Distribution of first live births by women with nonpsychotic predelivery hospitalizations and psychotic prenatal and psychotic predelivery hospitalizations, Sweden, 1987-2001. *One hundred seven nonpyschotic prepregnancy and 18 prenatal hospitalizations. †Includes paranoid and reactive states, delusional disorders, and other acute and transient psychotic disorders without symptoms of schizophrenia.
Distribution of postpartum psychoses and bipolar disorder by postpartum week of hospitalization.
Harlow BL, Vitonis AF, Sparen P, Cnattingius S, Joffe H, Hultman CM. Incidence of Hospitalization for Postpartum Psychotic and Bipolar Episodes in Women With and Without Prior Prepregnancy or Prenatal Psychiatric Hospitalizations. Arch Gen Psychiatry. 2007;64(1):42-48. doi:10.1001/archpsyc.64.1.42
Postpartum psychosis occurs in 1 to 2 cases per 1000 live births. Most studies have not distinguished postpartum psychosis from bipolar disorder or the proportion of the incidence attributable to prepregnancy psychiatric morbidity.
To determine the incidence of postpartum psychosis and bipolar disorder attributable to previous psychiatric hospitalization.
Population-based study using linked registry data to determine postpartum onset of psychotic and bipolar episodes within 90 days after the first birth, by women with and without prepregnancy or prenatal psychiatric hospitalization. We assessed the type, number, and recency of previous hospitalizations on the incidence of hospitalization for postpartum psychotic and bipolar episodes.
Nationwide Swedish Hospital Discharge and Medical Birth registers.
Swedish women delivering a first live infant between January 1, 1987, and December 31, 2001.
Main Outcome Measures
Postpartum hospitalization for psychosis or bipolar disorder.
The cumulative incidences for postpartum psychotic and bipolar episodes (adjusted for age at first birth) were 0.07% and 0.03%, respectively. The incidence of psychiatric hospitalizations for postpartum psychotic or bipolar episodes among women without previous psychiatric hospitalizations was 0.04% and 0.01% of first births, respectively; for women with any psychiatric hospitalization before delivery, the incidence was 9.24% and 4.48%, respectively. For postpartum psychotic and bipolar episodes, the risk increased significantly with the recency of prepregnancy hospitalizations, number of previous hospitalizations, and length of most recent hospitalization. More than 40% of women hospitalized during the prenatal period for a bipolar or a psychotic condition were hospitalized again during the postpartum period. Approximately 90% of all postpartum psychotic and bipolar episodes occurred within the first 4 weeks after delivery.
Almost 10% of women hospitalized for psychiatric morbidity before delivery develop postpartum psychosis after their first birth. This underscores the need for obstetricians to assess history of psychiatric symptoms and, with pediatric and psychiatric colleagues, to optimize the treatment of mothers with psychiatric diagnoses through childbirth.
Previous studies have suggested that postpartum psychosis has a prevalence of 1 to 2 cases per 1000 postpartum women,1,2 and this prevalence does not appear to vary widely from culture to culture.2 These rates include both psychotic and bipolar episodes that occurred post partum. However, episodes of bipolar disorder may or may not include psychotic symptoms. Therefore, these estimates do not truly represent the incidence of postpartum psychosis. Second, this estimate of risk may be inflated because earlier studies included women with a history of postpartum illness, which is a substantial independent risk factor for recurrence of a postpartum psychotic or bipolar episode at subsequent births.3- 5 Third, it is unclear what proportion of the incidence of 1 per 1000 births is attributable to women who are ill during pregnancy and whether specific characteristics of prepregnancy psychiatric illnesses influence the risk of developing postpartum psychotic or bipolar episodes.
The nationwide Swedish Medical Birth and Hospital Discharge registers include information about all births and hospitalizations in Sweden. Each Swedish resident receives a unique national registration number at birth or as a result of residency that allows for a linkage between registries. We used these data sources to determine the extent to which hospitalization for previous psychiatric illnesses influences the incidence of postpartum psychotic and bipolar episodes.
