5.1. INTRODUCTION
Pregnancy and the postnatal period are critical times of psychological adjustment for women, and there is increasing evidence that a woman’s mental state during this time influences both obstetric outcomes and the future development of the infant (for example, Jablensky et al., 2005; Nulman et al., 2002). Mental ill health in the antenatal and postnatal periods can also affect other children in the family, as well as the woman’s partner and their relationship. Severe mental illness, such as psychosis, bipolar disorder or severe depression, may be particularly detrimental, both during pregnancy and subsequently, given the dependence of an infant on its mother and the rapid adjustment to motherhood faced by first-time mothers. Therefore, accurate identification of both those at risk of developing, and those currently suffering from, mental illness during this time is highly desirable.
Although mental disorders experienced during the antenatal and postnatal periods, with the possible exception of puerperal psychosis, are broadly no different in terms of aetiology and diagnostic characteristics to disorders occurring at other times, women typically have frequent contact with a range of healthcare professionals during pregnancy, childbirth and the postnatal period. Such routine contact provides an important opportunity to identify those at risk of developing, or currently suffering from, mental disorders.
This chapter addresses whether it is possible to predict postnatal mental illness during pregnancy and how to detect current illness.
Vignette: A woman with a history of depression and abuse who had anxiety postnatally but did not go on to develop depression in the postnatal period
I am 40 years old and had my first child 11 months ago. My husband and I planned to have a baby and I became pregnant right away. I enjoyed being pregnant and kept well. My son was born in March 2005, 10 days overdue. I coped with labour using gas and air, but needed a ventouse delivery and stitches for a third-degree tear.
I experienced clinical depression in my twenties. I was sexually assaulted by a boyfriend and this triggered memories and flashbacks of sexual abuse from my father when I was a child. I went to my GP and received counselling and a psychiatric assessment. I was prescribed an antidepressant, which I took for about 6 months. I also contacted the Scottish Society for the Prevention of Cruelty to Children who ran a counselling service for adult survivors of child sexual abuse and I got a lot of support individually and in a group.
When I considered starting a family, I was scared that I would develop postnatal depression. I read about it and spoke to my midwife about the risk factors and ways to help avoid postnatal depression. I talked about being depressed before, but I did not go into details about the reasons. My health visitor gave me good advice about postnatal depression and kept a good check on me after my son was born.
I was also scared about how I would cope with labour. I was worried that I would panic because I would feel very exposed and vulnerable. I talked to my husband about this and we discussed ways that he could reassure me and support me during labour. But I did not discuss this with the doctor or midwife. I reckoned that everyone in labour probably feels scared and vulnerable and I hoped that the staff would be reassuring no matter what the circumstances. Also, there never seemed to be an appropriate time to talk about emotional concerns. The emphasis at the antenatal clinic was on physical aspects of pregnancy. In the event, I felt able to cope in labour up until I needed a ventouse. At that point, a lot of staff came into the room, both for me and for my son, and I became very upset. My husband asked some people to leave and so only the very necessary staff stayed and the consultant dimmed the lights for me and reassured me and calmed me down.
After my son was born, I experienced a lot of emotional ups and downs, but my health visitor reassured me that these feelings were all very normal after having a baby. She encouraged me, telling me that I was doing very well. I attended a breastfeeding support group where I got lots of advice and support from staff and other mums. I am still friendly with some of the girls I met there. Talking to them really helps when I am worried or feel down.
The hardest thing I have found to deal with since my son’s birth is anxiety about being separated from him. I was very anxious about returning to work because of leaving him in a nursery. My GP reassured me that this was a normal feeling. I attended a talk given by STEPS, an NHS self-referral psychological service, which gave general advice on coping with stress after having a baby. I saw a STEPS counsellor for just one session to talk about my worries. I told her about how I had been abused as a child and how this made me scared about leaving my son with anyone else in case something happened to him. She reassured me that my feelings were valid and she gave me good information about how to minimise the chances of such things happening to him and what signs there would be to look out for. She gave me more confidence in being able to keep my son safe.
