- Top of page
Postpartum psychosis affects one or two in every 1000 childbearing women.[1-3] Its onset is rapid and usually occurs in the first 4 weeks after delivery. Affected women experience mania, severe depression, delusions, and confusion.[1, 3] Postpartum psychosis is a critical condition that usually requires hospitalisation,[1, 3] presenting risks for women and their offspring. For example, suicidal and homicidal ideation about their infants, although not common, are higher among women with postpartum psychosis than among other women.[3-5]
The potential to prevent and treat postpartum psychosis is linked to understanding its etiology and its triggers. It has been positively associated with primiparity,[2, 7] older maternal age, sleep loss,[9, 10] low social support,[8, 11] the use of prescribed drugs, and estrogen withdrawal. It is most strongly and consistently associated with a pre-conception history and family history of psychosis.[1-3, 6, 14-18] Harlow et al. reported that pre-conception psychotic or bipolar illness significantly increased the likelihood of postpartum hospitalisation for psychotic disorder, with increased likelihood among women with more recent, more frequent, and lengthier pre-conception psychiatric hospitalisations. Although women with pre-conception psychosis may account for less than 1% of a prenatal population, they account for more than half of all cases of postpartum psychosis, and may have a 100 times greater risk of postpartum psychiatric hospitalisation than women without such a pre-conception history. The question of whether postpartum psychosis is a distinct clinical entity is controversial, but many consider pregnancy and postpartum to be vulnerable periods for psychiatric episodes among women with pre-existing illness. A postpartum psychiatric hospitalisation may thus be considered ‘…a manifestation of a lifetime vulnerability to affective disorders with childbirth as the precipitating factor’. One common theory to explain why women may have a psychotic episode postpartum relates to their sensitivity to the loss of estrogen and progesterone after childbirth. Some hormones increase 200-fold over the course of pregnancy, and then abruptly decline with childbirth.[19, 20]
It is not clear whether prenatal, obstetric, or infant complications exacerbate the risk for a postpartum psychotic episode. Studies have found positive associations with caesarean delivery and delivery complications, including preterm delivery.[2, 7, 10, 15, 21, 22] Because postpartum psychosis is rare in the general population, many studies have been limited by small sample sizes. The most convincing evidence that postpartum psychosis may be associated with prenatal conditions comes from recent population-based epidemiological studies. A study that examined 1.1 million births in Sweden between 1975 and 2003 found that the risk for women with pre-pregnancy psychiatric hospitalisations was weakly, but positively, associated with preterm delivery and caesarean delivery, after adjustment for pre-pregnancy psychiatric history. Another recent population-based study of women who had no previous psychiatric hospitalisations (i.e. low-risk women) found that postpartum psychosis, although extremely rare, was positively associated with older maternal age and negatively associated with maternal diabetes and delivery of offspring with high birthweights.
We used the population-based Swedish medical birth register to assess the associations between prenatal, obstetric, and infant complications and postpartum psychosis. We examined the records of 1842 women who: (1) had a previous hospitalisation for psychiatric illness, putting them at high risk for postpartum psychosis;[14, 15] (2) were not hospitalised for psychiatric illness during pregnancy, thereby reducing the likelihood that postpartum hospitalisation was a continuation of a prenatal psychotic episode; and (3) were primiparous, thus excluding women whose only previous hospitalisation may have been for puerperal psychosis, which may be a discrete condition from other postpartum psychosis related to pre-existing illness.
- Top of page
Of the 1842 women with a pre-conception psychiatric hospitalisation, 1135 (62%) were hospitalised for schizophrenia or another psychotic disorder, and 7% of these women had psychotic paranoid states, 60% had reactive psychoses, and 33% had psychoses of childhood origin. The remaining 707 women (38%) had pre-conception admissions for bipolar disorder.
