Distinct Suicidal Ideation Paths Emerge Across Pregnancy and the Postpartum
Suicidal ideation during the perinatal period remains difficult to track in routine care, in part because clinicians often rely on a single item buried inside broader depression screens. In this UPWARD-S supplemental study, investigators followed 38 pregnant participants with a history of major depressive disorder through pregnancy and up to six months postpartum, using the suicidal ideation items from the Edinburgh Postnatal Depression Scale and the Patient Health Questionnaire-9. The central finding was not a single uniform course, but several distinct trajectories that behaved differently across time. Some participants showed persistently low endorsement, some remained persistently high, and a third group changed direction depending on the period being measured.
That pattern matters clinically because perinatal suicide risk is often treated as a binary screening result when it behaves more like a moving target.
How the Pregnancy Cohort Split Into Stable and Shifting Patterns
During pregnancy, group-based trajectory modeling identified three patterns for each screening item. Two of those groups were relatively stable: one with consistently low suicidal ideation endorsement and another with consistently high endorsement. The third pregnancy group was more dynamic, starting high and then declining over time. Across both the EPDS and PHQ-9 items, the overall structure looked broadly similar, but the proportion of participants assigned to each trajectory differed by measure.
That separation is clinically important because it suggests the same patient can look meaningfully different depending on which item is used to ask the question. In a perinatal setting, where symptoms of depression, anxiety, sleep disruption, and psychosocial strain can fluctuate quickly, a single screening time point may capture only a fragment of risk. A declining trajectory does not erase earlier danger, especially when suicidal thinking has already surfaced during pregnancy. It instead signals that the burden of risk may be changing shape rather than disappearing.
The study population itself narrows the interpretation. Participants were eligible for the supplemental study because they were pregnant and had a history of major depressive disorder, while those who endorsed suicidal ideation at baseline were ineligible for the parent UPWARD study. As a result, the cohort reflects a selected group already known to be vulnerable, which makes the observed persistence and movement in suicidal ideation especially relevant to real-world screening workflows.
Postpartum Patterns Shifted in the Opposite Direction
The postpartum results were not a simple continuation of the pregnancy signal. After delivery, two groups again showed relatively stable trajectories, one high and one low, but the third postpartum group began low and then increased. In practical terms, the period after childbirth did not simply repeat pregnancy patterns; it introduced a new subgroup whose suicidal ideation intensified over time.
That postpartum rise is clinically plausible. The weeks and months after birth bring abrupt sleep disruption, hormonal change, physical recovery, feeding stress, and often a widening gap between expected and lived experience. For patients with a prior mood disorder, those pressures can make risk emerge later rather than at the first postpartum contact. A screen that is reassuring early in the postpartum period may miss a later climb in suicidal thinking, particularly if clinicians assume that a negative response at one visit is durable.
The study does not claim that one screening instrument alone captures the full burden of risk. Instead, it shows that timing and item choice can change how many people are labeled as having persistent, improving, or worsening suicidal ideation. In a cohort this small, the exact proportions should not be generalized, but the directionality of the trajectories is difficult to ignore.
EPDS and PHQ-9 Were Similar in Shape, Not in Classification
Across pregnancy and postpartum, the suicidal ideation trajectories were broadly similar between the EPDS and PHQ-9. Even so, the classification of participants into each trajectory group differed by measure. In other words, the screening tools told a comparable story in broad strokes, but they did not place the same participants into the same risk buckets.
That modest agreement is a meaningful warning for clinicians who use these items interchangeably. The final question on the EPDS and item 9 on the PHQ-9 both serve as quick probes for suicidal thinking, yet they sit inside different instruments, use different wording, and may be interpreted differently by patients. A patient who appears low risk on one tool may not land in the same trajectory on the other. In a perinatal population with known psychiatric vulnerability, even modest discordance can affect referral decisions, follow-up intensity, and whether a clinician feels reassured after a single negative screen.
The mechanics of risk detection matter here. Suicidal ideation is not a fixed trait; it can rise and fall with mood state, stress load, and contextual changes between pregnancy and postpartum. A screening system that captures only one moment may miss the arc. A system that uses two commonly deployed tools may improve detection, but only if clinicians recognize that the instruments are not perfectly aligned.
What This Small Cohort Can and Cannot Support
The evidence comes from a supplemental study nested within UPWARD, and the sample size was small. That limits any attempt to estimate population-level prevalence or to define precise predictive thresholds. The study was not designed to establish which item is superior for identifying imminent self-harm risk, nor does it replace clinical assessment. The trajectories, however, provide a useful map of how suicidal ideation can behave across the perinatal window in women with prior major depressive disorder.
The most actionable signal is the persistence of distinct patterns rather than a single average course. Some patients remained low-risk by item endorsement, some stayed high, and some changed direction either during pregnancy or after delivery. Averages would have blurred those differences. The trajectory approach preserved them.
For perinatal care, that means screening should be treated as a repeated process, not a one-time administrative step. A low score early in pregnancy or after delivery should not end vigilance, and a rise later in the postpartum period may be just as important as a high score during gestation. The data also suggest that future larger studies should compare these instruments against each other more directly, with enough participants to determine how often each item flags the same patient and how often it diverges.
The study closes with a practical gap rather than a final answer: larger cohorts are needed to determine how perinatal suicidal ideation trajectories should shape screening frequency, clinical follow-up, and intervention planning.