Figure 1. Flowchart illustrating the identification of the study population.
Figure 2. Kaplan-Meier curves illustrating conversion rates to diagnoses of bipolar affective disorder during a 15-year follow-up period. CI indicates confidence interval; Other points, women having their initial psychiatric episode before giving birth to their first live-born child or later than 12 months after giving birth to their first live-born child; 1-30 d Post partum, women having their initial psychiatric episode within the first 30 days after giving birth to their first live-born child; and 31-365 d Post partum, women having their initial psychiatric episode from 31 to 365 days after giving birth to their first live-born child.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Munk-Olsen T, Laursen TM, Meltzer-Brody S, Mortensen PB, Jones I. Psychiatric Disorders With Postpartum Onset: Possible Early Manifestations of Bipolar Affective Disorders. Arch Gen Psychiatry. 2012;69(4):428–434. doi:10.1001/archgenpsychiatry.2011.157
Author Affiliations: National Centre for Register-Based Research, Aarhus University, Aarhus, Denmark (Drs Munk-Olsen, Laursen, and Mortensen); Department of Psychiatry, UNC Center for Women's Mood Disorders, University of North Carolina at Chapel Hill (Dr Meltzer-Brody); and Department of Psychological Medicine and Neurology, Medical Research Council Centre for Neuropsychiatric Genetics and Genomics, Cardiff University, Cardiff, Wales (Dr Jones).
Context Childbirth has an important influence on the onset and course of bipolar affective disorder, and it is well established that there may be a delay of many years before receiving a diagnosis of bipolar disorder following an initial episode of psychiatric illness.
Objective To study to what extent psychiatric disorders with postpartum onset are early manifestations of an underlying bipolar affective disorder.
Design Survival analyses were performed in a register-based cohort study linking information from the Danish Civil Registration System and the Danish Psychiatric Central Register.
Participants A total of 120 378 women with a first-time psychiatric inpatient or outpatient contact with any type of mental disorder excluding bipolar affective disorder.
Main Outcome Measures Each woman was followed up individually from the day of discharge, with the outcome of interest being an inpatient or outpatient contact during the follow-up period with a first-time diagnosis of bipolar affective disorder.
Results A total of 3062 women were readmitted or had an outpatient contact with bipolar affective disorder diagnoses. A postpartum onset of symptoms within 0 to 14 days after delivery predicted subsequent conversion to bipolar disorder (relative risk = 4.26; 95% CI =3.11-5.85). Approximately 14% of women with first-time psychiatric contacts during the first postpartum month converted to a bipolar diagnosis within the 15-year follow-up period compared with 4% of women with a first psychiatric contact not related to childbirth. Postpartum inpatient admissions were also associated with higher conversion rates to bipolar disorder than outpatient contacts (relative risk = 2.16; 95% CI = 1.27-3.66).
Conclusions A psychiatric episode in the immediate postpartum period significantly predicted conversion to bipolar affective disorder during the follow-up period. Results indicate that the presentation of mental illness in the early postpartum period is a marker of possible underlying bipolarity.
Childbirth has an important influence on the onset and course of bipolar affective disorder,1 and studies have shown that episodes of postpartum psychosis are often best considered as presentations of bipolar affective disorder occurring at a time of dramatic psychological and physiological change.2 Despite diverse presentations, the specific link between childbirth and bipolar disorder is well documented in a range of different studies including population-based register studies,3 family studies,4,5 and genetic studies.6
It is also clear, however, that a high number of women presenting with the new onset of a psychiatric disorder in the immediate postpartum period do not receive a diagnosis of bipolar disorder. In a previous study, we showed that the relative risk (RR) of having a first-time inpatient psychiatric admission with bipolar disorder during the first 30 days post partum was 23.33 (95% CI, 11.52-47.24).3 However, among the total of more than 1100 women with first-time psychiatric admissions 0 to 12 months post partum, only 44 were diagnosed with bipolar disorder. This may, to some extent, be due to specific puerperal diagnostic categories being applied to women with bipolar episodes but clearly not all postpartum cases fall into the bipolar spectrum, and childbirth may be a trigger for a range of psychiatric diagnoses. It is also possible, however, that a significant proportion of postpartum episodes that receive another diagnosis do in fact occur in women with underlying bipolar illness. This may result from misdiagnosis if, eg, features of high mood in a mixed episode are missed.
