Depression occurring in the year following childbirth remains a serious public health concern affecting 10–15% of women in the USA (O'Hara & McCabe, Reference O'Hara and McCabe2013; Shorey et al., Reference Shorey, Chee, Ng, Chan, San Tam and Chong2018). Evidence suggests that the burden of postpartum depression may be especially high among adolescent and young adult women, Black women, and women living in under-resourced environments (Goyal, Gay, & Lee, Reference Goyal, Gay and Lee2010; Pooler, Perry, & Ghandour, Reference Pooler, Perry and Ghandour2013), likely reflecting higher levels of concomitant stress, financial strain, and inequities in mental health access and care. Depression in the postpartum period often goes undetected and untreated (Pearlstein, Howard, Salisbury, & Zlotnick, Reference Pearlstein, Howard, Salisbury and Zlotnick2009) and is associated with long-term adverse effects for the mother, child, and her family (Letourneau et al., Reference Letourneau, Dennis, Benzies, Duffett-Leger, Stewart, Tryphonopoulos and Watson2012; Weissman et al., Reference Weissman, Wickramaratne, Nomura, Warner, Pilowsky and Verdeli2006) even when symptoms do not reach diagnostic threshold (Meaney, Reference Meaney2018). These multiple concerns point to an urgent need for the early identification of women who would benefit from targeted, preventive interventions.

Across the lifespan, a history of depression is known to be a strong risk factor for subsequent depression (Rudolph, Flynn, Abaied, Groot, & Thompson, Reference Rudolph, Flynn, Abaied, Groot and Thompson2009). Around the time of childbirth, short-term prospective studies have shown high levels of continuity in depression from pregnancy to the postpartum period (Davé, Petersen, Sherr, & Nazareth, Reference Davé, Petersen, Sherr and Nazareth2010; Eastwood et al., Reference Eastwood, Ogbo, Hendry, Noble, Page and Group2017; Guintivano et al., Reference Guintivano, Sullivan, Stuebe, Penders, Thorp, Rubinow and Meltzer-Brody2018) that have led to the broader conceptualization of ‘peripartum depression’ encompassing symptoms during and/or after pregnancy (American Psychiatric Association, 2013). There is also consistent evidence that a history of depression at any time prior to pregnancy is associated with increased risk for peripartum depression (Davé et al., Reference Davé, Petersen, Sherr and Nazareth2010; Dennis, Heaman, & Vigod, Reference Dennis, Heaman and Vigod2012; Guintivano et al., Reference Guintivano, Sullivan, Stuebe, Penders, Thorp, Rubinow and Meltzer-Brody2018; Räisänen et al., Reference Räisänen, Lehto, Nielsen, Gissler, Kramer and Heinonen2014; Silverman et al., Reference Silverman, Reichenberg, Savitz, Cnattingius, Lichtenstein, Hultman and Sandin2017). However, the vast majority of these data come from retrospective self-reports, which may be subject to recall bias in ways that are congruent with mood (Eich & Macaulay, Reference Eich and Macaulay2000; Walker, Skowronski, Gibbons, Vogl, & Thompson, Reference Walker, Skowronski, Gibbons, Vogl and Thompson2003), and which typically lack clinical and temporal specificity (Gjerdingen & Yawn, Reference Gjerdingen and Yawn2007; Moffitt et al., Reference Moffitt, Caspi, Taylor, Kokaua, Milne, Polanczyk and Poulton2010). In addition, extant research often reflects limited sample diversity in terms of age, race, and income level. Adolescent mothers, for example, are frequently excluded from studies of peripartum depression (e.g. Segre, O'Hara, Arndt, & Stuart, Reference Segre, O'Hara, Arndt and Stuart2007; Smith-Nielsen, Matthey, Lange, & Væver, Reference Smith-Nielsen, Matthey, Lange and Væver2018) and in the USA, health insurance-based registry studies are unlikely to represent all sectors of the population. In some cases, research excludes high-risk women covered on Medicaid (Dietz et al., Reference Dietz, Williams, Callaghan, Bachman, Whitlock and Hornbrook2007). Similarly, despite significant racial-ethnic disparities in peripartum outcomes, Black women and other women of color are often under-represented in studies investigating the prevalence of depression around the time of childbirth (Gaynes et al., Reference Gaynes, Gavin, Meltzer-Brody, Lohr, Swinson, Gartlehner and Miller2005). The limited representativeness of prior samples has made it difficult to estimate the burden of depression history on peripartum depression, particularly for those who may be most at risk. Furthermore, these and other studies have often relied on binary indicators such as a depression diagnosis, a psychiatric referral, or prescribed medication at an unspecified time (Vliegen, Casalin, & Luyten, Reference Vliegen, Casalin and Luyten2014). Yet, the severity of depression varies according to age of onset (Hölzel, Härter, Reese, & Kriston, Reference Hölzel, Härter, Reese and Kriston2011), and typically follows a remitting, episodic course (Kennedy, Abbott, & Paykel, Reference Kennedy, Abbott and Paykel2004; Mueller et al., Reference Mueller, Leon, Keller, Solomon, Endicott, Coryell and Maser1999) that can itself be prognostic (Brodaty, Luscombe, Peisah, Anstey, & Andrews, Reference Brodaty, Luscombe, Peisah, Anstey and Andrews2001). Therefore, developmental patterns of depression across the lifespan may be a critical determinant of peripartum adjustment that could also elucidate optimal timing of screening, treatment planning, and follow-up. Finally, studies have shown that depression in adolescence, even at subthreshold levels, is associated with an increased risk for depression in adulthood (Fergusson, Horwood, Ridder, & Beautrais, Reference Fergusson, Horwood, Ridder and Beautrais2005; Johnson, Cohen, & Kasen, Reference Johnson, Cohen and Kasen2009). Given that the incidence of depression increases dramatically between adolescence and early adulthood, prospective studies that span this age range contribute particularly important information about future depression risk (Pine, Cohen, Brook, Gurley, & Ma, Reference Pine, Cohen, Brook, Gurley and Ma1998).