The Swedish Medical Birth Register, held by the National Board of Health and Welfare, Stockholm, includes information on more than 99% of all births in Sweden. To distinguish the independent effect of hospitalization for nonpuerperal psychiatric disorders from that of previous postpartum events on the incidence of postpartum psychotic or bipolar illness episodes, we restricted the analyses to first births only. From the Medical Birth Register, we identified all women delivering their first infant between January 1, 1987, and December 31, 2001. After excluding 3 women because of incorrect or missing ages, 612 306 women were available for our analysis. To rule out the potential influence of adverse obstetric prenatal events on hospitalization for a postpartum psychotic or bipolar episode, we repeated all analyses in a subset of women whose first live birth was a singleton delivered at term (≥37 weeks) with no congenital malformations (n = 547 414).
Information in the Medical Birth Register is prospectively recorded on standardized antenatal, obstetric, and pediatric medical records. Copies of these records are forwarded to the Medical Birth Register for computerized data entry. At the time of the first prenatal visit, details are recorded regarding prepregnancy height and weight, current smoking status, maternal cohabitation with the biological father, and county of residence. Information on pregnancy-related complications is recorded during pregnancy. At the time of delivery, details on the date of delivery, labor and delivery complications, and congenital anomalies are recorded. Maternal age and parity are recorded at delivery, and information on the mother's country of birth is obtained by cross-linking the Medical Birth Register with the Register of Total Population held by Statistics Sweden, Stockholm. We also linked information on the mother's highest level of formal education (obtained from the Education Register, held by Statistics Sweden) to the Medical Birth Register. All births are validated annually against the Register of Total Population by using the mothers' and the infants' unique National Register numbers. This study was approved by the Research Ethics Committee of the Karolinska Institutet in Stockholm, Sweden.
Since 1973, the nationwide Swedish Hospital Discharge Register has been prospectively capturing details on virtually all psychiatric hospitalizations. Dates of each hospital admission and discharge and the main discharge diagnosis (and secondary diagnoses, if applicable) assigned by the treating physician are recorded according to the International Classification of Diseases, Eighth Revision (ICD-8) through 1986, the International Classification of Diseases, Ninth Revision (ICD-9) from 1987 through 1996, and the International Statistical Classification of Diseases, Tenth Revision (ICD-10) from 1997 through 2001. The diagnostic assessment is then forwarded electronically to the Hospital Discharge Register with standardized algorithms across Sweden. The Hospital Discharge Register had a nationwide coverage of inpatient treatment facilities during the study period and includes psychiatric and somatic clinics.
Several studies have investigated the agreement between Swedish register diagnoses and corresponding medical records.6- 8 In the most recent study, Ekholm et al6 found a 94% agreement between register diagnoses of schizophrenia and research diagnoses based on semistructured interviews and medical records. Kristjansson et al7 compared ICD-8 discharge diagnoses (including latent schizophrenia) with DSM-III diagnoses via medical record review and found 82.7% agreement. Dalman et al8 compared ICD-9 discharge diagnoses with DSM-IV diagnoses via medical record review of individuals aged 10 to 22 years and found 85.7% agreement. The conceptualization of schizophrenia in the Nordic countries is traditionally conservative. The diagnoses are assigned with great care, and prevailing validation studies confirm a low number of false-positive diagnoses.6
Using the person-unique national registration number assigned to all Swedish residents, we obtained information on all psychiatric admissions of mothers delivering a first birth between 1987 and 2001. By comparing the date of each hospitalization with the date of the first-birth delivery, we determined those admissions that occurred before pregnancy, during the prenatal period, and during the postpartum period. We considered a psychiatric admission to be prepregnancy if the date of admission occurred 42 weeks before the delivery of the first infant, prenatal if it occurred within 42 weeks before delivery, and post partum if the admission occurred within the first 90 days after delivery.
We classified psychiatric hospitalizations according to a discharge diagnosis of psychotic, bipolar, and other psychiatric disorders. Psychotic admissions were categorized as schizophrenia (ICD-8 and ICD-9 codes 295 and subcategories but excluding schizoaffective disorders, and ICD-10 codes F20, F23.1, and F23.2), schizoaffective disorders (ICD-8 and ICD-9 codes 295.7 and 295H, and ICD-10 codes F21 and F25), and other nonaffective psychoses (ICD-8 codes 297-299, ICD-9 codes 297-298, and ICD-10 codes F22-F24 without F23.1 and F23.2, and F28-F29). Approximately 7% of the nonaffective psychoses included women with psychotic paranoid states; 60% were termed reactive psychoses; and the remaining were psychoses with origin specific to childhood.