I returned to work part-time when my son was 9 months old. I still don’t enjoy leaving him at nursery, but I am enjoying being back at work and am better able to leave him with a trusted babysitter so that I can get a break.
At this moment in time, I am mostly enjoying being a mum. I find it very tiring and I worry a lot, but from talking to other mums I realise that they all feel the same too! Being depressed in the past made me very aware of the risks of becoming depressed again, so I tried my best to prepare for that and to get advice and support to help avoid illness. I am also aware that I might be overprotective of my son, so I am trying to avoid that. I want him to grow up feeling safe and secure and confident.
I feel that I have had appropriate support from my GP, psychiatrist, counsellor, midwife and especially my health visitor. I also got good support from my husband and friends from the breastfeeding support group.
5.1.1. Shifting the emphasis from screening to prediction and detection
Screening has been defined as the systematic application of a test or enquiry to identify individuals at high risk of developing a specific disorder who may benefit from further investigation or preventative action (Peckham & Dezateux, 1998). Screening programmes detect people at risk of having the condition or at risk of developing the condition in the future. They do not establish a diagnosis but give some indication of any action that may be required, such as further diagnostic investigation, closer monitoring or even preventative action. Screening is not necessarily a benign process (Marteau, 1989). Since screening tools are never 100% accurate, people who are incorrectly identified as being at risk of developing a condition (false positives) can be subject to further possibly intrusive, harmful or inappropriate investigations, management or treatment. Those falsely identified as not being at risk of developing a condition (false negatives) will also suffer by not being given the further investigation they need.
The National Screening Committee (NSC), in its criteria for determining whether a national screening programme should be undertaken for any disorder, has set 22 criteria for appraising the viability, effectiveness and appropriateness of a programme for large population screening. These include: the need for a simple, safe, precise and validated screening test; an agreed policy on the further evaluation of individuals with a positive test result; the availability of an effective treatment for those identified through early detection, with evidence of early treatment leading to better outcomes than later treatment; adequate resources available prior to commencement; and acceptability to the population. It is important that the majority of these criteria are satisfied before a screening programme is adopted, not least because screening can cause adverse effects, including distress secondary to asking specific questions, raising concerns and expectations of care7.
Existing NICE mental health guidelines have considered the case for general population screening for a number of mental health disorders and concluded that screening should only occur for specific high-risk populations where benefits outweigh risks (for example, NICE, 2004a, 2005a). Whether women have a higher prevalence of mental disorder during pregnancy and the postnatal year than at other times is unclear (with the exception of puerperal psychosis) (Brockington, 1996; Gavin et al., 2005). However, mental disorders occurring during pregnancy and the postnatal period may have greater adverse consequences for all concerned than they do at other times (for example, Sharp et al., 1995). There is therefore a need to try to discover whether it is possible to predict which women are likely to develop a disorder, as well as to ensure that existing illness is detected in order to initiate appropriate treatment as quickly as possible.
The guideline uses the terms prediction and detection instead of screening in order to distinguish between the two functions involved in screening. Prediction is used to refer to the identification of risk factors, either current or past, which increase the probability of developing mental disorder or the probability of relapse of a previous mental disorder at some point in the future. These can include hereditary or congenital factors, psychiatric history, early life experiences and current circumstances, as well as current mood and functioning. Detection is used to refer to the identification of current disorder.
Since accurate prediction might lead to treatments to prevent the onset of, or ameliorate the course of, mental disorder, and accurate detection leads to treatments to treat disorder, prediction and detection require different systems and tools. This chapter will therefore attempt to answer two questions: can we successfully predict who may become ill, and can we improve the detection of women who have become ill?
Vignette: A woman with a history of mental health problems diagnosed with depression in the postnatal period
In 2004 when my son was about 5 months old, I knew that I didn’t feel right; I didn’t think I was feeling enough. I was looking after him quite well, I think, keeping the house tidy, cooking dinner and so on, but I felt so flat.
