Suicide is a leading cause of maternal mortality.1 Although being pregnant may actually protect against suicide,2 subgroups of women may be at elevated risk of suicide after or before delivery.3 Up to half of pregnant women referred for psychiatric care already have a history of self-harm,4 yet only 20% of women with suicidal thoughts share this information with a health care provider5 so detecting vulnerability is difficult, although women who complete suicide often have proximate medical visits before their deaths.6
Many depression screening tools for perinatal women ask about thoughts of self-harm, but the implications of a positive response are uncertain. Because mood screening has been endorsed by multiple medical organizations7–9 and enacted by state legislation,10 understanding the significance of patients' reporting suicidal ideation appears essential if we are to follow the clinical recommendation of universal screening. In the present study, we estimate the overall incidence of suicidal ideation in an unselected pregnant and postpartum population and examine its clinical correlates as well as the profiles of patients reporting these thoughts and their associated explanations to estimate the magnitude of suicide risk.
MATERIALS AND METHODS
In a large suburban health system with two obstetric services on separate campuses, the Edinburgh Postnatal Depression Scale was routinely administered during pregnancy and after delivery. The Edinburgh Postnatal Depression Scale was chosen for its established validity in both the antepartum and postpartum timeframes.11 The Edinburgh Postnatal Depression Scale is a 10-question, self-report instrument that does not provide a clinical diagnosis of a perinatal mood disorder but does identify patients at risk. Items on the scale correspond to various clinical depression symptoms such as guilt, sadness, anhedonia, and suicidal ideation. Responses are scored 0, 1, 2, or 3 according to increased severity of the particular symptom. Total score is determined by adding together the scores for each of the 10 items and higher scores indicate more depressive symptoms.
Screeners were provided validated versions of the Edinburgh Postnatal Depression Scale in multiple languages so that each patient could be assessed in her native language. The Edinburgh Postnatal Depression Scale was administered to patients at 24–28 weeks of gestation and 6 weeks postpartum. These screening times were based on guidelines of Edinburgh Postnatal Depression Scale authors12 and also to coincide with routine clinical visits for glucose tolerance testing and the postpartum checkup. Edinburgh Postnatal Depression Scale question 10 reads, “The thought of harming myself has occurred to me.” Answer choices to this question are “never,” “hardly ever,” “sometimes,” and “yes, quite often” and are assigned 0, 1, 2, and 3 points, respectively. The Edinburgh Postnatal Depression Scale questionnaires were designated “positive” for scores 12 or greater or for a response other than “never” to question 10 (independent of the total score) based on recommendations by the Edinburgh Postnatal Depression Scale authors.13
The institutional review board of NorthShore University HealthSystem approved this study. Scored screens were inputted into electronic health records using Epic 12. Deidentified Edinburgh Postnatal Depression Scale data collected between January 2003 and March 2011 and demographic, obstetric, neonatal, and mental health variables associated with those data were retrieved for study. Antepartum and postpartum Edinburgh Postnatal Depression Scale questionnaires were analyzed separately to account for the possibility that different variables would be relevant in each period. Because obstetric outcomes are only directly relevant as potential predictors of postpartum suicidal ideation, those variables were included for postpartum analyses only.
Group comparisons were conducted using SAS 9.2. Univariable assessments with P<.25 in logistic regression were entered as potential predictors in the multivariable models. Any two-way interactions with P<.1 were included for the next step in variable selection. Nonsignificant predictors were eliminated backward and the final model retained predictors, interactions, and confounders with P<.05. Multicolinearity diagnosis was checked based on variance inflation factor and log-linearity was tested for continuous predictors. Goodness of fit was assessed by the Hosmer and Lemeshow test. For the final multivariable models, the C-statistic was used to quantify the models' capacity in discriminating the presence or absence of suicidal ideation report.