Admissions for an episode of bipolar illness were coded separately (ICD-8 and ICD-9 codes 296 and subcategories, and ICD-10 codes F30 and F31). The codes did not allow us to distinguish between episodes that were or were not psychotic or whether the episode was depressive or manic. When we identified women with multiple hospitalizations with different psychiatric diagnoses, each subject underwent evaluation by a board-certified psychiatrist (H.J.) blinded to postpartum status and was classified according to the predominant diagnosis across all prepregnancy and pregnancy-related hospitalizations. The remaining psychiatric hospitalizations included all other nonpsychotic and nonbipolar psychiatric mental health codes not listed from ICD-8, ICD-9, or ICD-10. More than 90% of women with postpartum psychotic or bipolar hospitalizations had no other postpartum mental disorder codes indicated. Among the small pool of women with multiple postpartum diagnostic codes, virtually all of the secondary codes were nonpsychotic or nonbipolar disorders.
We first categorized all women as having (1) no prepregnancy psychiatric hospitalizations, (2) prepregnancy or prenatal nonpsychotic and nonbipolar psychiatric hospitalizations, and (3) any prepregnancy or prenatal psychotic or bipolar hospitalizations. Differences in demographic and anthropometric characteristics were assessed among these groups. The percentages of postpartum psychotic and bipolar episodes were calculated among all women and then stratified by the type of prepregnancy or prenatal psychiatric admission. Adjustment for age at first birth was accomplished by direct standardization to the distribution among all women with no previous psychiatric hospitalizations using the formulas outlined by Greenland and Rothman.9 Using unconditional logistic regression, we assessed the influence of the length of hospitalization stay, interval between the hospitalization and first-birth delivery, and number of psychiatric hospitalizations on the risk of being hospitalized for a postpartum psychotic or bipolar episode.
To maintain consistency throughout the article, we refer to incidence as hospitalization for an episode of a psychotic or bipolar disorder during the postpartum period. We recognize that the episode may not be a first diagnosis but rather a continuation of an earlier prepregnancy or prenatal diagnosis or episode. However, the registry data do not allow us to make this distinction.
Among 612 306 women delivering a first infant, 2259 (0.37%) had a psychiatric hospitalization before delivery (Figure 1). Of these psychiatric admissions, 94.47% (n = 2134) were for an episode of a psychotic or bipolar disorder, and 13.45% (n = 287) occurred during the prenatal period, whereas the remaining 86.55% (n = 1847) occurred before the pregnancy.
Compared with women delivering their first infant with no previous psychiatric hospital admissions, those with previous psychiatric admissions were older at the time of their first birth, more likely to be current smokers, and less likely to be cohabiting with the father of the offspring (Table 1). There was little difference between women with and without prepregnancy or prenatal psychiatric hospitalizations by urban or rural residential location, education, country of birth, and year of first birth.
The overall age-adjusted prevalence of hospitalization for a postpartum psychotic episode after a first birth was 0.07%; for a postpartum episode of bipolar disorder, the prevalence was 0.03% (Table 2). These rates remained essentially unchanged when we restricted our population to women whose first infants were delivered live at term (≥37 weeks) with no congenital abnormalities (data not shown).
Of the 1348 women hospitalized for a psychotic disorder at any point before delivery, 14.54% experienced a postpartum recurrence that required hospitalization (Table 2). When stratified by women who were hospitalized for a predelivery psychotic episode that occurred before pregnancy and not prenatally, 9.07% were hospitalized for a postpartum recurrence, whereas among women hospitalized for a psychotic episode during the prenatal period, 44.05% had a postpartum recurrence. Similarly, of the 786 women hospitalized for an episode of bipolar disorder at any point before delivery, 8.5% had a recurrent episode during the postpartum period that required hospitalization. When stratified by the 710 women who were hospitalized for a predelivery bipolar episode before pregnancy and not prenatally, 5.15% were hospitalized for a postpartum recurrence, whereas among the 76 women hospitalized for bipolar disorder during the prenatal period, 41.26% had a postpartum recurrence. Only 2.6% of the 1348 women with predelivery hospitalizations for a psychotic disorder were diagnosed as having bipolar disorder during the postpartum hospitalization, and only 1.4% of the 786 women with predelivery hospitalizations for bipolar disorder were diagnosed as having a psychotic disorder during the postpartum hospitalization.