I had experienced a similar health problem many years before, but this was never brought up or discussed by me or my healthcare professionals when I became pregnant. I didn’t think it would be an issue as previously it had been related to my personal circumstances, which were totally different from how they were when I was pregnant. The only thing I did was complete the form that they ask all pregnant women to fill in – a form on which you could easily lie. My ‘score’ suggested that it was highly unlikely that I would suffer from postnatal depression.
5.1.2. Sensitivity and specificity
The terms sensitivity and specificity are used in relation to prediction and detection methods discussed in this chapter.
The sensitivity of an instrument refers to the proportion of those with the condition who test positive. An instrument that detects a low percentage of cases will not be very helpful in determining the numbers of patients who should receive a known effective treatment, as many individuals who should receive the treatment will not do so. This would make for poor planning and underestimating the prevalence of the disorder and the costs of treatments to the community. As the sensitivity of an instrument increases, the number of false negatives it detects will decrease.
The specificity of an instrument refers to the proportion of those without the condition who test negative. This is important so that well individuals are not given treatments they do not need. As the specificity of an instrument increases, the number of false positives will decrease.
To illustrate this: from a population in which the point prevalence rate of depression is 10% (that is, 10% of the population has depression at any one time), 1,000 women are given a test which has 90% sensitivity and 85% specificity. It is known that 100 women in this population have depression, but the test detects only 90 (true positives), leaving 10 undetected (false negatives). It is also known that 900 women do not have depression, and the test correctly identifies 765 of these (true negatives), but classifies 135 incorrectly as having depression (false positives). The positive predictive value of the test (the number correctly identified as having depression as a proportion of positive tests) is 40% (90/90 + 135), and the negative predictive value (the number correctly identified as not having depression as a proportion of negative tests) is 98% (765/765 +10). Therefore, in this example, a positive test result is correct in only 40% of cases, whilst a negative result can be relied upon in 98% of cases.
5.2. PREDICTION – RISK FACTORS FOR THE ONSET OF MENTAL DISORDER DURING PREGNANCY AND THE POSTNATAL PERIOD
5.2.1. Introduction
The mental disorders most associated with pregnancy and the postnatal period are depression and puerperal psychosis, although the point prevalence of the latter is relatively low (around 1 to 2 per 1000 compared with 100 to 150 per 1000 with depression). However, any mental disorder can occur during this time, including eating disorders and anxiety disorders such as PTSD.
Prediction tools are based on the use of risk factors. In order to predict future disorder accurately, it is necessary to know which factors are associated with development of future illness. These can include psychosocial and physical factors, as well as past illness and family history. Once the factors with the highest predictive value are known, a reliable prediction tool needs to be developed, which is usable by busy healthcare professionals in the clinical milieu. This is particularly important with regard to mental health during pregnancy and the postnatal period, since those caring for women during this time need robust skills in looking after the mental health as well as the physical health of women and their infants.
Vignette: A woman with pre-existing depression and depression after the birth of both of her children
Before I first became pregnant, I had been taking an antidepressant for around 3 years for what a consultant psychiatrist termed ‘classic diurnal depression’, which included debilitating symptoms of claustrophobia and agoraphobia. Although in a long-term relationship, I had previously dismissed the notion of having children – being able to look after them and build a loving relationship –because of my depression.
The pregnancy was an extremely happy time. Although I came off the antide-pressant as soon as I discovered I was pregnant, apart from some short-term, unpleasant physical symptoms (dizziness and so on) my mood was positive and I coped well with work and the new demands on my body. Already aware that there was a good chance of becoming depressed once my baby was born, I emphasised my worries on this score right at the beginning of my antenatal care at the GP practice. But, although I raised it time and again, and it was written clearly in my notes, it was never referred to by any healthcare professional.
Vignette: A woman with depression diagnosed 1 year after the birth of her first child and with suspected PTSD from traumatic birth
I began to experience a decline in my mental health from week 8 of my pregnancy. This was due to extreme nausea, insomnia, anxiety attacks and exhaustion. I was unable to work and became increasingly isolated, while receiving poor antenatal midwifery care via my GP’s surgery. I was simply told that, as my family had no history of mental illness, I was in no danger from either antenatal or postnatal depression.
