(6.14)	1.22∗∗∗	1.149, 1.297	6.46
Non-Hispanic	23,870	(92.25)	41,814	(93.51)	Reference
Race
Asian	994	(3.84)	1,823	(4.08)	1.00	0.926, 1.086	0.07
Black	7,462	(28.84)	11,528	(25.78)	1.19∗∗∗	1.149, 1.234	9.59
Other	3,390	(13.10)	5,562	(12.44)	1.12∗∗∗	1.069, 1.175	4.73
Whitea	13,994	(54.08)	25,741	(57.57)	Reference
Insurance
Private	7,412	(28.65)	33,627	(75.20)	Reference
Medicaid	18,315	(70.78)	10,801	(24.16)	1.26∗∗∗	1.217, 1.304	13.09
Practice location
Suburban	18,256	(70.55)	33,107	(74.04)	Reference
Urban	7,619	(29.45)	11,608	(25.96)	1.19∗∗∗	1.150, 1.232	10.02
Age, y					1.03∗∗∗	1.019, 1.037	6.07
	Median	(IQR)	Median	(IQR)
	14	(12-16)	14	(12-16)
	Mean	(range)	Mean	(range)
	14.098	(12-17)	14.016	(12-17)
Sleep disturbances item score ≥2 (half the days or more)
Characteristic	Sleep item score ≥2 (n = 9,891)		Score <2 (n = 60,699)		OR	95% CI	z
n	(%)	n	(%)
Sex at birth
Male	4,118	(41.63)	31,424	(51.77)	Reference
Female	5,773	(58.37)	29,274	(48.23)	1.51∗∗∗	1.442, 1.571	18.61
Ethnicity
Hispanic	780	(7.89)	3,881	(6.39)	1.25∗∗∗	1.156, 1.357	5.53
Non-Hispanic	9,076	(91.76)	56,608	(93.26)	Reference
Race
Asian	311	(3.14)	2,506	(4.13)	0.92	0.812, 1.035	−1.37
Black	3,558	(35.97)	15,432	(25.42)	1.71∗∗∗	1.627, 1.789	22.07
Other	1,281	(12.95)	7,671	(12.64)	1.24∗∗∗	1.156, 1.320	6.24
Whitea	4,731	(47.83)	35,004	(57.67)	Reference
Insurance
Private	6,345	(64.15)	45,597	(75.12)	Reference
Medicaid	3,498	(35.37)	14,715	(24.24)	1.71∗∗∗	1.633, 1.787	23.19
Practice location
Suburban	6,451	(65.22)	44,912	(73.99)	Reference
Urban	3,440	(34.78)	15,787	(26.01)	1.52∗∗∗	1.450, 1.587	18.08
Age, y					1.03∗∗∗	1.019, 1.044	5.01
	Median	(IQR)	Median	(IQR)
	14	(13-16)	14	(12-16)
	Mean	(range)	Mean	(range)
	14.127	(12-17)	14.033	(12-17)

Note: ORs refer to fixed effects from simple logistic regressions predicting Patient Health Questionnaire-9 modified for teens sleep item threshold from each sociodemographic characteristic. Reference indicates the reference category for fixed effects for categorical predictor variables. IQR = interquartile range; OR = odds ratio.

∗∗∗ p < .001

^aRace is a sociopolitical construct; White race was selected as the reference group to reflect racial advantage/privilege.

Receiver Operating Characteristic Curve Analyses

Figure 1 depicts receiver operating characteristic test characteristics of the PHQ-9-M sleep disturbances item scores using our depressive symptoms and suicidality thresholds. The optimal cut point for the PHQ-9-M sleep item for detecting threshold depressive symptoms was ≥2 (half the days or more), at which the sleep disturbances item had sensitivity of 0.74 (95% CI [0.73, 0.75]) and specificity of 0.90 (95% CI [0.90, 0.90]) for detecting threshold depressive symptoms. Conversely, the optimal cut point for suicidality was ≥1 (several days or more), at which the sleep disturbances item had sensitivity of 0.70 (95% CI [0.68, 0.71]) and specificity of 0.66 (95% CI [0.66, 0.66]) for detecting suicidality. AUC was good for threshold depressive symptoms (AUC 0.88, 95% CI [0.87, 0.88]) and fair for suicidality (AUC 0.71, 95% CI [0.71, 0.72]).