Trained mental health professionals conducted structured phone interviews with all women with positive Edinburgh Postnatal Depression Scale scores within 48 hours of screening. An interpretative language phone line allowed for evaluation in each woman's primary language. Women's explanations pertaining to suicidal ideation were analyzed according to qualitative research methods.14 Only those interviews conducted after 2009 were included in this qualitative analysis because earlier interview documentation was handwritten and limited in content. Case codes were developed by mental health professionals, each with a minimum of 3 years' clinical experience in mood disorder detection and treatment. Unifying categories were created by the theoretical framework of Kleiman.15 Phone evaluations were classified by four raters (clinical psychologist, social worker, and trained research assistants), who independently coded and then exchanged cases, compared between-rater assessments, and resolved discrepancies. Interrater reliability greater than 95% was reached and ongoing crosschecking of interpretations by at least two raters per passage was performed, flagging discordant results for group review.
Over the study timeframe, 34,567 Edinburgh Postnatal Depression Scale questionnaires were processed. Although each patient is intended to be screened twice per pregnancy (antepartum and postpartum), some patients were not screened at all, others were screened only once (antepartum or postpartum), and others were screened in association with multiple pregnancies. For our analyses, each screened patient was only represented once in each of the antepartum and postpartum cohorts. In all cases, the first available antepartum and postpartum Edinburgh Postnatal Depression Scale result per patient was included, even if those screens were from different pregnancies. This resulted in 22,118 unique screens (13,724 antepartum and 8,394 postpartum; Fig. 1). Demographic characteristics of the study group included unpartnered relationship status (17%); Caucasian (55%), Hispanic (12%), African American (7%), or other race (26%); English-speaking (93%), Spanish-speaking (6%); and publicly insured (13%).
Suicidal ideation was indicated on 842 of the 22,118 Edinburgh Postnatal Depression Scale questionnaires (3.8%, 95% confidence interval [CI] 3.5–4.1%), with report rates similar across antepartum (4.1%, 95% CI 3.7–4.4%) and postpartum (3.4%, 95% CI 3.0–3.8%) questionnaires. The distribution of antepartum and postpartum scores (Fig. 2) includes the proportion of patients within each numeric score who reported thoughts of self-harm. Notably, of all positive Edinburgh Postnatal Depression Scale questionnaires (defined as a total score 12 or greater or report of thoughts of self-harm), 41% indicated some degree of suicidal ideation. Although a correlation with higher scores on the remaining nine Edinburgh Postnatal Depression Scale items suggests greater symptomatology in patients at risk for suicide (P<.001), a total of 497 women who reported thoughts of self-harm had total scores that fell below the positive cutoff score of 12.
In univariable analyses (Table 1), antepartum suicidal ideation was associated with younger age (mean age 30.9 compared with 31.9 years, P=.001), unpartnered relationship status (29.5% compared with 16.5%, P<.001), non-Caucasian race (62.1% compared with 43.8%, P<.001), speaking a non-English language (11.0% compared with 6.6%, P<.001), having public insurance (19.9% compared with 12.5%, P<.001), and having a preexisting mental health diagnosis (8.4% compared with 4.2%, P<.001). Report of postpartum suicidal ideation was similarly associated with unpartnered relationship status (25.0% compared with 17.5%, P<.01), non-Caucasian race (62.4% compared with 45.2%, P<.001), non-English language (12.4% compared with 7.7%, P<.01), public insurance (18.2% compared with 14.2%, P<.001), preexisting mental health diagnosis (12.0% compared with 5.8%, P<.001), and lower neonatal birth weight (mean 3,159 g compared with 3,282 g, P=.013; Table 1). For the antepartum, the final multivariable model (Table 2) retained relationship status, language, and relationship status by language interaction, race, sum of Edinburgh Postnatal Depression Scale items 1–9 (EPDS9), and EPDS92 (quadratic term). The association between relationship status and suicidal ideation was different between English and non-English speakers. For English speakers, partnered patients were less likely (odds ratio [OR] 0.75, 95% CI 0.575–0.987) to report thoughts of self-harm. For non-English speakers, partnered patients had 1.55 (95% CI 0.813–2.971) times the risk of reporting suicidal thoughts. For both antepartum and postpartum, EPDS9 and EPDS92 also were associated with self-harm report.