In Table 3, we assessed the incidence of hospitalization for postpartum psychotic episodes by the specific predelivery psychotic diagnosis, and then we further stratified these episodes by prepregnancy vs prenatal hospitalization. Of the 276 women diagnosed as having schizophrenia or the 47 women diagnosed as having a schizoaffective disorder during hospitalizations that occurred before the delivery of their first child, 21.72% and 22.21%, respectively, were hospitalized again during the postpartum period. In comparison, only 12.16% of the 1026 women diagnosed as having other nonaffective psychotic disorders during predelivery hospitalizations were hospitalized again during the postpartum period for a psychotic illness. In general, for all psychotic diagnoses, hospitalizations during the prenatal period resulted in higher rates of hospitalization for postpartum psychosis than those observed in women with hospitalizations that occurred only before pregnancy. Among women with prenatal hospitalizations, more than 90% of the 211 women diagnosed as having bipolar disorder, schizophrenia, or schizoaffective disorder during prenatal hospitalizations were diagnosed as having the same disorder during the postpartum hospitalization. Among 1137 women with prepregnancy but not prenatal hospitalizations, approximately 65% to 75% of diagnoses made during the prepregnancy hospitalization were concordant with the diagnosis made during the postpartum hospitalization.
Among the 1137 women with prepregnancy (but not prenatal) psychiatric hospital admissions, we observed a significant trend of increasing risk for postpartum psychotic episodes with more recent prepregnancy hospitalizations, more previous prepregnancy hospitalizations, and longer length of the most recent prepregnancy hospitalization stay (Table 4). We also observed similar findings for increasing risk of postpartum bipolar disorder with the recency and number of prepregnancy bipolar hospitalizations (Table 4).
Finally, as shown in Figure 2, regardless of the specific diagnosis made during a postpartum hospitalization, most of the postpartum admissions occurred within the first 2 weeks after delivery and then gradually declined through the remaining 3 months.
The combined incidence of hospitalization for an episode of psychotic and bipolar disorder during the postpartum period observed in this Swedish population (0.10%) was nearly identical to that observed by Terp and Mortensen10 (0.09%) using Danish registry data. Furthermore, our findings suggest that the incidence of hospitalization for an episode of postpartum psychotic or bipolar disorder is largely confined to women with a previous psychotic or bipolar illness. This supports with more accuracy and power an earlier study by McNeil4 in which 25 (28%) of 88 women with a history of nonorganic psychosis developed puerperal psychosis compared with none of the 104 pregnancy control subjects with no history of psychosis.
We chose to stratify predelivery hospitalizations by those hospitalized during the prenatal period as opposed to those hospitalized before pregnancy and not during the prenatal period. This stratification allowed for an assessment of the incidence and estimates of risk based on the characteristics of previous hospitalizations that were independent of the current pregnancy, because it is unclear whether postpartum hospitalizations for psychotic and bipolar episodes in women also hospitalized during the prenatal period represent a continuation of the same illness episode.
The strengths of our data are the nationwide coverage of the Swedish population and the limitation of our population to women delivering their first birth. Several investigators have suggested that a previous postpartum psychiatric illness substantially influences the risk of a new episode during a subsequent pregnancy. By restricting our analyses to women delivering their first infants, we effectively controlled for this confounding effect.
Kendell et al1 suggested that women with a history of manic-depressive illness had a much higher risk of psychiatric admission during the postpartum period compared with women with a history of schizophrenia. Others have also suggested that the proportion of schizophrenia among puerperal psychoses is small11 and that the postpartum period does not necessarily provoke a recurrent episode of a schizophrenic illness.12 Our data showed that 21.72% of women with a prepregnancy hospital admission for schizophrenia were hospitalized for a psychotic disorder during the postpartum period. Among women with prepregnancy bipolar disorder, 8.46% had another episode of bipolar disorder during the postpartum period. Thus, prepregnancy psychotic or bipolar illness substantially increases the risk of a postpartum psychotic or bipolar event.