A total of 179 women (10%) were hospitalised for a psychiatric disorder within 90 days of delivery: 115 (64%) were hospitalised for psychotic disorders and 64 (36%) were hospitalised for bipolar disorder. In analyses adjusted for variables related to pre-conception psychiatric hospitalisation history, non-affective disorder diagnoses and recent, longer term, and multiple pre-conception psychiatric hospitalisations were independently associated with postpartum psychiatric hospitalisation (Table 2). Compared with the 620 women who had been hospitalised more than 6 years previously, postpartum hospitalisation was more likely for the 623 women who were hospitalised within the previous three years (5 and 15%, respectively; adjusted OR 3.2, 95% CI 2.1–5.0). Compared with the 502 women with pre-conception hospital stays of 0–7 days, the 665 women who had hospital stays of more than 28 days were 1.9 times more likely (95% CI 1.2–3.0) to have had a postpartum psychiatric hospitalisation (6.5 and 12.0%, respectively). Compared with the 937 women with one pre-conception psychiatric hospitalisation, the 545 women who had two or three pre-conception psychiatric hospitalisations had a 2.4 increased odds ratio of postpartum hospitalisation (4.5 and 11.9%, respectively; 95% CI 1.6– 3.7), and those with four or more pre-conception hospitalisations had a 5.1 times increased odds of postpartum hospitalisation (4.5 and 21.1%, respectively; 95% CI 3.4–7.7).
Table 2. Pre-conception psychiatric hospitalisation characteristics of 1842 primiparous women who had singleton live births and a history of pre-conception psychiatric hospitalisation, by postpartum hospitalisation for psychosis, taken from the Swedish medical birth register for 1987–2001
|Characteristic of pre-conception hospitalisation||No postpartum hospitalisation for psychosis (n = 1663)||Postpartum hospitalisation for psychosis (n = 179)||Adjusted OR (95% CI)a|| P b |
| Diagnosis at most recent pre-conception psychiatric hospitalisation |
|Schizophrenia or schizoaffective disorder||11.7%||11.8%||0.72 (0.45–1.18)||0.187|
|Bipolar disorder||39.5%||28.5%||0.55 (0.38–0.80)||0.002|
|Non-affective disorder||48.8%||57.0%||1.0|| |
| Recency of last pre-conception psychiatric hospitalisation |
|>6 years||35.5%||16.8%||1.0|| |
|3–6 years||32.5%||32.4%||2.15 (1.35–3.43)||0.001|
|<3 years||32.0%||50.8%||3.22 (2.07–5.01)||≤0.0001|
| Length of most recent pre-conception psychiatric hospitalisation |
|0–7 days||28.2%||18.4%||1.0|| |
|8–28 days||36.6%||36.9%||1.48 (0.94–2.33)||0.088|
|>28 days||35.2%||44.7%||1.90 (1.22–2.95)||0.004|
| Number of pre-conception psychiatric hospitalisations |
|4+||17.1%||42.5%||5.09 (3.38–7.67)||< 0.0001|
There was no difference in mean age between women who were and were not hospitalised postpartum for a psychiatric condition (mean 29.7, SD 4.8, and mean 29.6, SD 5.4 years, respectively; P = 0.79). Compared with women having fewer than 9 years of completed education, those with more education had two-fold higher odds of postpartum psychiatric hospitalisation (Table 3).
Table 3. Characteristics of 1842 primiparous women who had singleton live births and a history of pre-conception psychiatric hospitalisation, by postpartum hospitalisation for psychosis, taken from the Swedish medical birth register for 1987–2001
|Maternal characteristic||No postpartum hospitalisation for psychosis (n = 1663)||Postpartum hospitalisation for psychosis (n = 179)||Adjusted OR (95% CI)a || P b |
| Year of delivery |
| Age at delivery |
|<25 years||18.6%||14.5%||1.13 (0.63–2.02)||0.688|
|25–29 years||30.9%||35.8%||1.56 (0.96–2.53)||0.075|
|30–34 years||31.3%||35.8%||1.58 (0.97–2.56)||0.066|
|≥35 years||19.2%||14.0%||1.0|| |
| Country of birth |
|Other Nordic country||3.9%||5.6%||1.56 (0.71–3.35)||0.737|
|Non-Nordic country||7.2%||5.0%||0.69 (0.23–1.41)||0.658|
|≥9 years||84.5%||92.2%||2.05 (1.16–3.62)||0.014|
| Co-habitation with father of infant |
Women had between zero and five documented pregnancy-related complications, with no difference between women who were and were not hospitalised postpartum in the mean number of complications (mean 0.98, SD 0.99 and mean 0.86, SD 0.99 complications, respectively; P = 0.13) or the prevalence of at least one complication (63 and 55%, respectively; P = 0.27) (Table 4). The only statistically significant (P ≤ 0.05) difference was that, compared with non-hospitalised women, those hospitalised postpartum for a psychiatric condition had two-fold higher odds of puerperium complications (e.g. infections, lactation problems, and venous complications).