5.2.2. Evidence search
In order to determine whether a particular factor accurately predicts future mental disorder, large-scale prospective studies are required which clearly define the risk factor under question and assess mental health status at an appropriate time point using a well-validated diagnostic tool. It is important to note that studies that use a simple correlational design simply show that there is a link between factor and outcome but can not establish whether the factor plays any causal role in the onset or maintenance of disorder.
The evidence search aimed to identify longitudinal prospective studies of risk factors for depression in the postnatal period, puerperal psychosis, eating disorders and anxiety disorders. High-quality reviews of such studies were also found in the general search for systematic reviews relevant to the guideline (see Chapter 3). Inclusion criteria for individual studies included diagnosis being established by diagnostic interview rather than by self-report tests such as the EPDS. Further details of the search process are in Appendix 6.
Very little data was found for conditions other than depression in the postnatal period. For depression in the postnatal period, three existing reviews were identified (see below), but for other disorders similar rigorous systematic reviews were not identified. Therefore, a search of the literature for primary research studies from 1996 to 2005 was undertaken for depression and from database inception to 2005 for anxiety disorders and eating disorders. Inclusion criteria were widened for the latter search to include studies in which diagnosis was not established by diagnostic interview. A search was also undertaken to identify prediction tools for depression in the antenatal and postnatal period from 2001 onwards. Update searches were undertaken during the remainder of the guideline development process. shows the databases searched.
Databases searched and inclusion/exclusion criteria for studies of risk factors for mental disorder during pregnancy and the postnatal period.
5.2.3. Risk factors for depression in the postnatal period
Existing reviews
Three reviews of risk factors for depression in the postnatal period were identified: Beck (2001), O’Hara and Swain (1996) and Robertson and colleagues (2004). A brief description of each is provided in .
Reviews of risk factors for depression in the postnatal period.
The reviews are difficult to reconcile since there is little overlap between included studies. O’Hara and Swain (1996) and Beck (2001) have only 18 studies in common (out of 59 and 84 respectively). In addition, neither list excluded studies or describe reasons for exclusion. Of the 26 additional studies reviewed by Robertson and colleagues (2004) (which used both the O’Hara and Swain [1996] and Beck [2001] reviews and added new studies), four were already in Beck (2001) and three in O’Hara and Swain (1996). Weaknesses of Beck (2001) are the inclusion of cross-sectional designs and studies not assessing depression according to a diagnostic interview. Similar weaknesses apply to O’Hara and Swain (1996), although a higher proportion of included studies used a clinical interview to assess depression compared with those in Beck (2001) (approximately 50%, compared with 36%). It is not clear whether Robertson and colleagues (2004) excluded studies included by the two previous reviews if they did not meet inclusion criteria, for example, the cross-sectional studies in Beck (2001). lists the risk factors identified by each review.
Risk factors for depression in the postnatal period identified by existing reviews.
Of the identified risk factors, depressed mood or anxiety during pregnancy are the strongest factors associated with developing depression in the postnatal period. Other potentially important risk factors are the level of social support, life events and psychiatric history, including previous experience of depression. With psychiatric history, the level of increased risk appears to be related to the severity and duration of the previous depression. For example, women reporting depression in the postnatal period after their first child that resolved in less than 2 months in the absence of any other risk factors were not at increased risk of depression after their second child (Elliott et al., 2000).
Social support can be defined in various ways. For example, it can be defined in terms of sources of support, such as spouse, friends and relatives, or in terms of the type of support received. This includes informational support, instrumental support (such as practical help) and emotional support (Robertson et al., 2004). Robertson and colleagues (2004) found that both emotional and instrumental support was negatively correlated with depression in the postnatal period, and that perceived social isolation was strongly predictive of depression in the postnatal period.
New studies
Eight additional studies were found that were published since these reviews. Study characteristics are in . These largely support the findings of earlier studies, with vulnerable personality, past history of depression, dissatisfaction with partner relationship, recent life events and poor social support associated with higher depression symptoms. In addition, a systematic review found no link between caesarean section and depression in the postnatal period (Carter et al., 2006).
Characteristics of studies of risk factors for depression published since existing reviews.