Figure 1.

Receiver Operator Curves for (A) Threshold Depressive Symptoms and (B) Suicidality

Note:AUC = area under the curve.

Sleep and Depressive Symptoms

Adolescents who scored ≥2 on the PHQ-9-M sleep item (half the days or more) were at increased odds of also endorsing concurrent threshold depressive symptoms by 23.31 times compared with adolescents who scored <2 (OR 23.31, 95% CI [21.64, 25.13]). Adolescents who endorsed sleep disturbances were also more likely to report threshold depressive symptoms when determining threshold symptoms from a scaled PHQ-9-M score (sleep item removed) (OR 13.09, 95% CI [12.15, 14.11]) (Table 3).

TABLE 3.

Sleep Disturbances Item Score Associations With Concurrent Depressive Symptoms and Suicidality

Outcome	Sleep disturbances item score ≥1				Sleep disturbances item score ≥2
	OR	95% CI	z	Model R2	OR	95% CI	z	Model R2
Threshold depressive symptoms	21.71∗∗∗	19.41, 24.37	53.11	0.21	23.31∗∗∗	21.64, 25.13	82.61	0.28
Prorated depressive symptoms	11.40∗∗∗	10.34, 12.61	48.11	0.16	13.09∗∗∗	12.15, 14.11	67.26	0.20
Any suicidality	4.14∗∗∗	3.89, 4.41	44.23	0.10	4.94∗∗∗	4.64, 5.26	50.61	0.10
Covarying for threshold depressive symptoms	2.09∗∗∗	1.95, 2.25	20.17	0.22	1.58∗∗∗	1.45, 1.72	10.73	0.21
Covarying for prorated depressive symptoms	2.47∗∗∗	2.30, 2.65	25.49	0.22	2.27∗∗∗	2.10, 2.45	21.32	0.22
Item 9 score ≥1	5.76∗∗∗	5.30, 6.37	41.01	0.10	6.25∗∗∗	5.80, 6.73	48.54	0.11
Covarying for threshold depressive symptoms	2.19∗∗∗	1.96, 2.38	15.27	0.27	1.24∗∗∗	1.11, 1.37	3.90	0.28
Covarying for prorated depressive symptoms	2.76∗∗∗	2.52, 3.04	21.20	0.29	2.10∗∗∗	1.91, 2.31	15.37	0.28
Past month serious ideation	4.25∗∗∗	3.81, 4.73	26.30	0.08	5.02∗∗∗	4.54, 5.54	31.76	0.09
Covarying for threshold depressive symptoms	1.75∗∗∗	1.54, 1.99	8.59	0.20	1.25∗∗∗	1.09, 1.43	3.27	0.20
Covarying for prorated depressive symptoms	2.09∗∗∗	1.85, 2.36	11.94	0.21	1.79∗∗∗	1.58, 2.02	9.35	0.20
Lifetime attempt	3.23∗∗∗	2.96, 3.54	25.35	0.09	3.70∗∗∗	3.38, 4.04	28.60	0.09
Covarying for threshold depressive symptoms	2.03∗∗∗	1.84, 2.25	13.75	0.14	1.74∗∗∗	1.55, 1.94	9.57	0.13
Covarying for prorated depressive symptoms	2.22∗∗∗	2.01, 2.45	15.97	0.14	2.09∗∗∗	1.88, 2.31	13.86	0.14

Note: Fixed effects from multiple logistic regressions predicting risk outcomes from the Patient Health Questionnaire-9 modified for teens sleep disturbances item at several days or more threshold (≥1) and half the days or more threshold (≥2). All models covaried for age, sex at birth, race, ethnicity, practice location, and insurance type. OR = odds ratio.