It is well known that there are difficulties in diagnosing bipolar disorder, with often a considerable delay in an accurate diagnosis being made.7-9 In the Bipolar Comprehensive Outcomes Study, individuals with bipolar disorder showed an average of 9 years between the first episode of major depression and diagnosis.10 Similarly, this study revealed an average 5-year delay in receiving a bipolar diagnosis following the first symptoms of a manic episode. Depressive episodes occur in patients with both unipolar and bipolar disorder. For all individuals presenting with depression, there is therefore a risk of subsequent episodes of high mood and a diagnostic conversion from unipolar disorder to bipolar disorder. A study by Angst et al11 has quantified this risk and found a diagnostic change from unipolar major depression to bipolar I disorder in 1% of patients per year and to bipolar II disorder in about 0.5% of patients per year.
Although a diagnostic shift to bipolar disorder is more unusual than for other initial diagnoses, it can occur. For example, there are issues in the diagnostic stability of a first psychotic episode, with the later emergence of affective features resulting in a revision from a nonaffective psychosis to bipolar or schizoaffective disorder in a proportion of patients.7,9
Given these diagnostic challenges and the very strong relationship between childbirth and bipolar disorder, we hypothesized that a postpartum onset of psychiatric illness could help in the difficult task of diagnosis with the triggering of illness by childbirth being a marker of underlying bipolarity. We aimed to study to what extent nonbipolar postpartum episodes convert to bipolar affective disorders over time using registry data in a longitudinal study based on data derived from the entire Danish population.
We established a population-based cohort using information from the Danish Civil Registration System.12 The register was established in 1968 and holds information on sex, date of birth, and parents' Civil Registration System numbers and is updated daily on vital status and migration. All citizens in Denmark are assigned a personal identification number (Civil Registration System number), which is used in all registers and makes linkage between and within registers possible. We included all women with records of 1 psychiatric contact who were born in Denmark from January 1, 1950, until December 31, 1991, who were alive on their 15th birthday (N = 120 378) of which 2870 had their first psychiatric contact within the first year after delivery of their first child.
Data on psychiatric disorders came from the Danish Psychiatric Central Register, which holds information on all admissions to psychiatric hospitals in Denmark since 1969. Furthermore, it contains records of psychiatric contacts at outpatient clinics since 1995. At present, the register holds information on more than 725 000 persons and 3.25 million contacts. From 1969 to 1993, the diagnostic system used in the registries was the Danish adaption of the International Classification of Diseases, Eighth Revision (ICD-8)13 and ICD-10 from 1994 and onward.14 The study population and their parents were linked to the register to obtain information on mental illness to study psychiatric disorders in both the probands and presence or absence of psychiatric disorders in their families. Women and their parents were classified as having a psychiatric disorder if they had records of either inpatient or outpatient contacts at a psychiatric hospital with any type of mental illness (main diagnoses only). For the women in the study, diagnoses were grouped as follows: bipolar affective disorders (ICD-8 codes 296.19, 296.39, and 298.19; ICD-10 codes F30 and F31), unipolar affective disorders (ICD-8 codes 296.09, 296.29, 296.99, 298.09, 300.49, and 300.1x; ICD-10 codes F32, F33, F34.1, F38.8, and F39.0), schizophrenia-like disorders (ICD-8 codes 295.xx, 297.xx, 298.39, and 301.89; ICD-10 code F2x.xx), adjustment disorders (ICD-8 codes 307.xx and 308.4x; ICD-10 code F43.xx), personality disorders (ICD-8 code 301.xx minus 301.83; ICD-10 codes F60 and F61), puerperal disorders not elsewhere classified (ICD-8 code 294.49; ICD-10 code F53.xx), and other disorders (remaining diagnoses). Parents of the women in the study were categorized by their history of psychiatric disorders in the following way: history of psychiatric contact with bipolar affective disorder (ICD-8 codes 296.19, 296.39, and 298.19; ICD-10 codes F30 and F31), history of psychiatric contact with any type of mental illness excluding bipolar disorder (remaining diagnoses), and no history of psychiatric contacts.
