Endorsement of a single-item measure of sleep disturbance during pregnancy and risk for postpartum depression: a retrospective cohort study

We conducted a retrospective cohort study in an integrated healthcare system that provides health insurance and clinical services to more than 640,000 individuals in the metropolitan Denver area and northern and southern CO communities. The study cohort included perinatal patients between 2012 and 2017 who had a live birth and who had completed PHQ-9 documented in the electronic health record between their pregnancy onset date and 8 weeks postpartum (Fig. 1). For participants with multiple pregnancies in the dataset, only the most recent was included. Outlier participants with four or more PHQ-9 administrations within a trimester were excluded. For participants with two to three PHQ-9 administrations within a trimester, the first observation was used. Finally, the analyzed sample included those with a PHQ-9 total score < 10 at any trimester, indicating minimal to mild depressive symptom severity. This study was approved by the institutional review board in the integrated healthcare system.

All the data were extracted from the Virtual Data Warehouse (VDW), which contains electronic health record data (ICD-9/10 diagnosis codes, procedure codes, participant demographics, and participant-reported outcome data including PHQ-9) and insurance claims data.

The PHQ-9 is a nine-item self-report measure of depressive symptom severity (Kroenke et al. 2001b). The sleep disturbance item (#3) assesses both insomnia and hypersomnia (“trouble falling asleep or staying asleep, or sleeping too much”). The fatigue item (#4) assesses feeling tired or having low energy. The appetite disturbance item (#5) assesses both poor appetite and overeating. Respondents indicate how often they have been bothered by symptoms, with response options including “not at all (0),” “several days (1),” “more than half the days (2),” or “nearly every day (3).” Endorsement of these items was defined as any score above “not at all (0).” Total scores range from 0 to 27, and scores of 10 or above are typically used to identify possible depression. In the present study, non-depressed pregnant participants were defined as those with PHQ scores < 10. In this health system, PHQ-9 is administered during the first and second trimesters of pregnancy and at 5–8 weeks postpartum.

Gestational length (based on ultrasound dating) and other participant characteristics (age at the first trimester visit, Hispanic ethnicity, race, and parity) were also extracted from the VDW for statistical analyses.

Participant characteristics were summarized by descriptive statistics. Frequency distribution and percentage were used to summarize categorical variables, and mean with standard deviation (SD) was used to summarize continuous variables.

First, to investigate prospective associations between prenatal endorsement of sleep disturbance and elevated postpartum depressive symptoms (PHQ-9 ≥ 10), univariate and multivariate logistic regression models were used, with separate models for each pregnancy trimester. Covariates with documented associations with postpartum depression were selected for inclusion in the models (Guintivano et al. 2018; Hassdenteufel et al. 2021). The same analytic approach was used for investigating prospective associations between prenatal endorsement of fatigue and elevated postpartum depressive symptoms, and prospective associations between prenatal endorsement of appetite disturbance and elevated postpartum depressive symptoms. Because the predictor variable (i.e., sleep disturbance, fatigue, or appetite disturbance) was included in the outcome variable (PHQ-9 total score ≥ 10) albeit at different timepoints, we conducted sensitivity analyses that omitted this variable from the outcome variable and used a PHQ-9 cutoff that was pro-rated for eight items (i.e., PHQ-9 ≥ 8.89).

Second, to aid in the identification of those most at risk of experiencing sleep disturbance during pregnancy, we examined participant characteristics that correlated with endorsement of sleep disturbance. Specifically, two-sample t-tests were used for continuous variables and chi-squared test for categorical variables. Statistical significance was declared at p < 0.05, and all the statistical analyses were done by the statistical computing software R.

A total of 3619 participants were eligible for analysis (see Fig. 1 for sample selection details). On average, the participants were 31 years old, and the majority was White, non-Hispanic, and multiparous (see Table 1).

