In what is billed as possibly the largest study of its kind ever, a team led by psychiatrist Katherine Wisner approached over 17,000 women during their maternity hospitalization. The investigator team succeeded in screening over 10,000 for depression and in providing diagnostic interviews to 973 of those who screened positive.
It turned out that 14 percent of the women were at an increased risk of postpartum depression, which is consistent with what other studies have found. But in Sit’s [Dorothy Sit, M.D., one of the study investigators], home visits for further evaluation revealed that in many cases, the symptoms were very serious.
“We discovered 20 percent had suicidal thoughts — these are thoughts of death, thoughts of wanting to die, not wanting to wake up, just escape,” Sit says. “In fact, some patients with very severe symptoms had made the decision to take their lives.”
Sit believes that all pregnant women and new moms should be screened for depression ideally at home so they can get diagnosed and receive treatment sooner, but even a clinical questionnaire may help. Lead author Dr. Katherine L. Wisner, a professor at Northwestern Medicine in Chicago, agreed. She added…that suicide makes up 20 percent of postpartum deaths, and is the second most common cause of death in postpartum women.
“Most of these women would not have been screened and therefore would not have been identified as seriously at risk,” Wisner explained to UPI. “We believe screening will save lives.”
June Horowitz, Professor of Nursing at Boston College and a postpartum depression researcher but who was not involved in the study, chimed in with CBS News:
“I think we still in this country really do not recognize mental health issues, we still have that Puritan, pull-yourself-up-by-your-bootstraps, be-tough attitude,” she said.
“It should be the gold standard that everyone gets screened for postpartum depression,” she added.
Yet, the American College of Obstetrics and Gynecology is still withholding a full endorsement of routine screening, saying that more evidence is needed before a recommendation can be made. How could that be?
The abstract of the JAMA psychiatry article is available here. Unfortunately the article itself is not open access, but if you email the lead author at firstname.lastname@example.org with the message…
Dear Professor Wisner: I would very much appreciate a PDF of Onset Timing, Thoughts of Self-harm, and Diagnoses in Postpartum Women…, which appeared in JAMA Psychiatry. Thank you in advance.
…I’m sure she would be pleased to send you the PDF.
The article reports a large, well resourced, NIMH funded study, but I am going to argue that it has serious limitations. Strong claims are made about the value of routinely screening pregnant and postpartum women for depression, the risk of suicide that screening would reveal among these women, the presence of unrecognized and untreated bipolar disorder, and the potential for saving lives with screening.
While the authors did an impressively effective job of packaging the research, getting it published in a high impact journal, and publicizing it in the media, the take away message was simplistic and misleading.
I know, these criticisms are rather harsh. If you have accessed this article through a University library website or successfully obtained a PDF from the lead author, you can read it ahead of time or you can read it as you follow along, and decide for yourself.
As with the last blog post, this study also needs to be situated in the context of the American healthcare system.
Many low income women in the United States have only inadequate insurance or none at all. They immediately qualify for Medicaid when they are found to be pregnant, but this insurance is canceled six weeks after delivery of their infant. Even having Medicaid is not a guaranteed access to depression care, because many psychiatrists and psychologists refuse to accept Medicaid because of what they consider too low reimbursement for services. But when their Medicaid eligibility is canceled, care for depression that low income women had been getting while pregnant would have to shift to community mental health centers. As is immediately apparent when someone enters a waiting room of a community mental health center, they serve mainly the severely mentally ill and dual diagnosis patients with substance abuse and severe mental illness. The waiting rooms of such settings can be frightening places for a new mother, particularly one who does not have a babysitter and must bring her infant with her.
As many as half of all low income women do not show up for their well baby visit and so are lost from the medical system. Many American women, both low income and otherwise, concentrate on medical care for their infant rather than themselves, and so pediatric care becomes the only possible site for identifying mothers’ postpartum depression. Unfortunately, pediatricians are not comfortable diagnosing and treating adult depression. These facts need to be taken into account when discussing proposals for screening and treating postpartum depression that include getting women into treatment quickly. It just won’t happen without a radical redesign of the system..