Only a few longitudinal studies have examined preconception risk for depression during and after pregnancy. In one of two landmark studies, the 20-year Victorian Intergenerational Health Cohort Study (VIHCS, Spry et al., Reference Spry, Olsson, Hearps, Aarsman, Carlin, Howard and Brown2020) followed an adolescent cohort to identify pregnancies among participants when they were 29–36 years old. Among 384 women with 564 pregnancies, Patton et al. (Reference Patton, Romaniuk, Spry, Coffey, Olsson, Doyle and Brown2015) reported that mental health problems (i.e. mixed depression/anxiety) occurring during adolescence and young adulthood conferred an 8.4-fold risk for peripartum depression (assessed across late pregnancy and at 2 and 12 months postpartum) relative to no history. Mental health problems in young adulthood only were associated with a 4.6-fold increased risk, whereas adolescent-limited mental health problems were unrelated to peripartum depression risk. These results were replicated in a second longitudinal cohort study, the 36-year Australian Temperament Project (ATP) Generation 3 (Edwards et al., Reference Edwards, Hawkins, Letcher, Little, Macdonald, Oberklaid and Sanson2013) in which 393 women reported on 686 pregnancies (Thomson et al., Reference Thomson, Romaniuk, Greenwood, Letcher, Spry, Macdonald and Hutchinson2021). Subsequent analyses of the VIHCS data also showed that prenatal depression explained only a small part of the association between preconception mental health problems and symptoms of postpartum depression (Spry et al., Reference Spry, Moreno-Betancur, Middleton, Howard, Brown, Molyneaux and Thomson2021). Taken together, these findings highlight the importance of pregnancy psychopathology for peripartum depression and suggest that timing (i.e. recency) is an important dimension of risk. However, it is also possible that earlier developing symptoms have cumulative effects that only become evident with increasing age. Some methodological considerations include the focus on mixed depression/anxiety rather than specific depression history, changes in measurement from adolescence to adulthood that raise questions about developmental timing effects, and the inclusion of multiparous women for whom the preconception and prior peripartum periods may coincide.