First, we investigated the prospective associations between prenatal endorsement of sleep disturbance and elevated postpartum depressive symptoms. Among pregnant participants who were not depressed (i.e., PHQ-9 total < 10), endorsement of sleep disturbance at any trimester was associated with increased risk of elevated postpartum depression symptoms in adjusted analyses. Endorsement of sleep disturbance in the second and third trimesters was associated with over three-fold higher odds of elevated postpartum depressive symptoms, adjusting for age, race, ethnicity, and parity (second trimester sleep disturbance aOR 3.74, 95% CI 1.47–11.49, p = 0.01; third trimester sleep disturbance aOR 3.43, 95% CI 1.88–6.78, p < 0.001), followed by nearly twofold higher odds for endorsement of sleep disturbance in the first trimester (aOR 1.90, 95% CI 1.17–3.13, p = 0.009).

Secondary analyses examined the prospective association of other somatic symptoms of pregnancy with postpartum depression. In the first and third trimesters, both fatigue and appetite disturbance were associated with increased odds of elevated postpartum depressive symptoms in adjusted analyses. For example, endorsement of fatigue in the first trimester was associated with over three-fold higher odds of elevated postpartum depressive symptoms in the adjusted analyses (OR 3.44, 95% CI 1.27–14.11, p = 0.038). Endorsement of appetite disturbance in the third trimester was associated with over two-fold higher odds of elevated postpartum depression symptoms in the adjusted analyses (OR 2.15, 95% CI 1.23–3.65, p = 0.006).

The results did not substantively change in sensitivity analyses, suggesting that the associations between prenatal sleep disturbance and elevated postpartum depressive symptoms were not driven by the association between prenatal sleep disturbance and postpartum sleep disturbance. The same was true for the adjusted analyses of appetite disturbance. In contrast, in the adjusted analyses, fatigue no longer had a statistically significant association with elevated postpartum depressive symptoms when omitting the fatigue item (see Supplemental Table 1).

Endorsement of sleep disturbance varied significantly by race during the first and second trimesters but not the third trimester. For example, 48.4% of Asian participants reported sleep disturbance in the first trimester compared to 65.1% of Black or African American participants. During the first two trimesters, endorsement of sleep disturbance was highest among the Black or African American participants (ranging from 61.8 to 65.1%) and lowest among the Asian participants (ranging from 33.3 to 56.2%). Endorsement of sleep disturbance also varied significantly by parity during the first and third trimesters, such that sleep disturbance was less common among nulliparous participants.

Our findings show that pregnant participants without clinically significant depression symptoms who reported “trouble falling or staying asleep or sleeping too much” on the  PHQ-9 at any trimester were at increased risk for elevated depressive symptoms at 6 weeks postpartum. Endorsement of fatigue and appetite disturbance, which also are somatic symptoms of both pregnancy and depression, was associated with increased risk of elevated postpartum depressive symptoms. However, endorsement of sleep disturbance in the second trimester was unique in predicting postpartum depression.

These findings are consistent with previous research reporting similar associations using longer, multiple-item assessments of prenatal insomnia (Tomfohr et al. 2015). For example, non-depressed pregnant participants who reported nightly difficulties falling asleep during the third trimester were nearly four times as likely to experience depression by 6 months postpartum compared to those who reported no or occasional difficulty falling asleep (24% vs 7%, respectively) (Suri et al. 2017). Findings from the current paper add to the evidence base by showing that even a single-item measure of sleep disturbance has utility in prospectively identifying risk for postpartum depression.

Our findings are also consistent with previous research showing that pregnant women who are Black are more likely to report short or very short sleep (≤ 6 h per night) relative to those who are non-Hispanic White (Amyx et al. 2017). Racial disparities in endorsement of sleep disturbance are likely a proxy for other unmeasured factors; for example, research among perinatal samples pinpoints racism and discrimination as a driver of poor sleep in this population (Gaston et al. 2020; Slopen et al. 2016). This finding is particularly notable in the context of disparities in access to and/or quality of perinatal care that may exacerbate the burden of prenatal sleep disturbance among Black or African American people (Vedam et al. 2019; McLemore et al. 2018).