Summary of risk factors for the development of depression in the postnatal period
Compared with other disorders, depression in the postnatal period is relatively well studied, and a wide variety of risk factors have been investigated. Those factors consistently associated with the onset of depressive symptoms during the postnatal period include depressed mood and depression during pregnancy, anxiety during pregnancy, poor social support (although this is defined in a number of different ways), recent life events and a history of depression or other psychiatric history.
5.2.4. Risk factors for puerperal psychosis
Puerperal psychosis is a rare event particularly compared with depression in the post-natal period (see Chapter 4). Many commentators consider much puerperal psychosis to be a variant of an episode of bipolar disorder, with a third of episodes meeting criteria for mania or schizoaffective mania (Gelder et al., 2000). In an epidemiological study following 470,000 women over an 8-year period, 21.4% of those with bipolar disorder whose last episode was manic and 13.3% of those with bipolar disorder whose last episode was depressed had a psychiatric admission (a proxy for psychosis) within 3 months of delivery. These proportions were much higher than in those with other psychiatric diagnoses, for example, 3.4% of those with schizophrenia and 1.9% of those with depressive neurosis (all ICD-9 diagnostic categories) (Kendell et al., 1987). In a later study, episodes (defined as a DSM-IV episode of mania or psychotic episode within 6 weeks of delivery) followed 26% of deliveries in women with a diagnosis of bipolar I disorder or schizoaffective disorder (313 deliveries in 152 women) (Jones & Craddock, 2001). It is important to note that these figures do not represent the natural history of the untreated disorder postnatally as the populations studied will have included an unknown proportion of women receiving treatment. Thus these proportions should not be quoted as the relapse risk without preventative treatment, which is likely to be higher and can only be derived from untreated samples. Small control samples of women with bipolar disorder on no medication postnatally indicate that the natural postnatal relapse risk for bipolar disorder may be considerably higher, at up to 70% (that is 700 in 1,000) postnatally (Viguera et al., 2000) and 50% (that is 500 in 1,000) antenatally. Nevertheless, there appears to be a substantial number of women who experience an episode of psychosis in the early postnatal period who do not have an existing bipolar diagnosis and/or who never experience a non-puerperal episode (Dean et al., 1989). In the relatively small study by Dean and colleagues (1989), these women had better overall outcomes than women with a bipolar diagnosis.
A study of the relationship between obstetric factors and onset of puerperal psychosis found 60% of women with puerperal psychosis had a diagnosis of bipolar disorder or schizoaffective disorders (Sharma et al., 2004). However, the risk of recurrence of illness at any time (defined as readmission) after a puerperal psychosis appears to be much greater in women with a diagnosis of schizophrenia, with half of women with this diagnosis followed in a Danish sample relapsing within a year of discharge, compared with half with other diagnoses being readmitted within 2 years (Terp et al., 1999). In all, 98% of women with schizophrenia who had a puerperal psychosis relapsed. Women who had had a puerperal psychosis but who did not have a diagnosis of schizophrenia were also at risk of relapse (ibid.). Within 10 years, 65% of those with no previous admission were readmitted, and 85% of those with a previous admission were readmitted. Specific risk factors for readmission in women with a functional psychosis other than schizophrenia were not being married and preterm delivery (ibid.). A Swedish cohort study of over 600,000 women concluded that most episodes of puerperal psychosis occurred in women with a previous psychotic or bipolar illness, and that the majority of these episodes were of a psychotic disorder rather than an episode of bipolar disorder (Harlow et al., 2007).
Social and obstetric risk factors associated with psychiatric admission (used as a proxy for puerperal psychosis) include not being married, not having living children and perinatal death (index delivery) (Kendell et al., 1987). Family history of puerperal illness may also be a risk factor (for example, Jones & Craddock, 2001), although family history of non-puerperal illness does not appear to be a risk factor (for example, Robertson et al., 2005). Older age at delivery also appears to be a risk factor, with those aged 40 to 44 having five times the risk of those in the general female population compared with around twice the risk in younger women (Nager et al., 2005). This study also found that not living with the infant’s father was a risk factor. In women with bipolar disorder or schizophrenia (that is, those at high risk of puerperal psychosis), links with factors such as increased sensitivity of dopamine receptors in the hypothalamus and sleep loss have also been suggested (Sharma et al., 2004; Wieck et al., 1991), as have primiparity, difficult labour, genetic predisposition and hormonal changes (Brockington, 1996).