Based on information from these registries, we studied the largest possible homogenous population for which we had sufficient information. The study population consisted of women born in Denmark from 1950 to 1991 who were alive and had a history of a first-time psychiatric contact with any type of psychiatric disorder excluding bipolar affective disorder (N = 120 378) (Figure 1). Data derived from the registries were handled and analyzed prospectively, and follow-up started on January 1, 1970, or at age 15 years, whichever came last. Specifically, women 15 years and older were followed up individually from date of discharge of first inpatient or outpatient contact. Follow-up ended on the date of first-ever psychiatric inpatient or outpatient contact with bipolar affective disorder, death, or emigration or by December 31, 2006. The outcome of interest was incident psychiatric contacts (defined as inpatient or outpatient contacts at psychiatric hospitals) with a bipolar affective disorder.
We excluded women from the study who had bipolar affective disorder diagnoses at their initial contact (first-time contact at either an inpatient or outpatient psychiatric facility). Women diagnosed with bipolar affective disorder during a psychiatric contact immediately following the first contact (discharge and readmission within 24 hours) were also excluded from the study together with women younger than 15 years at first psychiatric contact (Figure 1).
Each woman was followed up individually during the specified follow-up period using survival analysis techniques. Data were analyzed using Poisson regression with the logarithm of the person-years as an offset using SAS version 9.1 (SAS Institute Inc). This method approximates a Cox regression.15,16 The main outcome measures were number of new cases per period (1000 years), ie, incidence rates of first psychiatric contacts with bipolar affective disorder diagnoses. Age, calendar period, and psychiatric disorders in parents of the women were treated as time-dependent variables, whereas the remaining variables were treated as time-independent variables.17 When comparing incidence rates between 2 groups, the outcome measures were incidence rate ratios, which can be interpreted as RRs. Kaplan-Meier plots were made to illustrate the conversion rates to bipolar affective disorders in women having initial psychiatric contact within the first months after delivery, between 2 to 12 months post partum, or at other points.17
The study was approved by the Danish Data Protection Agency. Informed consent from the women in the cohort was not required for use of registry data in this type of research.
During 1970 to 2006, 120 378 women had an initial psychiatric contact (admission or outpatient contact) with any type of psychiatric disorder excluding bipolar affective disorder. A total of 2870 of these women had their first psychiatric contact within the first year after delivery of their first child. During follow-up, 3062 of the 120 378 women received diagnoses of bipolar affective disorder at a subsequent psychiatric admission or outpatient contact, of which 132 had their initial psychiatric contact 0 to 12 months post partum (Figure 1).
Women having a first-ever psychiatric contact within the first month post partum demonstrated an increased probability of converting to bipolar affective disorder at a later stage: initial contact 0 to 14 days post partum, RR = 4.26 (95% CI = 3.11-5.85) and initial contact 15 to 30 days post partum, RR = 2.65 (95% CI = 1.72-4.07) (Table 1). This was compared with the reference category, which was women with initial psychiatric contacts at other points excluding the first year after delivery. A total of 47 women specifically received a diagnosis of mental and behavioral disorders associated with the puerperium, not elsewhere classified (ICD-8 code 294.49; ICD-10 code F53.XX). Since these disorders are restricted to an onset within the first 6 weeks post partum, women with these diagnoses will, by necessity, at subsequent nonpostpartum psychiatric contact receive an alternative diagnosis. We have therefore conducted additional analysis (Table 2) excluding the women with an initial diagnosis of puerperal disorders in which a similar pattern of results were observed.