The study was conducted at an urban obstetrical hospital in Pittsburgh. Women who delivered a live infant were visited by a nurse or social worker on the ward who gave them information about the screening study. Those who agreed to participate were then screened by telephone for six weeks postpartum.
Screening was done with the Edinburgh Postpartum Depression Scale.
The 10 item scale is the one most commonly used for such purposes, but is hardly ideal. From item to item, the scoring key varies and reverses in ways that can be very confusing to patients and to clinic staff trying to score the instrument. You can see it here. Try to imagine filling it out in a visit as a pregnant woman in a busy obstetric clinic waiting room and decide whether you could keep track of the reversing of item content and flipping of scoring keys, without being alerted to them.
The EPDS was constructed in British colloquial English, and a quaint item 6, “Things are getting on top of me” confuses American women. When we were using it in a screening project at a Medicaid clinic, a woman muttered aloud in the waiting room, “Damn, that’s how I got here!” in response to item 6. Although there is poor validation of cut points for the EPDS, scores of 10 or 13 and higher are commonly used to determine positive screens and therefore the need for further evaluation.
Determination of self-harm in this study-or what was later called “suicidal thoughts” in the media-was entirely based on responses to item 10 of the EDPS, “The thought of harming myself as occurred to me,” With a timeframe of the past 7 days, the 4 response options being yes, quite often; sometimes; hardly ever; and never.
All women who screen positive were offered a psychiatric evaluation at their homes, and if they declined, they were offered a shorter telephone assessment of whether they had major depressive disorder. Those who opted for the home interview received the complete Structured Clinical Interview for DSM-IV (SCID), but those opted for the telephone assessment received only the portion of the SCID focusing on major depression.
Of the 17601 women approached, 17426 were eligible and agreed to telephone screening. The investigators were able to contact and screen 10,000 of them. Of the 1396 women who screened positive, 826 accepted a home interview and 147 accepted the telephone screen for major depression.
This is an ambitious study, but lots of women are lost from it, from the initial request for them to consent to participation in the study, and at each step forward. Let’s do some simple calculations. The investigators succeeded in screening 10,000/17426 or 57% of the eligible women they approached. They were then able to do diagnostic interviews in the homes of 826/1396 or 59% of the women screening positive. So, a rough estimate is that results of were obtained from (.57) x (.59) = 34% of the women eligible for an extensive interview, and it is on the basis of these interviews that the strongest claims are made.
This is a low proportion of recruiting and retaining pregnant and postpartum women, certainly lower than what my colleagues and I have obtained in other studies, and could bias results. Busy new mothers may not be interested in getting an extensive psychiatric interview at home. The investigators offered $40 which undoubtedly increased participation, but may have biased the final sample toward women who needed the money. And you can also see in the method section that the screening was part of recruitment for a randomized trial for management of depression. If women have adequate insurance, they would be less motivated to sign up for screening leading them to a clinical trial in which they may not get the treatment they prefer. On the other hand, low income, uninsured women may be attracted to the possibility of getting treatment they could otherwise not afford. So, a there is a further bias of oversampling low income women seeking the possibility of free care that they could not otherwise obtain.
Consistent with this criticism, the article indicates that the women who accepted a home visit had higher mean EPDS scores. They were younger, less educated, and more likely to be African American, on Medicaid or uninsured, and single.
Among the women receiving the more extensive diagnostic interview at home, the most common primary diagnosis was unipolar depressive disorder (566 women, 68.5%) with 514 (90.8%) having major depressive disorder. Most women with unipolar depression also had a comorbid anxiety disorder.
A “striking” 187 women (22.6%) were found to have a bipolar disorder.
Note that these percentages are calculated using the biased sample of women who completed a home interview, who, in turn, are a biased sample of those who scored positive on the questionnaire, who, in turn, are a biased sample of those approached in the waiting room. If the authors had provided simple estimates of the prevalence of psychiatric disorder they found in the waiting room, these estimates could be evaluated by direct comparisons with the existing literature, but unfortunately these numbers were not provided.