Summary of risk factors for the development of psychosis in the postnatal period
Risk factors for the development of puerperal psychosis are a history of previous serious mental disorder, particularly bipolar disorder, and previous puerperal psychosis, schizoaffective disorder and schizophrenia.
5.2.5. Risk factors for developing other disorders
While the amount of literature on risk factors predicting the development of depression in the postnatal period is relatively large, literature for risk factors predicting the development of other mental disorders during this time is lacking. However, there are a few studies and reviews describing risk factors for disorders including PTSD, panic disorder and eating disorders.
Post-traumatic stress disorder
Some work has been done looking at risk factors for symptoms of PTSD following childbirth; for example, a study by Czarnocka and Slade (2000) identified perceptions of low levels of support from partners and staff, patterns of blame and perceived low control in labour as predictive of symptoms. Personal vulnerability factors, such as previous mental health difficulties and trait anxiety, were also related to symptoms. These factors were supported by a review of available studies (Olde et al., 2006). The review found two small studies of PTSD symptoms in women following emergency caesareans, which reported high levels of symptoms.
Panic disorder
There is little systematic research on panic disorder during pregnancy, particularly on risk factors for developing panic disorder during the postnatal period. A review of studies looking at its occurrence (without concurrent affective disorder) during pregnancy and/or the postnatal period found ten studies, all except one of which were retrospective and uncontrolled (Hertzberg & Wahlbeck, 1999). Only one study was solely concerned with onset of panic disorder and the postnatal period (Sholomskas et al., 1993). Overall, a collation of all the studies found that 6% of pregnancies had postnatal onset (278 pregnancies), although in most studies (n = 8) the postnatal period was defined as up to 3 months after delivery. No specific risk factors were identified that predicted the onset of panic disorder.
Generalised anxiety disorder
There appears to be little research on the predictive factors of GAD in pregnancy or the postnatal period, although anxiety itself has been shown to be predictive of depression in the postnatal period (Heron et al., 2004).
Eating disorders
It is uncommon for women to develop an eating disorder de novo during pregnancy and the postnatal period, and women with severe eating disorders tend to have reduced fertility and therefore are less likely to become pregnant. There are also data suggesting that symptoms of existing disorders improve during pregnancy, with 68% of women in a study of 530 women attending an antenatal clinic reporting symptoms above threshold in the 2 years before conception reporting symptoms below threshold in pregnancy (Turton et al., 1999). However, some women continue to experience symptoms during pregnancy, and factors associated with higher symptomatology in pregnancy include younger age (less than 30 years), previous high levels of symptoms, lower educational attainment, poorer housing, employment status and previous miscarriage (ibid.). Women with a history of bulimia nervosa or binge eating disorder may be at increased risk of depression in the postnatal period (Mazzeo et al., 2006).
Obsessive-compulsive disorder
A review of OCD symptoms in pregnancy and the postnatal period found some evidence for onset of OCD associated with pregnancy and childbirth, although studies were of OCD populations and relied on retrospective self-report (Abramowitz et al., 2003). The authors could find no evidence for predictive factors for the development or exacerbation of OCD in the postnatal period. Similarly, they found no data on the possible causal link between OCD and depression in this period, which commonly co-occur.
Summary of risk factors for the development of other mental disorders in the postnatal period
Disorders other than depression and psychosis are much less well studied in pregnancy and the postnatal period. Symptoms of PTSD in the postnatal period may be associated with perceptions of low levels of support from partners and staff, patterns of blame and perceived low control in labour. Exacerbation of eating disorders during pregnancy was associated with younger age (less than 30 years), previous high levels of symptoms, lower educational attainment, poorer housing, employment status and previous miscarriage. It is not clear whether specific risk factors can predict onset of panic disorder or OCD.