For postpartum, the final model (Table 2) included relationship status, language, relationship status by language interaction, race, EPDS9 and EPDS92, cesarean delivery, and severe perineal laceration (ie, involving the anal sphincter or rectum). Relationship status and language were significant confounders—partnered patients (OR 0.80, 95% CI 0.56–1.16) and English speakers (OR 0.78, 95% CI 0.44–1.38) were less likely to report suicidal ideation. We examined the specificity and sensitivity of the risk factors associated with suicidal ideation (Table 3). The predictor with the greatest sensitivity (antepartum 79.6, 95% CI 76.22–82.92 and postpartum 84.5, 95% CI 80.30–88.72) and specificity (antepartum 78.2, 95% CI 77.49–78.90 and postpartum 75.9, 95% CI 74.95–76.81) with respect to suicidal ideation was the sum of the other nine responses excluding question 10 (EPDS9).
Phone evaluations associated with 574 suicidal ideation–positive Edinburgh Postnatal Depression Scale questionnaires (371 antepartum and 203 postpartum) underwent qualitative assessment. Six coded categories best represented patient explanations of suicidal ideation report (Table 4). A majority of women revealed that they were, in fact, experiencing some degree of suicidal thought (low or high risk) at the time of their clinical assessment (330/574 [57.5%], 95% CI 53.5–61.5%); most reported that their thoughts of self-harm were experienced as passive or fleeting, and these women were coded as “low risk” for suicide. These same patients further characterized their feelings as escapist thoughts with no plan or intent toward self-harm. In contrast, the code “historical” was used for the 89 (15.5%, 95% CI 12.5–18.5%) women who indicated their reports were referring to suicidal thoughts occurring before pregnancy. These women typically described wanting to be “completely honest” in answering the question and did not feel right stating that they had “never” had thoughts of self-harm when, in fact, at some point in their lives, they had experienced such thoughts.
Misunderstanding Edinburgh Postnatal Depression Scale question 10 was also common (13.1%, 95% CI 10.3–15.8%) with women citing language barriers, although the Edinburgh Postnatal Depression Scale was offered in more than 20 languages and each non-English-speaking patients identified their preferred language for the screening instrument. It is speculated that limited literacy in a native or designated language may have been a factor in this group. Among others in this category, the English version of the Edinburgh Postnatal Depression Scale had been administered to reportedly fluent but nonnative English speakers who ultimately described being confused by the wording of the self-harm item. Still other women reported interpreting the question in the context of anxiety as meaning “I am fearful of accidental harm coming to me”—a phenomenon that has been reported elsewhere—or “the thought that I am indirectly harming myself through less-than-optimal health habits has occurred to me.”
A small proportion (3.1%, 95% CI 1.7–4.6%) of women reported answering question 10 positively in error as a result of simply being distracted or hurried; in cases where this explanation seemed credible to coders, these were coded as “mistakes.” In contrast, a substantial minority (10.8%, 95% CI 7.6–12.5%) denied suicidal ideation but could not provide any reasonable explanation for their initial documentation of thoughts of self-harm. These women were distinct from those in other categories in that respondents acknowledged understanding the question, did not claim to have answered in error, but either could not or would not provide any further details when questioned directly about their responses.
Finally, in a small number of patients representing 1.1% (95% CI 0.2–1.9%) of the sample of women reporting suicidal ideation and 0.07% (95% CI 0.03–0.16%) of all screens included in this subanalysis, clinical assessments uncovered high-risk, active suicidal ideation with a plan, intent, and access to means. Three of these women actually reported a suicide attempt to the evaluator within the current perinatal episode. The following case is illustrative of one such patient and the effect that Edinburgh Postnatal Depression Scale screening appeared to have on the clinical outcome:
Ms. X was a 19-year-old partnered, African American woman with a history of depression seen for her 6-week postpartum check. After reviewing the completed Edinburgh Postnatal Depression Scale that was positive for self-harm report, the obstetric resident documented the following: “I asked patient about thoughts of suicide and she replied that she took large amounts of Nyquil with ibuprofen and Vicodin two nights ago. I asked her if this was her first suicide attempt since her delivery and she said yes. I asked her if she was thinking of still hurting herself and she nodded yes. After this, I contacted social work and crisis support and walked the patient over to the emergency department.”
Absent the Edinburgh Postnatal Depression Scale alert and the direct physician questioning, it is entirely possible that this patient could have left the clinic with her recent suicide attempt undiscovered and her ongoing ideation undetected.