∗∗∗p <.001.

Sleep and Suicidality

Adolescents who scored ≥1 on the PHQ-9-M sleep item (several days or more) were at greater odds for endorsing for suicidal ideation (item 9) (OR 5.76, 95% CI [5.30, 6.37]), serious ideation in the past month (OR 4.25, 95% CI [3.81, 4.73]), lifetime suicide attempt (OR 3.23, 95% CI [2.96, 3.54]), and overall suicidality (OR 4.14, 95% CI [3.89, 4.41]). Adolescents who endorsed sleep disturbances remained more likely to endorse any suicide item even after covarying for threshold depressive symptoms (OR 2.09, 95% CI [1.95, 2.25]) (Table 3).

Discussion

The current study examined the prevalence of adolescent sleep disturbances and concurrent associations with depressive symptoms and suicidality in a large, sociodemographically diverse PPC sample. Results indicated high endorsement rates of sleep disturbances overall, with higher rates in youth who were female, of racial and ethnic minoritized backgrounds, and insured by Medicaid. We observed medium and large total effect sizes between sleep disturbances and concurrent threshold depressive symptoms and suicidality. These results underscore the clinical utility of screening for sleep disturbances in PPC, as these disturbances signal coinciding depression and STBs. Given that sleep is a fundamental pillar of health and often less stigmatized than other depressive symptoms and indicators of suicide risk, our findings can meaningfully inform PPC screening and intervention strategies.

More than one-third (37%) of youth in the study endorsed trouble falling asleep, staying asleep, or sleeping too much at least several days in the 2 weeks before screening; 14% reported sleep disturbances more than half the days. These figures are consistent with prior research showing 25% to 35% of youth experience sleep disturbances.^15,46 Also more likely to endorse sleep disturbances were youth who were female; those of Hispanic/Latinx, Black, or another racial background; those insured by Medicaid; and those seen in urban practice locations. The magnitude of these sociodemographic effect sizes was relatively small and should be interpreted with caution, as it is possible that the statistical significance of such effects may be a result of the large sample size. However, it is noteworthy that these effects are consistent with well-documented gender, racial, ethnic, and socioeconomic sleep disparities in US children and adolescents^47,48 and, as such, warrant further examination in future research, especially considering the larger implications that small effects can have on health disparities when considered on a national scale. Our findings also map onto known racial and ethnic disparities in risk for depression and STBs^4,5,41 and poorer health care outcomes in general in the United States.⁴⁹ Given robust evidence demonstrating a prospective association between sleep and poor mental and physical health outcomes, insufficient and poorer quality sleep may contribute to and exacerbate other health disparities.⁴⁷ As race and ethnicity are sociopolitical constructs reflecting historical and ongoing marginalization due to racism and discrimination,⁵⁰ additional research that directly measures these experiences is needed to inform efforts to promote sleep health equity, particularly among youth with co-occurring depression and STBs.^51,52

Youth who endorsed sleep disturbances were also more likely to report concurrent depression and suicidality. Despite the limited scope of the PHQ-9-M sleep item, scores predicted threshold depressive symptoms and suicidality with good accuracy for depressive symptoms and fair accuracy for suicidality. Although PHQ-9-M sleep item cut points differed for predicting depression and suicidality (ie, higher cut point for depression), results were stark for both outcomes at the several days or more level. Nearly 3 in 20 youths who reported sleep disturbances several days or more screened positive for clinically significant depressive symptoms, and more than 1 in 10 screened positive for suicidality. Youth who endorsed sleep disturbances several days or more were over 20 times more likely to have a PHQ-9-M total score indicating a probable major depressive disorder diagnosis and 4 times more likely to endorse any item indicating past or current suicidality. The large effect size between sleep and co-occurring depression was consistent with that found in a previous meta-analysis.³⁵ Of note, the current study did find slightly larger effect sizes than those in meta-analyses linking sleep and concurrent suicidality,²⁶ possibly attributable to less measurement heterogeneity in our US-based PPC sample and use of a single measure to assess both constructs.