Table 4. Pregnancy-related complications and postpartum psychiatric hospitalisation for 1842 primiparous women who had singleton live births and a history of pre-conception psychiatric hospitalisation, taken from the Swedish medical birth register for 1987–2001
|Complication||No postpartum hospitalisation for psychosis (n = 1663)||Postpartum hospitalisation for psychosis (n = 179)||Unadjusted OR (95% CI)||Adjusted OR (95% CI)a |
| Pregnancy complications |
|Hypertensive diseasesb||5.7%||3.9%||0.67 (0.31–1.47)||0.62 (0.28–1.40)|
|Tobacco user at first prenatal visit||30.9%||40.8%||1.54 (1.13–2.12)g||1.30 (0.93–1.83) |
|Other pregnancy complicationsc||15.8%||22.4%||1.54 (1.06–2.24)f||1.27 (0.85–1.89)|
| Obstetric complications |
|Malposition or malpresentation||6.3%||6.2%||0.97 (0.51–1.85)||0.91 (0.46–1.78)|
|Early, false, or long labour||9.7%||10.6%||1.10 (0.67–1.83)||1.21 (0.71–2.06)|
|Caesarean delivery||17.6%||18.4%||1.06 (0.71–1.58)||1.03 (0.68–1.57)|
|Laceration or damage from instruments||8.3%||11.7%||1.47 (0.90–2.39)||1.42 (0.84–2.39)|
|Premature rupture of membranes||4.2%||1.7%||0.39 (0.12–1.25)||0.37 (0.11–1.21)|
|Other disorders of amniotic fluid and membranes||1.8%||1.1%||0.62 (0.15–2.60)||0.68 (0.15–3.07)|
|Other labour and delivery complicationsd||32.2%||34.1%||1.09 (0.79–1.51)||1.13 (0.80–1.60)|
| Puerperium complications not during or immediately after deliverye ||3.3%||5.0%||1.58 (0.77–3.25)||2.26 (1.00–4.90)f|
| Any of the above complications ||55.4%||62.6%||1.35 (0.98–1.85)||1.32 (0.94–1.84)|
In post hoc chi-square analyses, 78% of women who had puerperium complications and 55% of those who did not also had at least one pregnancy complication (P = 0.0004). In logistic regression analyses, adjusted for pre-conception hospitalisation variables, maternal age, and maternal education, women with pregnancy complications were 2.8 times more likely (95% CI 1.5–5.1) to have puerperium complications than those with no complications. Maternal death was not associated with postpartum psychiatric hospitalisation. Death was recorded for 20 women, with time of death ranging between 8 months and 13 years postpartum. There was no difference in the mean time of death between women with and without postpartum psychiatric hospitalisations (mean 4.1, SD 3.7 and mean 6.9, SD 3.3 years, respectively; P = 0.16).
There was no difference in the mean birthweight of offspring of women who were and were not hospitalised postpartum (mean 3401, SD 611 and mean 3382, SD 625 grams, respectively; P = 0.70), nor was there any difference in gestational age at delivery (39.2 weeks for both groups; P = 0.81). The mean Apgar score at 5 minutes was 9.6 for both groups: 96% of all infants had scores between 8 and 10. Fourteen infants were delivered after 42 weeks and 34 infants (1.9%) were large for gestational age, with no difference by group (P = 0.89).
Postpartum psychiatric hospitalisation was not associated with other infant complications, except for infant death (Table 5). Seventeen infants died within their first year of life. The odds of infant death was 4.1 times higher (95% CI 1.3–12.6) for infants of women who had a postpartum psychiatric hospitalisation (2.8%) compared with infants of those who did not (0.7%). Among the 17 infant deaths, 80% of the offspring of hospitalised women and 75% of the offspring of women who were not hospitalised postpartum died within the first month of life (P = 0.85). In post hoc analyses, infant mortality was not associated with any maternal characteristic, including variables related to pre-conception psychiatric hospitalisation. Infant mortality was positively associated with infant low birthweight (P = 0.0001), preterm delivery (P < 0.0001), and congenital anomalies (P = 0.0002). Five of the 17 infants who died were of low birthweight (29%), six infants were preterm (35%), and four infants had congenital anomalies (24%).