As discussed earlier, the rates of subsequent conversion to bipolar disorder are likely to differ depending on the initial diagnosis and it is well established that a family history of bipolar disorder predicts higher rates of bipolar affective disorders in offspring.18 Therefore, we conducted multivariate analyses to adjust for these and other potential confounders. After taking first diagnoses and family history of psychiatric illness into account, conversion rates to bipolar disorder were still significantly predicted by the timing of initial contact. Although reduced compared with the results presented in Table 1, there was a significantly higher conversion rate to bipolar affective disorder in women having their initial psychiatric contact within the first postpartum month: initial contact 0 to 14 days post partum, RR = 2.53 (95% CI = 1.57-4.07) and initial contact 15 to 30 days post partum, RR = 1.82 (95% CI = 1.07-3.10) (Table 3). Additionally, there is evidence that the severity of the initial postpartum episode may be important because inpatient admissions were associated with a higher conversion rate than outpatient contacts (RR = 2.16; 95% CI = 1.27-3.66). Specific diagnosis at the initial postpartum psychiatric contact also predicted increased conversion rates to bipolar disorder: unipolar affective disorder, RR = 2.88 (95% CI = 1.53-5.43); schizophrenia-like disorders, RR = 2.57 (95% CI = 1.19-5.57); and puerperal disorders, RR = 2.99 (95% CI = 1.51-5.92). An initial diagnosis of a personality disorder did not predict an increased conversion rate. Additionally, if the father of the woman had a history of bipolar affective disorder, conversion rates were significantly higher (RR = 3.49; 95% CI = 1.22-10.01) compared with both women with a family history of other psychiatric disorders (RR = 0.98; 95% CI = 0.53-1.82) and women with psychiatrically healthy parents (reference category) (Table 3).
Kaplan-Meier curves were made to illustrate the conversion rates to bipolar disorder among the women in the cohort. Fifteen years after initial contact, 13.87% (95% CI = 10.00-17.58) of women with onset in the immediate postpartum period (0-30 days after delivery) had converted to bipolar affective disorder. In comparison, after 15 years, 4.69% (95% CI = 3.47-5.90) of women with later postpartum onset (31-365 days after delivery) and 4.04% (95% CI = 3.88-4.21) with onset at other points had converted to bipolar disorder (Figure 2), with the later numbers not being significantly different. We extended the Kaplan-Meier curves to include up to 22 years after first admission. Because there were few cases in the final years of the follow-up period, results should be interpreted cautiously; however, 18.98% (95% CI = 13.86-23.80) of the women with immediate postpartum onset had converted to bipolar disorders within 22 years after the initial postpartum psychiatric contact. In comparison, only 6.51% (95% CI = 4.81-8.18) with later postpartum onset and 5.43% (95% CI = 5.19-5.66) with onset outside the first postpartum year converted to bipolar disorder within the same follow-up period.
The results of this study clearly demonstrate that a postpartum onset raises the possibility of underlying bipolarity. Almost 14% of women with a first-time psychiatric contact during the first month after delivery converted to bipolar affective disorder within the first 15 years after their initial postpartum episode, which was 3 times more often than women with initial psychiatric contact at other points. Our results show additional factors that are important determinants of the risk of conversion to bipolar affective disorder.
Timing of initial psychiatric contact was a major predictor for conversion to bipolar disorder. Contact with psychiatric services during the first 30 days post partum predicted conversion to a subsequent diagnosis of bipolar affective disorder, whereas psychiatric contacts later in the first postpartum year did not show an increased risk of conversion.
Inpatient admissions in the postpartum period were associated with a higher conversion rate to bipolar disorder than outpatient contacts.