Taken together, these results suggest that although sleep disturbances are not deterministic of risk, their presence may indicate that additional screening for depression and suicide is warranted, especially in health care settings that do not administer routine screenings. Similar to a fever, sleep disturbances could be an indicator of a broad range of co-occurring health problems, including, as demonstrated by our data, significant depressive symptoms and suicidality. Indeed, whereas most youth with sleep disturbances did not screen positive for depression or suicidality, the increased odds of a concurrent positive screen associated with even just several days of sleep disturbances (20-fold for depressive symptoms and 4-fold for suicidality) are noteworthy. Although strongly recommended by national leaders in adolescent mental health,^7,8 many PPC practices lack procedures for conducting routine adolescent depression and suicide screenings or miss cases by failing to screen all adolescents.⁴⁰ Therefore, identifying other risk factors such as sleep disturbances to prompt additional screening and case management could help address these limitations. Furthermore, considering that sleep disturbances are a common precursor of subsequent depression and suicidality^19,20,27 (next-day,²⁴ next-month²⁵), attention to the PHQ-9-M sleep item and other sleep complaints could identify prodromal youth.

The clinical implications of our study are significant. Current protocols for documenting and addressing sleep disturbances in PPC are ambiguous.³⁸ Our results suggest that endorsing even several days of trouble falling asleep, staying asleep, or sleeping too much over 2 weeks can indicate clinically significant risk for concurrent depression and suicidality. Considering that youth may be more willing to disclose physical over internalizing symptoms due to stigma,³⁰ routine screening for sleep disturbances in PPC appears to be warranted. This could include attention to the PHQ-9-M sleep item for health care systems that administer universal screenings and/or administration of additional sleep screening tools.³⁸ Furthermore, PPC clinicians may consider targeting sleep disturbances among at-risk youth. Indeed, evidence suggests that addressing sleep with youth may aid in the prevention and treatment of depression³² and STBs.^33,34 Moreover, considering gender, racial, ethnic, and socioeconomic disparities in sleep health and the transdiagnostic nature of sleep disturbances, improving access to sleep health education and interventions could help reduce system-wide health care disparities.⁴⁷ Currently, even when youth are diagnosed with a sleep disturbance by their PPC provider, most do not receive a recommendation, treatment, or referral.³⁸ Future research focused on brief PPC-based adolescent sleep interventions could help prevent and treat depression and STBs and potentially other mental and physical health problems.^20,51

Our results should be considered within the context of study limitations. First, we relied on a single PHQ-9-M item with a limited response scale as a measure of sleep disturbances over the prior 2 weeks. More comprehensive and reliable sleep measures may reveal more nuanced prevalence rates and associations between sleep, sociodemographic factors, and concurrent depression and suicidality. Second, our measure of depression risk (ie, PHQ-9-M total scores) includes the sleep item score, creating potential for multicollinearity and construct overlap. While model variance inflation factors were small, and sensitivity analyses indicate large effects between sleep and depressive symptoms when calculating a prorated PHQ-9-M risk score without the sleep item, distinguishing the 2 constructs can be difficult considering sleep disturbances are a symptom of depression, and risk thresholds for PHQ-9-M total scores are based on inclusion of sleep as a core symptom.⁴² Moreover, despite evidence that PHQ-9-M total scores are reasonably sensitive and specific for detecting depression in youth,⁴² GLAD-PC guidelines still recommend semistructured diagnostic interviews to confirm a depression diagnosis.⁷ The current study did not have access to follow-up diagnostic data to either confirm or refute depression diagnosis or prognosis in this sample, and therefore an important future direction of this work will be to examine prospective associations between sleep, PHQ-9-M screener outcomes, and clinician-rated diagnoses to better understand the course and interrelationship between sleep and elevated depressive symptoms vs diagnosis in PPC.