Table 5. Fetal and newborn complications and postpartum psychiatric hospitalisation for 1842 primiparous women who had singleton live births and a pre-conception history of psychiatric hospitalisation, taken from the Swedish medical birth register for 1987–2001
|Complication||No postpartum hospitalisation for psychosis (n = 1663)||Postpartum hospitalisation for psychosis (n = 179)||Unadjusted OR (95% CI)||Adjusted OR (95% CI)a |
|Congenital anomalies ||4.5%||6.7%||1.54 (0.82–2.90)||1.58 (0.82–3.07)|
|Low birthweight||6.5%||6.2%||0.94 (0.50–1.79)||0.85 (0.44–1.66)|
|Preterm delivery||8.8%||7.9%||0.90 (0.51–1.59)||0.80 (0.44–1.45)|
|Small for gestational age ||5.8%||6.3%||1.09 (0.57–2.08)||1.20 (0.61–2.35)|
|Other fetal or newborn complicationsb||19.1%||21.8%||1.18 (0.81–1.72)||0.96 (0.64–1.44)|
|Any of the above complications||22.5%||26.8%||1.26 (0.89–1.79)||1.06 (0.73–1.55)|
|Infant mortality (within 365 days of birth)||0.7%||2.8%||3.96 (1.38–11.36)c||4.10 (1.33–12.64)c|
- Top of page
We restricted our analyses to women who had been hospitalised for psychosis or bipolar disease prior to pregnancy. Such women are at 100 times greater risk for postpartum hospitalisation than women with no pre-pregnancy history, but this group of women are rarely studied. Our findings suggest that prenatal complications in women with pre-conception psychiatric hospitalisations will not help clinicians identify who, among these high-risk women, will have a postpartum psychotic or bipolar recurrence. Furthermore, we saw no association between number of complications and postpartum hospitalisation. Similarly, Blackmore et al. examined several maternal and pregnancy-related variables in 129 women with histories of bipolar spectrum disorder. They found that only primiparity and delivery complications were associated with postpartum psychosis; other factors, including a broad range of pregnancy complications, caesarean delivery, and infant gestation, were unrelated.
Several other studies have reported mixed associations with maternal and prenatal characteristics and postpartum psychosis, depression, and bipolar disease in general populations. One study, using the same data source we used, examined women with no history of psychiatric hospitalisation (i.e. women at very low risk) for a broad range of postpartum psychiatric illnesses. The authors reported that postpartum psychiatric illness was positively associated with older maternal age and negatively associated with infant birthweight and some maternal complications. We examined the same correlates and found no associations for women at high risk, suggesting that the etiology of postpartum psychiatric conditions may be different for women with a history of psychiatric illness than for women at very low risk with no such history. Another recent study using data from the Swedish medical birth register examined obstetric complications and postpartum psychosis in more than a million women who gave birth between 1975 and 2003. After statistical adjustment for pre-pregnancy psychiatric hospitalisation, the authors found that postpartum psychosis was slightly associated with preterm birth (OR 1.2), but was not associated with 16 other pregnancy (e.g. anaemia or pre-eclampsia) or infant (e.g. malposition, asphyxia or small for gestational age) complications. Similarly, Videbech and Gouliaev examined the Danish medical birth register data and found that only preterm delivery was associated with a first episode of psychosis within 1 year postpartum in the general population, but other pregnancy and birth complications were not. They also found that a first episode of postpartum psychosis was rare (one in every 2000 women), and that bipolar disorder was the most prevalent; their sample included no cases of schizophrenia.