5.2.6. Clinical summary of risk factors for the development of mental disorders in the postnatal period
There are a large number of studies looking at risk factors for depression in the postnatal period and some data on risk factors for psychosis and other disorders. The risk factors that consistently show reasonable predictive value, particularly for the development of depression, psychosis and recurrence of bipolar disorder, are past psychiatric history, including previous puerperal episodes, and current disorder or symptomatology. There is also some suggestion that family history of psychosis in the postnatal period is predictive.
5.4. METHODS FOR DETECTING MENTAL DISORDER DURING PREGNANCY AND THE POSTNATAL PERIOD
5.4.1. Detection methods
The evidence considered in Chapter 4 indicates that psychiatric disorders during the postnatal period may carry considerable risk to both the woman and the infant. In light of this, it is important that reliable methods of detection of current mental disorders in the postnatal period are available, as most mental disorders experienced during the antenatal and postnatal period respond to appropriate and timely treatments.
Detection of depression in the antenatal period
There is one study validating the EPDS in pregnancy (Murray & Cox, 1990). This validated the EPDS against RDC diagnoses of depression using a standardised psychiatric interview (Goldberg et al., 1970) in 100 women of between 28 and 34 weeks’ gestation. Six per cent had a diagnosis of major depression and 8% minor depression. At the 12/13 cut-off rate, the EPDS had a sensitivity of 100% for major depression but a specificity of only 87%. Specificity improved at the higher cut-off of 14/15 (96%). For minor depression, the sensitivity was 71% and the specificity 72% at the 10/11 cut-off. Therefore, a higher cut-off may be required to use the EPDS to detect depression in pregnancy.
Detection of depression in the postnatal period
Again, depression in the postnatal period is the area where most work has been done to develop detection methods, including both self-report and clinician-completed measures.
Eight self-report measures that had been assessed in mothers in the first postnatal year were reviewed by Boyd and colleagues (2005), including the BDI and BDI-II (Beck et al., 1961, 1996), the Bromley Postnatal Depression Scale (BPDS; Stein & Van den Akker, 1992), the Center for Epidemiological Studies Depression Scale (CES-D; Radloff, 1977), the EPDS (Cox et al., 1987), the GHQ (Goldberg, 1972), the Inventory of Depressive Symptomatology (IDS; Rush et al., 1986), the Postnatal Depression Screening Scale (PDSS; Beck & Gable, 2000, 2001a) and the Zung Self-Rating Depression Scale (Zung SDS; Zung, 1965).
When describing the sensitivity, specificity and positive predictive value of the different instruments, the review defined ‘excellent’ as values above 0.9, ‘good’ as 0.8 to 0.9, ‘moderate’ as 0.5 to 0.7, ‘low’ as 0.3 to 0.5, and ‘poor’ as less than 0.3. Based on these categorisations, it calculated that the BDI had good specificity, variable sensitivity, depending on the sample used, and low positive predictive value. The BDI-II had excellent specificity, moderate sensitivity and excellent positive predictive value, although this reduces to moderate for minor depression if a prevalence rate of 13% is assumed (based on O’Hara & Swain, 1996).
The review found only a single study of the BPDS that showed moderate sensitivity and positive predictive value and excellent specificity. However, a self-report diagnostic tool was used. The CES-D had moderate sensitivity and positive predictive value and excellent specificity. However, the review concludes that the CES-D may miss 40% of depressed women. The review found that the positive predictive value for the EPDS varied depending on the sample used. In a population with the ‘standard’ prevalence rate of 13%, the positive predictive value was low to moderate, with low specificity. Sensitivity and positive predictive values were low in studies using it to detect minor and/or major depression, rather than just major depression. The review included several studies of non-English language versions of the scale.
Boyd and colleagues (2005) found studies validating the GHQ (various versions) in depression in the postnatal period, although the GHQ does not measure solely depression but rather general psychiatric morbidity. The 12-item version was found to have the highest sensitivity and positive predictive value. The review found that the IDS had excellent sensitivity, good specificity and moderate positive predictive value. The PDSS was found to have excellent sensitivity and specificity, and good positive predictive value.