Overall, 98% of the women who scored above the cut point of 10 had a psychiatric disorder.
Having conducted this kind of research, I find this result astonishing and inconsistent with the rest of the literature. A low cut point of 10 on the EPDS is sometimes set so that no depression will be missed, but that comes at the cost of having to interview a lot of women who don’t turn out to be depressed. Most will not be. Technically, the issue is one of sacrificing specificity for sensitivity.
Those of us who are concerned about such things keep emphasizing to clinicians that a score on a questionnaire is no substitute for a careful interviewing of patients to see if they are actually depressed. We do not want to encourage overdiagnosis and overtreatment. Yet, if we take these results at face, clinician should be able to assume that most of the women who score above even this low threshold have a psychiatric disorder.
Then there is the issue of such a high proportion of women above this modest cut point being found to have a bipolar disorder. There is not a lot of literature concerning identifying postpartum women with bipolar disorder using this particular questionnaire, but I think many postpartum depression researchers like myself would find this figure so out of the range of expectations as to cause doubts about the validity of the study.
Picking up on the potential bias that I noted, many low income pregnant women are not married and often have conflictful relationships with the biological father, whom they do not necessarily live with, certainly by the time they have delivered their baby. When interviewers lack cultural sensitivity, these intermittent periods of irritability, outright , and emotional upset that might be interpreted as hypomanic episodes. Any substance abuse by the women or even the biological fathers can further complicate making diagnosis of a bipolar disorder in the women: women’s episodes of anger and conflict associated with their own or the father’s substance abuse might be misconstrued as a hypomanic episode. Certainly, there is evidence that insufficiently trained and supervised interviewers misdiagnose bipolar disorder and, in general, that the diagnosis is being made too casually in the community. Many persons reporting having been diagnosed are not even at risk for bipolar disorder.
There may be an additional bias of attaching a psychiatric label to normal phenomena in doing psychiatric interviews with new mothers. I am sure you have heard of the common and normal experience of postpartum blues. But what is less known that in the first five days after giving birth, about 15% of women experience a fleeting period in which their mood is elevated. They may feel particularly good and talk a lot. These women typically do not experience impairment in the social functioning or as a mother. If you examine the criteria for hypomanic episode, you can see that these criteria might be stretched to include a lot of these women, because they only need to meet four of the nine vague criteria. Yet it is not at all clear that these women should be considered as having a psychiatric disorder, and there are no evidence-based treatments or even an established need for treatment. Nonetheless, it’s conceivable that overzealous interviewers would classify them as having hypomanic episode and therefore bipolar 2 disorder. Making statements to the media counting these patients as having bipolar disorder is misleading.
Yet, the authors’ interpretation dismisses that possibility of bias and turns a weakness of the EPDS into a strength:
This figure is likely to be an underestimate of bipolar disorder episode frequency because the EPDS does not screen specifically for the hypomanic/manic phase of the disorder. The postpartum period carries the highest lifetime risk for first-onset and recurrent episodes of bipolar disorder.
They go on to highlight the importance of their finding
Recognition of bipolar disorder is the most important prerequisite for adequate treatment. Many patients receive treatment for comorbid psychiatric disorders, but lack of recognition of the underlying bipolar disorder results in few receiving appropriate treatment. Half of women with “treatment-resistant” PPD actually have bipolar disorder…Treatment of the depressed phase of bipolar disorder with a mood stabilizer and an antidepressant does not confer benefit beyond treatment with a mood stabilizer alone. Given the critical importance of birth as a life event for families, detection and treatment of bipolar disorder among childbearing women has major public health significance.
I find the claims made in the direct quotes from the investigator team to the media concerning self harm, or, as it is described there, suicidality, even more controversial, and so I will quote results as directly described in the article:
In the sample of 10 000 women who underwent screening, 319 (3.2%) had thoughts of self-harm, including 8 who endorsed “yes, quite often”; 65, “sometimes”; and 246, “hardly ever.”