Other variables not available through the electronic health record were also collected from women during the phone assessments. A low rate of current mental health treatment engagement was noted among women reporting suicidal ideation. Of the 574 women included in this portion of the study, only 44 (7.7%, 95% CI 5.49–9.84%) were receiving care from a psychiatrist, 55 (10%, 95% CI 7.17–11.99%) were linked with a therapist, and 72 (13%, 95% CI 9.83–15.25%) were taking psychotropic medication. However, the majority of these women (521 [90.8%], 95% CI 88.4–93.13%) were not engaged in any form of mental health treatment.
The American College of Obstetricians and Gynecologists supports follow-up evaluation of screen-positive women despite insufficient evidence to make a firm recommendation for screening.16 Our data support universal screening as a means to identify women with mood disorders and also to uncover those with suicidal ideation. Using the Edinburgh Postnatal Depression Scale, we found significant rates of suicidal ideation among nearly 22,000 screens during (4.1%) and after (3.4%) pregnancy. We also identified subgroups that are more likely to report suicidal ideation, including unpartnered, non-Caucasian women and patients with a prior mental health diagnosis. The interaction of language and culture may also affect patient interpretation of the Edinburgh Postnatal Depression Scale even when native language screening is used.
Because the majority of women with suicidal ideation were not at high risk for attempting suicide, it is tempting to conclude that the self-harm question has limited clinical utility. However, the consequences of undetected suicidal ideation suggest that false-positive screens can be tolerated to maximize sensitivity.5 The American College of Obstetricians and Gynecologists asserts that sensitivity should be the determining factor in maximizing the number of depressed patients identified.16 This strategy is further supported by the finding that childbearing-aged adults have a relatively low threshold for acting on suicidal ideation (34% of those with suicidal thoughts developed a plan and 72% of those with a plan attempted suicide).17
Although most women with suicidal ideation were not acutely suicidal, timely assessment of these women is essential because women with high acuity are potentially “time-sensitive.” Our false-negative rate was low, with no known instance of suicide attempt among screened women who did not report suicidal ideation. It is also instructive that the one completed suicide involved a patient whose obstetrician chose not to screen her because of her “positive affect” during her postpartum visit.
When suicidal ideation is reported, even if there is no formal triage system, a few simple questions can be applied to gauge risk. Simply stating, “Many women who feel this way during or after pregnancy experience thoughts they would not normally have. Sometimes these can include not wanting to be here any more or thoughts of harming oneself. Is this happening to you?” Standardized tools such as the Columbia-Suicide Severity Rating Scale18 also can be used to identify women at highest risk.
We also found that predictors of suicidal ideation were similar during and after pregnancy while obstetric outcomes modified postpartum risk in a plausible manner. Reduced risk was found in association with cesarean delivery without labor, whereas greater risk was seen with severe lacerations. Avoiding physical discomfort and uncertain outcomes associated with labor could be experienced as an emotionally favorable circumstance, whereas the rigors of labor followed by perineal injury and protracted recovery could augment emotional distress. These clinical associations may prompt anticipatory guidance for selected women to counteract their potential mental health vulnerability.
The strengths of this study include its large sample size, accurate data retrieval through linkage of our screens to the electronic health record, uniform evaluations for all positive screens, and a universal screening strategy, which reduces the potential for selection bias. Although our screening efficiency improved over time, we still missed the opportunity to screen a portion of eligible patients. This challenge may not be easily surmounted because most of these women have limited prenatal care, late transfer of care, nonadherence to their postpartum visit, or other unmodifiable factors. We cannot determine whether nonscreened women would have different suicide risk, but their demographics and obstetric outcomes (data not shown) do not distinguish them from our screened cohort.
Additional study limitations include data from a single health system, which may limit generalizability, and access only to data retrievable from the electronic health record. Relevant missing variables include education level, domestic violence, current and historical mental health and substance use data, and a better estimate of income than the suboptimal surrogate of Medicaid. Furthermore, our patients are generally insured, partnered, and more often Caucasian, and these factors are negatively correlated with depression and suicide risk.19,20 Also, universal rather than selective mood screening results in ascertainment of mostly asymptomatic women, which may account for lower rates of depression than previously reported in studies using patient selectivity.21
These limitations notwithstanding, our findings support the utility of perinatal depression screening and the importance of having protocols in place for timely evaluation of at-risk women. In our cohort, 1% of those who reported suicidal ideation were assessed to be at high risk for self-harm, and we referred six highly acute women for emergent psychiatric care. We believe the magnitude of maternal suicide risk and the attendant rates of maternal and infant death are ample justifications for routine screening.
