Suicidal Ideation in Pregnancy and Postpartum Does Not Behave Like a Single Continuous Risk State
Suicidal ideation during the perinatal period is often treated as if it rises and falls on one predictable curve. The UPWARD-S analysis shows a more complicated pattern. Across pregnancy and into the first six months after delivery, suicidal ideation separated into three recognizable trajectories: a stable low group, a group with more frequent endorsement, and a group whose endorsement changed over time.
The sample was small, but the signal was clinically important. In a cohort of 38 women followed through 26 time points, the pattern of self-harm or suicidal thoughts did not blur into one uniform experience. Some participants remained low throughout, some stayed consistently higher, and others shifted in either direction across the perinatal window. In practice, that means a single negative screen at one visit cannot be assumed to represent the full course of risk.
One of the sharpest findings was the difference between the two screening tools used in routine care. The Edinburgh Postnatal Depression Scale and the PHQ-9 did not classify participants in the same way. During pregnancy, 31.6% were placed in the stable low group by the EPDS, while 60.5% were classified that way by the PHQ-9. The divergence is not just statistical; it reflects two instruments asking related but not identical questions, over different time frames, and with different emphasis on self-harm versus suicidal thoughts.
What the UPWARD-S Cohort Captured Over Time
Participants came from a larger study designed around depressive relapse during pregnancy. Eligibility required a history of major depressive disorder, and the study intentionally enrolled individuals who endorsed suicidal ideation during pregnancy, were at least 16 weeks pregnant, and had major depressive disorder, current dysthymia, or current subsyndromal depression. Bipolar disorder, psychotic disorder, active mania, and active substance use disorder within the past 12 months were exclusion criteria.
After enrollment, each participant completed a clinical interview to confirm depressive history and a baseline assessment within one week. The Columbia Suicide Severity Rating Scale was used at baseline to assess current suicidal ideation and suicidal behaviors. From there, self-report questionnaires were collected every two weeks, generating a longitudinal view of how endorsement changed across pregnancy and after birth.
The pregnancy trajectories were described as stable low endorsement, declining endorsement, and stable more frequent endorsement. Postpartum, the pattern shifted slightly to stable low endorsement, increasing frequency of endorsement, and stable more frequent endorsement. That postpartum increase in one trajectory is clinically meaningful. Delivery does not automatically reduce risk, and for some women the months after birth may mark a period of worsening distress rather than relief.
The strongest bedside implication is straightforward: perinatal suicide risk assessment has to be repeated, not assumed. A woman who appears stable in late pregnancy may not remain stable after delivery, and another who begins with intermittent ideation may later settle into a lower-risk pattern. The trajectory matters because it informs how often clinicians should ask, how directly they should ask, and how urgently they should reassess if answers change.
EPDS and PHQ-9 Are Not Interchangeable for Suicidality Screening
The study’s most practically relevant comparison may be the mismatch between the EPDS and PHQ-9. Both instruments contain a single item used in practice to flag suicidal or self-harm ideation, but the wording and recall window differ. The EPDS item asks about thoughts of self-harm over the past week. The PHQ-9 item asks about thoughts of being better off dead or self-harm over the past two weeks.
Those differences matter because the tools are not sampling exactly the same clinical phenomenon. The EPDS item is broader in one sense, because it includes self-harm, while the PHQ-9 emphasizes passive death wish alongside self-harm. Their different recall windows also change what gets counted. A fleeting spike in distress may register on one tool and disappear on the other, while a more persistent pattern may appear similar across both.
In this cohort, the classification split suggests measurement itself can shape the apparent prevalence of suicidal ideation trajectories. In other words, the screening instrument is not just a passive observer. It influences the clinical picture by deciding which experiences are captured and how frequently they are counted.
Why Perinatal Suicide Risk Needs a Different Framework
Perinatal suicidal ideation is often discussed alongside depression, but the two do not always move together. Depression can be episodic, while suicidal thoughts may be more persistent or may fluctuate in a different rhythm. That difference has consequences for screening programs built around depression tools alone. If suicidal ideation does not map neatly onto mood symptoms, then depression severity scores can miss a separate and potentially unstable risk process.
The findings also argue against treating pregnancy and postpartum as one uninterrupted block of risk. Hormonal change, sleep disruption, role transition, relationship stress, and prior psychiatric vulnerability all interact across this interval. After delivery, those stressors can intensify rather than fade. For patients with prior depression, the postpartum period may expose a different vulnerability profile than pregnancy itself.
A single paragraph in the findings captures the clinical tension well: the stable low and stable more frequent groups generally stayed in the same trajectory from pregnancy into postpartum, yet other participants shifted. That split suggests some women show continuity of risk, while others undergo a change in pattern that routine screening may miss if it is too infrequent.
Limits of the Data and the Need for Better Real-World Measurement
The study’s limitations are substantial and should frame interpretation. The sample was only 38 women, and the analysis relied entirely on the EPDS and PHQ-9 for suicidal ideation measurement. Neither tool was designed to provide a full phenomenologic assessment of suicidality, and item-based screening cannot substitute for a clinical interview when risk is suspected.
The time windows embedded in the instruments also complicate interpretation. A one-week window and a two-week window do not produce identical estimates, especially when symptoms may shift rapidly in pregnancy or after birth. The EPDS item also captures self-harm thoughts more broadly, which makes it less specific for suicidal ideation alone.
Larger and more diverse cohorts are needed to determine whether these same trajectory groups appear across other populations and care settings. Ecological momentary assessment could add a more granular view of day-to-day change, especially in the postpartum period when sleep loss and mood instability may evolve quickly. Repeated real-time assessment may also identify the moments when ideation first emerges, rather than documenting it only after it has already become visible on a routine screen.
For now, the practical lesson is narrow but important. Repeated screening, direct questioning, and careful interpretation of the exact item used are essential when clinicians are trying to detect suicidal ideation in pregnancy and after delivery. The choice between EPDS and PHQ-9 can change who gets classified as low risk, and that discrepancy has direct consequences for follow-up and safety planning.