It has also been suggested that the risk of a first psychiatric admission for schizophrenic psychosis is not increased after childbirth.10 Although we did not specifically compare the incidence of first hospitalization for schizophrenia before and during the postpartum period, our data would likely support this assertion because only 6 of 610 047 women with no previous psychiatric hospitalization were hospitalized for an episode of schizophrenia during the postpartum period after their first birth.
It has been suggested that women with schizophrenia develop psychotic symptoms later in the postpartum period than those with bipolar disorder or affective illness.4 If this were true, short definitions of the puerperal length might likely underestimate the number of cases of schizophrenia and schizoaffective disorders affected. However, our observed distribution of postpartum psychotic bipolar disorders by postpartum week of hospitalization appeared to be similar for each category of psychosis. Furthermore, approximately 90% of all postpartum hospitalizations for psychiatric illnesses occurred within the first 4 weeks after birth.13
We observed a high diagnostic concordance between predelivery and postpartum psychotic hospitalizations. Given that the term postpartum psychosis is not a discrete nosologic entity, our data support the view of psychotic episodes that essentially do not differ from those occurring at other times in the women's lives. We cannot determine from these data whether symptom exacerbation during pregnancy or in the postpartum period is associated with discontinuation of medication therapy due to patient choice or physician recommendations. A recent survey of women with bipolar disorder suggested that regardless of their educational and socioeconomic background or the characteristics of the physicians who care for them, these women are often ill-informed about the risk of perinatal exposure to psychotropic medication and the high rates of relapse during pregnancy and the postpartum period when psychotropic treatment is discontinued.14
A limitation of our data is that we were not able to validate the ICD codes pertaining to psychiatric admissions through case-by-case reviews or through diagnostic interviews. However, the diagnoses in the Swedish Hospital Discharge Register are considered to have high validity and reliability,7,8 and conceptualization of psychoses has traditionally been conservative.7 Furthermore, the restriction of the register to inpatient care is likely to further enhance the reliability of the clinical diagnoses. We recognize that the use of hospitalization for psychotic and bipolar disorders is a limitation. However, these postpartum conditions are considered medical emergencies and almost always result in hospitalization. A further limitation was our inability to determine, based on the ICD codes, whether women diagnosed as having bipolar disorder before or during pregnancy or post partum were psychotic during their hospitalizations. Although we suspect that most of the women hospitalized for bipolar disorder have psychotic symptoms, we cannot determine whether psychotic symptoms occurred more or less frequently during the different periods surrounding pregnancy.
Given that our results showed a high degree of concordance between the prepregnancy and postpartum psychiatric diagnoses, our combined rates of postpartum psychosis and bipolar disorder and the rate of previous psychiatric hospitalizations are likely to be quite reliable.
Of great importance is the determination of other factors that explain why some women do and some do not develop a psychotic or bipolar recurrence. Our data suggest that the recency and number of prepregnancy psychiatric hospitalizations and, to a lesser extent, the length of the most recent hospitalization substantially increase the likelihood of a postpartum psychotic or a bipolar illness. Future studies should also assess pregnancy characteristics to determine whether certain prenatal or obstetric markers can help identify those at higher risk of a new or a recurrent onset of postpartum psychotic or bipolar episodes.
Pregnancy is likely to be a period that triggers a higher incidence of recurrent psychotic or bipolar episodes compared with nonpuerperal periods. This finding underscores the need for obstetricians to adequately determine the prevalence of a history of severe psychiatric morbidity and to convey that information to their pediatric and psychiatric colleagues.
Correspondence: Bernard L. Harlow, PhD, Department of Epidemiology and Community Health, University of Minnesota School of Public Health, 1300 S 2nd St, Suite 300, Minneapolis, MN 55454 (firstname.lastname@example.org).
Submitted for Publication: October 19, 2005; final revision received April 24, 2006; accepted April 25, 2006.
Author Contributions: Drs Joffe and Hultman contributed equally to this manuscript as senior authors.
Financial Disclosure: None reported.
Funding/Support: This study was supported by a grant from the National Alliance for Research on Schizophrenia and Depression as part of the Toulmin Research Partners Program (Dr Harlow).