Note that of the 10,000 women, few had thoughts of self harm and only 8 endorsed quite often. And, contrary to the way these results are presented in the media, this is not a question about suicidality, but about self harm. We actually get quite similar results with the EPDS item 10 in an ongoing study of women drawn from obstetric waiting rooms, but we give a very different interpretation. This item is notoriously nonspecific and only very weakly related to attempted or completed suicide. Think of it: it covers non-suicidal, fleeting thoughts like a woman feeling like she wants to bang her head on the wall or pinch or hit herself in frustration. Endorsing this item with other than a response of “never” is not strongly related to mental health or well being because most endorsements are not strong endorsements. Comparisons between answers to item 10 and what is said in an interview find that only a response of “yes, quite often” is related to suicidality and so only 8/10,000 of this sample is at risk.
This investigator team actually had much higher-quality data available from the SCID, which poses more specific questions about suicide and allows interviewers to explain the questions and probe patient responses. I’m quite confident that if the investigators had reported their own findings, they would be much less impressive, even if more accurate.
If we accepted at face what the investigators say in the article and are quoted as saying in the media.
We would conclude that women who are pregnant and who have just delivered a baby should be routinely screened for depression. Most having even modest elevations on the screening questionnaire have a psychiatric disorder. A considerable portion of those with elevated score are suicidal and suicide, we are told, is one of the leading cause of deaths in this time of women’s lives. So, when doing screening, a mental health professional should be nearby to deal with the prevalent “suicidality.” While routine screening could lead to effective treatment and saving lives, there is a danger because many of these women who screen positive because they actually have bipolar disorder, and so been given antidepressant could trigger a manic episode. So, we should have a psychiatrist readily available to determine whether they are suffering from a bipolar disorder.
Contrast with what is available in the literature.
Readers who uncritically accept what is available in the article and the media coverage might be surprised to find in the existing literature:
- Depression may be no more common during pregnancy and in the immediate postpartum than for women of similar age who are not childbearing.
- Pregnancy and the period after a live birth are particularly low risk times for suicide, which is not cost-effective to screen for in general medical settings, anyway.
- Reviews of the literature and evidence-based recommendations from some professional and government organizations consistently conclude that routine screening of pregnant and postpartum women for depression is not yet supported by the literature.
- Some states mandate and reward clinical settings for routinely screening pregnant and postpartum women, but the evidence is that mandating screening does not improve depression outcomes.
- Screening of pregnant and postpartum women is unlikely to lead to improvement in depression scores because of difficulties in linking women with affordable, accessible, and acceptable care. Both women and their care providers have reservations about the safety of antidepressants for the developing fetus and breast-fed infant.
- While psychotherapy may be appealing to some women, most cannot find time just after having a baby. There are substantial challenges in getting them to regular scheduled psychotherapy appointments, even when psychotherapy is affordable and accessible.
Conflict of interest in the promotion of treatment of pregnant and postpartum women with antidepressants and mood stabilizers
Considerable concern has been expressed that what professionals tell pregnant women about medication for depression and other mood disorders is tainted by substantial and sometimes undeclared conflicts of interest. Many prominent investigators receive funding for serving as advisors to pharmaceutical industry and giving talks to clinicians concerning the safety and efficacy of antidepressants. An article from on the bias in the published literature from the Investigative Fund at the Nation Institute specifically mentions Dr. Wisner’s ties to Pfizer, GlaxoSmithKline and Eli Lilly.
The conflict of interest disclosure for this article indicates that Dr. Wisner participates in the advisory board for Eli Lilly and that Dr. Wisniewski is a consultant with a number of pharmaceutical companies. You decide if this appearance of a conflict of interest might have influenced recommendations for screening with the need for a psychiatrist present and that is likely lead to overdiagnosis and overtreatment with antidepressants and mood stabilizers.