Due to scope of the current study and limited access to adolescents’ medical records, we did not examine the influence of other proximal risk factors or symptoms on depression and STBs, and we did not examine how these variables may relate to mental health or diagnostic treatment histories. Specifically, we did not have information regarding any major concerns for the adolescent’s current well visit or past or follow-up medical or mental health diagnostic information. Thus, it is difficult to place these results in the broader context of an adolescent’s health care and developmental trajectory. However, the consistency between our results and the broader literature is indicative of clear associations between sleep, depression, and suicidality detected in pediatric health care settings and highlight the importance of provider attention to sleep disturbances in this population. Moreover, and consistent with previous meta-analyses,²⁶ the association between sleep and suicidality remained significant when accounting for total and prorated threshold depressive symptoms, demonstrating that sleep disturbances account for variance above and beyond depression. Notably, nearly 60% of youth who endorsed suicidality did not screen positive for threshold depressive symptoms, indicating that these constructs should continue to be carefully assessed and screened for independently. Future studies would benefit from expanding on our investigation to include more in-depth analysis of adolescents’ health care histories as well as other proximal risk factors (eg, family history and stressful life events) to better understand the relative influence of sleep within the broader context of adolescent development. Moreover, future work examining how the onset of the COVID-19 pandemic in March 2020 may impact the current findings would be worthwhile, especially given other work demonstrating significant increases in the prevalence of adolescent depression and suicidality⁵³ as well as changes in sleep patterns in adolescents after the pandemic.⁵⁴

Lastly, because of the concurrent nature of data collection, we cannot make directional, causal, or mechanistic claims about the associations between sleep disturbances, depression, and suicidality in this population. The PHQ-9-M items ask about the frequency of sleep and other depressive symptoms within the past 2 weeks, but also lifetime and past month indicators of suicidality. Therefore, though prior literature suggests that sleep disturbances often precede and serve to maintain depressive symptoms and STBs through biopsychosocial mechanisms,^17,18 it is possible that the reverse association may be true or that these variables interact with one another in a bidirectional manner. Given these limitations, it will be important for future studies to examine longitudinal associations between these constructs in PPC. Specifically, administration and analysis of finer-grained and more frequent assessments of sleep, depressive symptoms, and suicidality in primary care settings could further elucidate how and when sleep disturbances emerge within a child’s treatment trajectory in relation to depression and suicidality, as well as other proximal and distal risk and protective factors. Regardless of temporality, researchers may also consider implementing brief interventions for sleep disturbances in PPC to assess the efficacy of treating sleep to improve depressive symptoms and suicidality in youth. This approach is especially warranted as a sleep intervention may be more acceptable for adolescents and families who decline direct mental health treatments such as cognitive-behavioral therapy and/or have concerns about disclosure and stigma of internalizing symptoms.⁵⁰

In conclusion, this study demonstrated high prevalence of sleep disturbances (nearly 40%) in youth, with significant sociodemographic differences in endorsement. We found large and medium associations between sleep disturbances and concurrent depressive symptoms and suicidality in youth attending PPC well visits. Considering the high likelihood for co-occurring depression and suicidality in adolescents who endorse sleep disturbances, PPC providers should consider implementing standardized protocols for screening, documenting, and managing sleep disturbances.

CRediT authorship contribution statement

Giana I. Teresi: Writing – review & editing, Writing – original draft, Visualization, Formal analysis, Conceptualization. Molly Davis: Writing – review & editing, Project administration, Data curation. Ariel A. Williamson: Writing – review & editing. Jami F. Young: Writing – review & editing, Conceptualization. John A. Merranko: Writing – review & editing, Formal analysis. Tina R. Goldstein: Writing – review & editing, Supervision, Resources, Investigation, Conceptualization.