Despite improvements in reliability over recent decades, the diagnosis of psychiatric disorders, particularly of first episodes, is often unclear and needs to be revised as the illness develops. Although conversion from bipolar disorder to another diagnosis may be rare, a conversion to bipolar disorder is more common. The findings of this study present further evidence that the onset of psychiatric illness in the early postpartum period is of potential diagnostic importance. Postpartum onset should be added to the list of features in the presentation of depression that raise the likelihood of latent bipolarity. These include early age at onset, the presence of psychosis, atypical features, and an abrupt onset.19 In addition, our work demonstrates that an early postpartum onset of illness has prognostic implications and raises the chance of subsequent conversion to a diagnosis of bipolar disorder. This does not apply merely to women with postpartum unipolar depression (RR = 2.88; 95% CI = 1.53-5.43) but also to women with nonaffective psychosis/schizophrenia-like disorders (RR = 2.57; 95% CI = 1.19-5.57) (Table 3).
The results of this study provide further evidence linking childbirth-related episodes and bipolar disorder, building on our previous studies that have demonstrated a strong and specific link.3-5,20 Despite the strength of this association, the proportion of new postpartum episodes receiving a diagnosis of bipolar disorder was relatively low since many other disorders can have their onset after childbirth. Figure 1 illustrates that 38 women had their first-ever psychiatric contact with a bipolar diagnosis 0 to 30 days post partum, 27 had initial bipolar contact 31 to 365 days post partum, and 1662 had their initial contact with bipolar disorder at other points. All these women were excluded from the study, as their first psychiatric contact was specifically recorded as a diagnosis of bipolar affective disorder. Our results suggest, however, that at least some of the women receiving other diagnoses following childbirth have an underlying bipolar spectrum illness and, moreover, that this is considerably more common than for first episodes occurring at other times.
Because of the specific link between childbirth and bipolar affective disorder, it is critically important to conduct a comprehensive assessment of women presenting in the postpartum period. In particular, the clinician should pay careful attention to symptoms of hypomania or a mixed episode because the misdiagnosis of a bipolar spectrum disorder can have serious consequences. For example, the prescription of antidepressants is exceedingly common in the treatment of unipolar major depression but can cause or exacerbate manic symptoms in bipolar spectrum disorders.21 The Mood Disorder Questionnaire was recently shown to be a useful screening instrument for bipolar disorder in the postpartum period,22 and there are other screening tools (both semistructured interviews and self-report measures) for bipolar disorder in general that can be used following childbirth.23
We speculated that severity of illness could predict conversion to bipolar disorder. For this reason, we studied differences in conversion rates between those women who received treatment at outpatient services vs those requiring inpatient hospitalizations at their first psychiatric contact. Interestingly, type of first contact did predict conversion to bipolar disorder since women having inpatient treatment were more likely to convert than women having outpatient treatment (RR = 2.16; 95% CI = 1.27-3.66). These results can most likely be explained by the severity of first-ever psychiatric contacts, with inpatient admissions being markers of more severe episodes vs milder episodes, which will be treated at outpatient clinics.
At a time when revisions to the DSM and ICD classification systems are being considered, our findings are very relevant to the debate about how postpartum disorders should be treated. Previous work has demonstrated that a postpartum onset has important prognostic implications; women with an episode in relation to childbirth are at considerably higher risk of a postpartum recurrence following future pregnancies.24 Our work provides evidence that the relationship of the onset of symptoms to childbirth has diagnostic significance as well. A postpartum episode of a mood disorder or nonaffective psychosis is a marker that raises the possibility of both a current and a subsequent bipolar diagnosis. This study adds to the evidence for retaining the postpartum onset specifier for mood disorders in DSM-5, and even expanding its application to psychotic diagnoses.
Using population-based registers for conducting research offers both advantages and limitations. The present study uses data from registers that cover the entire Danish population where all people have free access to health care. It included a large study population with low attrition rates and prospectively recorded information on exposures and outcomes, minimizing both recall and selection bias. To study diagnostic paths among women with first-time psychiatric contacts within the postpartum period requires large study populations because severe postpartum mental disorders necessitating admission are rare.3 Long follow-up periods are also needed to fully map conversion rates to bipolar disorder, and population registers are ideal for conducting these types of epidemiologic studies.