Although prenatal complications were not associated with postpartum psychiatric hospitalisation, we found that women with puerperium complications were twice as likely as those without them to be hospitalised for a psychiatric illness within 90 days of a live birth. Less than 4% of our sample had such complications, and thus we did not have sufficient data to examine the nature of this association. This is a shortcoming of our study, given that the category of complications included diverse conditions (e.g. lactation problems, haemorrhage, or infection) that could have been chronic or acute, severe or mild. It is plausible that pregnancy-related postpartum conditions could be sufficiently stressful to affect maternal mental health, but we have not located other literature with which to compare our results. We are unable to assess the temporal association of the puerperium complications and postpartum hospitalisation. It is plausible that women who were hospitalised for a psychiatric condition were more likely to have a complication detected and documented on their medical records, and thus the increased risk we report here is a result of detection bias, rather than a true association.
We found no differences in infant complications between hospitalised and non-hospitalised women during the postpartum period. This may not be surprising, given that the two groups did not vary in maternal characteristics or pregnancy complications, and they all entered their pregnancies with similar psychiatric backgrounds. Despite there being no differences in vulnerability, the infants of hospitalised women were four times more likely to die in the first year of life, after adjustment for the variables associated with pre-conception hospitalisation (i.e. markers for the severity of maternal mental health problems prior to pregnancy). Maternal postpartum psychiatric hospitalisation itself was thus associated with infant survival, independent of the severity of maternal mental illness. As recently reviewed by Webb et al., parental psychosis is a risk marker for infant mortality. Two studies of Scandinavian populations similar to ours provide evidence of this. Bennedsen et al. found that infants of schizophrenic women had an almost three-fold higher risk for postneonatal death, but no increased risk for neonatal death, compared with infants of mothers in the general Danish population. Nilsson et al. reported a two-fold higher risk of mortality in infants of schizophrenic women compared with the general population, using the same data source that we used. Our work extends these findings because we restricted our analyses to infants of high-risk women and we examined the severity of psychosis (defined by postpartum hospitalisation). Like the other studies, we did not have sufficient data to examine temporality. In our sample, at least three-quarters of the infant deaths occurred during the first 30 days of life. We saw no difference in the timing of deaths by maternal hospitalisation within 90 days of delivery, but we do not know whether hospitalisation preceded or followed infant death. Because postpartum psychosis often has a rapid onset after delivery, we assume that maternal hospitalisation and infant death occurred close together in time. Without actual dates of hospitalisation and infant death, however, we can only make two conclusions: (1) infant mortality and the severity of postpartum mental health (as measured by hospitalisation) are associated; and (2) infant mortality was relatively rare in our sample of women with pre-conception psychiatric illnesses.
The population-based sample provided by the Swedish medical birth register was large enough to allow us to examine a rare outcome (postpartum psychiatric hospitalisation) and some rare exposures. Although we examined broad categories of obstetric complications, many studies about fetal or maternal psychiatric outcomes have categorised the discrete conditions we examined as a single category of obstetrical complications. Cannon and colleagues, in a review of maternal complications and offspring risk for schizophrenia, concluded that this common practice of grouping complications together limits extant research, as it is meaningless to consider diverse conditions as a single risk factor. The external validity of our findings may be limited to Sweden or to the study period, as patterns of diagnosis, care for psychiatric illness, and institutional pregnancy support may vary geographically, and/or change even during a short period of time. Women whose symptoms were less severe, or for whom psychiatric diagnoses were unclear, may not have been hospitalised. If this is the case, our outcome – hospitalisation – is a clear marker of severe postpartum psychiatric illness.
Whereas medical registries are an accessible, and perhaps the only feasible, data source for population-based studies, the precision of the data rely entirely on physician and hospital staff. Incomplete data could bias results, but only if they resulted in the differential misclassification of women who were and were not hospitalised postpartum for psychiatric illnesses. Because the prenatal and delivery data were recorded before the outcome, this is unlikely. In addition to the quality of the existing data, some variables were not available in our data source. Our analyses would have been stronger if we could have explored the risk associated with the use of anti-psychotic medications in the perinatal period. We were also unable to examine the associations of postpartum psychiatric illness and other reproductive events, such as endocrine function or postpartum hormonal contraceptive use. Other unmeasured and possibly important variables that would have strengthened our study include sleep quality,[9, 10] and partner and/or general social support.[1, 11, 21] Our findings of few associations between routinely documented variables and postpartum psychosis do not preclude – in fact, they may encourage – the future examination of the social and environmental factors that may exacerbate postpartum psychological risks in vulnerable women.