Our findings suggest that item 3 of the PHQ-9 may be an efficient tool for identifying a risk factor for postpartum depression that may otherwise be under-recognized or even dismissed as a normative, harmless symptom of pregnancy (Felder et al. 2019). Clinically, this suggests that providers could examine the endorsement of sleep disturbance item on PHQ-9 to identify risk for future depression, in addition to examining total scores to identify current depression. This is akin to standard clinical practice that necessitates following up on endorsement of the suicidal ideation item of PHQ-9 even when the PHQ-9 total score is below the conventional cutoff of 10.

Our findings suggest that clinicians should be particularly attentive to endorsement of sleep disturbance in the second trimester as it may be a “canary in the coal mine” that is unique relative to other somatic symptoms of pregnancy in portending risk for depression. Additionally, unlike fatigue and appetite disturbance, sleep disturbances, particularly insomnia, have evidence-based treatments. A randomized controlled trial found that 64% of pregnant participants who received in-person cognitive behavioral therapy for insomnia (CBT-I) experienced symptom remission at post-intervention (Manber et al. 2019). There is also evidence that digital adaptations of CBT-I are effective during pregnancy (Felder et al. 2020; Kalmbach et al. 2020). Of particular relevance is the increasing evidence that CBT-I may prevent depression among adults with insomnia, older adults, and perinatal people (Felder et al. 2021; Cheng et al. 2019; Irwin et al. 2022). If left untreated, prenatal sleep disturbance can persist into the postpartum period (Tomfohr et al. 2015). The US Preventive Services Task Force recommends that pregnant people at risk for depression receive preventive counseling interventions (U. S. Preventive Services Task Force et al. 2019). Taken together, these findings underscore the importance of identifying and treating prenatal sleep disturbances.

The results of the present study should be considered in the context of several limitations. First, the single PHQ-9 sleep disturbance item assesses experiences of insomnia and hypersomnia; thus, we are unable to examine whether there are differences in the degree of risk conferred between these indicators. Second, data on depressive symptoms was only available through 6 weeks postpartum. The onset and course of postpartum depressive symptoms are highly variable and may emerge or persist up to 12 months following childbirth (American Psychiatric Association 2013). Although we were unable to investigate whether endorsement of sleep disturbance at 6 weeks postpartum predicted risk for elevated depressive symptoms later in the postpartum period, previous research suggests that mothers who reported having problems with sleep at 6 weeks postpartum were more likely to have elevated depressive symptoms at 6 or 12 weeks postpartum (Stremler et al. 2020). Relatedly, the focus of the current study was on elevated depressive symptoms (i.e., PHQ-9 > 10); although scores in this range are suggestive of depression, it is possible that the present results may not generalize to those with clinical diagnoses of depression. It is notable, however, that even subclinical depressive symptoms exert a substantial impact on maternal and child functioning (Meaney 2018). Third, although the present study is strengthened by its inclusion of a large sample of individuals drawn from a managed healthcare organization, the racial and ethnic background is not representative of the birthing population in the USA. For example, the Center for Disease Control 2018 Natality Data indicate that 52% of births were to a non-Hispanic White mother compared to the 69% in the current sample. Fourth, we were unable to examine whether prenatal sleep disturbance varied across other potentially important clinical characteristics such as history of depression or family mental health history. Finally, we acknowledge that the risk for depressive symptoms conferred by sleep disturbance was examined using a sleep item on a measure that also contributed to the total depressive symptom score derived from that same measure (i.e., item 3 on PHQ-9 assesses sleep disturbance and is also part of the total summed score on PHQ-9). However, this approach may align with the “real-world” conditions of busy obstetrics practices that may not have the time or providers available to administer and score separate measures of depression and sleep disturbance.