The present study has some limitations. First, the episodes studied were psychiatric inpatient and outpatient contacts, which restricts analyses to only those women who sought psychiatric care. Consequently, we can make no comment on less severe episodes of illness in which patients do not present to psychiatric services. It is possible that women in the postpartum period who have increased contact with health professionals are more likely to have milder episodes of illness identified and therefore referred to psychiatric services. If, as seems likely, women with less severe episodes of a mood disorder are less likely to go on to receive a bipolar diagnosis, we may have underestimated the true difference in conversion rates. Second, our study may have underestimated the true rate of conversion to bipolar spectrum disorders because a number of women may have developed significant bipolar symptoms but did not present to mental health services. This may account for the lower rates of conversion to bipolar illness we observed when comparing with studies involving follow-up of participants with a research interview focused on bipolarity.11 This issue applies equally across all the groups studied, both those with a postpartum and a nonpostpartum onset, and it is the substantial and significant difference between groups that is the important finding of our study rather than the absolute conversion rate. Third, in the registers we defined a postpartum contact as the date of psychiatric inpatient or outpatient treatment and the true onset of symptoms is likely to have been before this date. In addition, some patients may have had a prepartum onset with exacerbation or symptoms after the delivery. Fourth, lifetime conversion rates to bipolar disorder could not be established since women in the cohort were aged up to 57 years. Furthermore, the classification of specific psychiatric diagnoses was based on clinical diagnoses in the registry rather than research diagnostic criteria. However, previous validation studies have found high agreement between clinical and research diagnoses.25,26
Last and most importantly, from the available data it was not possible to study if women with subsequent bipolar diagnoses were misdiagnosed at initial contacts during the postpartum period or developed the disorder during the follow-up period. It is possible that the perinatal context has important influences on the presentation of mood and psychotic episodes and those features of severe postpartum psychiatric disorders. Perplexity and a rapidly changing clinical picture make it more difficult to make a correct diagnosis of the underlying clinical condition. Alternatively, it is possible that there are no differences in the accuracy of diagnosis in postpartum and nonpostpartum episodes and that the differences in subsequent conversion rates are real. We are unable to differentiate between these possibilities but whatever the balance between the two, the main implication for clinical practice—that a postpartum onset should raise the suspicion of an underlying bipolar disorder—would remain unchanged.
The present study confirms the well-established link between childbirth and bipolar affective disorder and specifically adds to this field of research by demonstrating that initial psychiatric contact within the first 30 days post partum significantly predicted conversion to bipolar affective disorder during the follow-up period. This was compared with both women with first psychiatric contact in the later part of the postpartum period and women with nonpostpartum initial contacts with psychiatric illnesses. The perinatal context of current or previous episodes is an important element to consider in psychiatric assessment, with both diagnostic and prognostic implications.
Correspondence: Trine Munk-Olsen, PhD, National Centre for Register-Based Research, Aarhus University, Taasingegade 1, Aarhus, 8000 Denmark (firstname.lastname@example.org).
Submitted for Publication: May 11, 2011; final revision received September 5, 2011; accepted October 7, 2011.
Published Online: December 5, 2011. doi:10.1001/archgenpsychiatry.2011.157
Author Contributions: Drs Munk-Olsen and Laursen had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosure: None reported.
Funding/Support: Dr Munk-Olsen receives financial support from the Danish Medical Research Council (reference number 09-063642/FSS). Drs Laursen and Mortensen are funded by the Stanley Medical Research Institute. Dr Meltzer-Brody is funded by National Institutes of Health grant K23MH85165-3.
Role of the Sponsors: The sponsors had no role in the design of the study, analysis and interpretation of the data, or writing of the manuscript.