1. Palladino CL, Singh V, Campbell J, Flynn H, Gold KJ. Homicide and suicide during the perinatal period: findings from the National Violent Death Reporting System. Obstet Gynecol 2011;118:1056–63.
2. Oates M. Suicide: the leading cause of maternal death. Br J Psychiatry 2003;183:279–81.
3. Appleby L. Suicidal behaviour in childbearing women. Int Rev Psychiatry 1996;8:107–15.
4. Healey C, Morriss R, Henshaw C, Wadoo O, Sajjad A, Scholefield H, et al.. Self-harm in postpartum depression and referrals to a perinatal mental health team: an audit study. Arch Womens Ment Health 2013;16:237–45.
5. Newport DJ, Levey LC, Pennell PB, Ragan K, Stowe ZN. Suicidal ideation in pregnancy: assessment and clinical implications. Arch Womens Ment Health 2007;10:181–7.
6. Luoma JB, Martin CE, Pearson JL. Contact with mental health and primary care providers before suicide: a review of the evidence. Am J Psychiatry 2002;159:909–16.
7. Yawn BP, Dietrich AJ, Wollan P, Bertram S, Graham D, Huff J, et al.. TRIPPD: a practice-based network effectiveness study of postpartum depression screening and management. Ann Fam Med 2012;10:320–9.
8. Earls MF, Committee on Psychosocial Aspects of Child and Family Health American Academy of Pediatrics. Incorporating recognition and management of perinatal and postpartum depression into pediatric practice. Pediatrics 2010;126:1032–9.
9. U.S. Preventive Services Task Force. Screening for depression in adults: U.S. preventive services task force recommendation statement. Ann Intern Med 2009;151:784–92.
10. Rhodes AM, Segre LS. Perinatal depression: a review of US legislation and law. Arch Womens Ment Health 2013;16:259–70.
11. Murray D, Cox JL. Screening for depression during pregnancy with the Edinburgh Depression Scale (EPDS). J Reprod Infant Psychol 1990;8:99–107.
12. Cox JL, Holden JM, Sagovsky R. Detection of postnatal depression. Development of the 10-item Edinburgh Postnatal Depression Scale. Br J Psychiatry 1987;150:782–6.
13. Cox JL, Holden JM. Perinatal mental health: a guide to the Edinburgh Postnatal Depression Scale (EPDS). London (UK): The Royal College of Psychiatrists; 2003.
14. Patton MQ. Qualitative evaluation and research methods. Thousand Oaks (CA): Sage Publications; 1900.
15. Kleiman K. Therapy and the postpartum woman: notes on healing postpartum depression for clinicians and the women who seek their help. New York City (NY): Taylor and Francis; 2008.
16. Screening for depression during and after pregnancy. Committee Opinion No. 453. American College of Obstetricians and Gynecologists. Obstet Gynecol 2010;115:394–5.
17. Kessler RC, Borges G, Walters EE. Prevalence of and risk factors for lifetime suicide attempts in the National Comorbidity Survey. Arch Gen Psychiatry 1999;56:617–26.
18. Posner K, Brown GK, Stanley B, Brent DA, Yershova KV, Oquendo MA, et al.. The Columbia-Suicide Severity Rating Scale: initial validity and internal consistency findings from three multisite studies with adolescents and adults. Am J Psychiatry 2011;168:1266–77.
19. Gavin AR, Tabb KM, Melville JL, Guo Y, Katon W. Prevalence and correlates of suicidal ideation during pregnancy. Arch Womens Ment Health 2011;14:239–46.
20. Lancaster CA, Gold KJ, Flynn HA, Yoo H, Marcus SM, Davis MM. Risk factors for depressive symptoms during pregnancy: a systematic review. Am J Obstet Gynecol 2010;202:5–14.
21. Pope CJ, Xie B, Sharma V, Campbell MK. A prospective study of thoughts of self-harm and suicidal ideation during the postpartum period in women with mood disorders. Arch Womens Ment Health 2013;16:483–8.