Two shorter-term prospective studies examined continuity in depression symptoms from preconception to the postpartum period. In a Singapore-based study, participants comprised 18- to 45-year-old women (primarily employed, married, and highly educated) who were planning a pregnancy (Kee et al., Reference Kee, Ponmudi, Phua, Rifkin-Graboi, Chong, Tan and Meaney2021). Depression symptoms were measured repeatedly across three 3-monthly intervals prior to conception, during the three trimesters of pregnancy and at 3 months postpartum. Results showed moderate within-person stability in depression severity from preconception to postpartum. However, the limited focus on ‘pregnancy planners’ may have introduced selection bias related to multiple psychosocial variables (e.g. life stress) with relevance for depression, and so the generalizability of the results to the broader population is unclear. Second, a US-based longitudinal study of 147 primiparous adolescent mothers (86% Black) drawn from the current sample examined within-person change in depression severity across two time points: a year prior to pregnancy and within the first 6 months postpartum (Hipwell, Murray, Xiong, Stepp, & Keenan, Reference Hipwell, Murray, Xiong, Stepp and Keenan2016). Using the same measure at both assessments, the results revealed no overall change in self-reported depression across this period, indicating high levels of continuity. However, this analysis was limited to a single pre-pregnancy assessment and, like the prior study, did not consider the longitudinal course or cumulative burden of depression that is associated with future depression severity in non-childbearing samples (Brodaty et al., Reference Brodaty, Luscombe, Peisah, Anstey and Andrews2001).

In summary, convergent evidence has shown substantial continuity in depression severity from pregnancy to postpartum and has also highlighted pre-pregnancy depression as a major risk factor for depression occurring during and after pregnancy. However, rigorous studies that include diverse and representative samples, followed prospectively across the lifespan are lacking, and so little is known about developmental patterns of preconception depression (e.g. worsening symptoms, cumulative effects) that could aid detection of vulnerable individuals and inform the timing of preventive interventions. The current study addresses this gap by leveraging data from a population-based, racially and socioeconomically diverse sample of girls who have completed annual assessments for 20 consecutive years from childhood to young adulthood. Utilizing repeated measures of depression severity across an extended developmental period encompassing preconception, pregnancy, and postpartum provides an unprecedented opportunity to examine change, chronicity, and cumulative effects of depression in first-time mothers, including adolescent mothers. Based on prior work, we hypothesized that there would be continuity in depression from pregnancy to the postpartum period, but that preconception depression would explain unique variance in postpartum depression risk. We also hypothesized that increasing depression symptom severity in the period prior to pregnancy, and a greater lifetime burden or accumulation of preconception depression, would predict higher levels of depression in the postpartum period.

Results

In the final sample of 414 first-time mothers, 408 (98.6%) completed the EPDS and 399 (96%) contributed at least three time points of data prior to pregnancy (the minimum number needed to estimate depression slopes) and so were included in the present analysis.

Descriptive statistics and bivariate correlations among sociodemographic and study variables are presented in Table 1. Participants ranged in age from 13 to 25 years at the time of delivery (mean = 19.20, s.d. = 2.16). Most women identified as Black (81.2%) or Black and another race (multiracial: 3%) and 15.8% identified as White. During pregnancy, 27.8% of participants lived in a household receiving public assistance. Postpartum depression scores, assessed on average at 4.24 months following delivery, ranged from 0 to 23 (M = 4.42, s.d. = 4.68), with 9% of this sample scoring in the clinically significant range (score ⩾12) and 15.9% scoring in the screening range (score ⩾10). Depression in the postpartum period was significantly correlated with history of depression severity across most previous time points (significant up to 11 years prior to pregnancy) and was strongly associated with depression during pregnancy and in the year before pregnancy. None of the sociodemographic variables was correlated with depression in the postpartum period.

Table 1. Descriptive statistics and correlations among demographic variables and depression scores across time

Dep, depression severity; pre-preg, pre-pregnancy.

^a Race coded as 1 = Black American (includes multiracial Black and another race) v. 0 = White American.

^b Public assistance coded as 0 = no v. 1 = yes. Postpartum depression was measured by the Edinburgh Postnatal Depression Scale, whereas depression during pregnancy and all pre-pregnancy time points was measured by the Adolescent/Adult Self-Report Inventory-4 major depressive disorder scale. All available pre-pregnancy time points were used; thus, depending on age of childbirth, some participants had more years of depression data available before pregnancy than other participants.

*p < 0.05, **p < 0.01.

Overall, there was a significant decrease in depression severity over time prior to childbirth, with an estimated mean slope of B = −0.20, p < 0.01 (95% CI −0.27 to −0.12). However, there was substantial variability in slopes (stable, increasing, decreasing) across individuals, with negative and positive slope gradients ranging from B = −3.10 to 3.20 respectively.

Predicting depression severity in the postpartum period

Results from the regression models predicting postpartum depression severity appear in Table 2. Adjusting for age at delivery, race, and low-income status, depression severity during pregnancy significantly predicted depression in the postpartum period (Model 1): a three-point increase in prenatal depression severity was associated with more than a one-point increase in severity of postpartum depression (B = 1.17). Without considering individual differences in depression severity during pregnancy, the slope of the preconception depression trajectory was not associated with depression in the postpartum period (Model 2). When Model 2 was re-run without adjusting for age at childbirth or receipt of public assistance, the results were unchanged, suggesting that these intermediary variables were not operating as mediators. When simultaneously considering the association of depression during pregnancy and the slope of preconception depression on depression in the postpartum period (Model 3), prenatal depression remained positively associated with depression in the postpartum period (B = 0.52, s.e. = 0.08, p < 0.001), whereas the slope of preconception depression trajectory was negatively associated with postpartum depression severity (B = −1.44, s.e. = 0.55, p < 0.01). After adjusting for differences in depression during pregnancy, a negative depression trajectory slope reflects greater levels of cumulative depression leading up to pregnancy (i.e. area under the curve of preconception depression trajectory). Consequently, these results suggest that beyond continuity in depression from pregnancy to postpartum, the cumulative history of depression explains additional unique variance in depression in the postpartum period such that an average three-point increase in depression severity per year over 8 years (the median preconception observation time in our sample) is associated with an estimated 1.08 increase in expected EPDS score (s.e. = 0.41, p < 0.01). To explore the possible effects of missingness, we refit the model using only data from women with complete data. In a sensitivity analysis, we compared inference from the reported model to this complete case model and found point estimates to be highly similar.

Table 2. Regression models predicting postpartum depression scores from history of depression

Note. Models are adjusted for race, prenatal public assistance, and slope of preconception depression. Significant effects are bolded for emphasis.

^a Race coded as 1 = Black American (includes multiracial Black and another race) v. 0 = White American.

^b Public assistance coded as 0 = no v. 1 = yes.

Regarding the role of demographic factors in the model, neither age at delivery nor public assistance predicted depression in the postpartum period after accounting for history of depression. Furthermore, adjusting for these variables did not significantly reduce estimates of depression continuities in Model 2. However, there was a main effect of race on depression in the postpartum period, with Black mothers reporting lower levels of depression than White mothers after adjusting for age, public assistance, depression during pregnancy, and slope of preconception depression trajectory (Table 2). Given the significant negative association between Black race and depression in the postpartum period, we further probed the nature of this association with respect to history of depression prior to childbirth. Black participants did not differ from White participants on depression severity reported during pregnancy (t = 1.10, p = 0.27) or in terms of slope of preconception depression trajectory (t = 0.04, p = 0.97).