For most scales, there were only between one and four studies, with the EPDS being the most studied (22 studies, with 11 being used to calculate sensitivity, specificity and positive predictive value). Only three were specifically designed to measure depression in the postnatal period (the BPDS, EPDS and PDSS). Boyd and colleagues (2005) concluded that, while more research was needed, the EPDS had been most widely studied but only had moderate psychometric properties. Other instruments such as the PDSS and the BDI may have value as methods for the detection of depression but further research is needed. The EPDS is considered below in more detail, since this is currently the most widely-used scale.
Another self-report scale in use in primary care is the Patient Health Questionnaire (PHQ; Spitzer et al., 1999). Although it does not appear to have been validated in postnatal women, it seems to have good sensitivity and specificity in detecting depression in primary care populations (Spitzer et al., 1999). A nine-item depression module (PHQ-9) is often used in isolation, for example by GPs, and a two-item version has also been tested and found to have good sensitivity and specificity (Kroenke et al., 2003). This is discussed in more detail below.
Also in use in primary care to assess both depression and anxiety symptoms is the clinician-completed Hospital Anxiety and Depression Scale (Zigmond & Snaith, 1983). Although this does not appear to have been validated in postnatal populations, it appears to have good case-finding properties (Bjelland et al., 2002). However, it was not designed as a detection tool.
5.4.2. The Edinburgh Postnatal Depression Scale (EPDS)
The EPDS (Cox et al., 1987) is a ten-item self-report questionnaire developed to assist healthcare professionals to identify depression in the postnatal period. It was developed in an attempt to address the problem of the pregnancy or postnatal status per se affecting experiences typically taken as indicators of depression, such as disturbances of appetite. It was piloted on childbearing women in Edinburgh. Although the EPDS was originally designed as a screening tool, its use has extended well beyond this in studies of the effectiveness of treatments for depression in the postnatal period. In research settings, the EPDS has at times been applied as though it were a tool to identify those who fall within the diagnostic category of depression and distinguish them from those that do not. The EPDS is strongly correlated with anxiety (Brouwers et al., 2001; Jomeen & Martin, 2005) and, for an unselected sample, the distribution of scores is continuous: it is simply not possible to determine the clear cut-offs that use as a diagnostic instrument would require because depressed mood lies on a continuum of severity (Murray & Carothers, 1990). It has been argued that the EPDS may be conceptualised as a continuous measure of emotional well-being (Green, 2005), although further studies would be required to assess its value as a continuous measure of severity or probability of clinical depression.
In order to assess the psychometric properties of the EPDS, a search was undertaken of validation papers of the English-language version, in which the validation was against diagnosis as established by diagnostic interview and for a sample of women from developed countries. Eight such studies were identified (see )8. In addition, another high-quality review was used to support this review (Gaynes et al., 2005).
Summary of validation studies of EPDS as a detection tool for depression in the postnatal period.
As can be seen from , the psychometric properties of the EPDS (sensitivity, specificity, positive predictive value and negative predictive value) vary considerably across the studies, particularly sensitivity and the positive predictive value, which varies from 33% to 93%. This variability probably reflects differences in the populations sampled and how the EPDS was administered in the studies, with most being research based. Also, the prevalence of depression differed between the studies, ranging from 8.7% (Boyce et al., 1993) to 25% (Cox et al., 1987). This is higher than the 6.8% calculated by Gaynes and colleagues (2005) in their systematic review (for the period up to 6 weeks postnatally, which is the most common time point used to assess depression in the included studies).
Gaynes and colleagues undertook a meta-analysis of three studies detecting major depression in the postnatal period using the EPDS at the ≥13 cut-off point. They removed a further study in order to calculate a meaningful standard error since this study estimated sensitivity of the EPDS at 100%, which is unlikely. They calculated a pooled sensitivity of 91% (95% CI = 0.84 to 0.99) but did not calculate a pooled specificity because of significant heterogeneity. However, when summing up, they report a specificity of 95%, although it is not clear how they calculated this.
Nevertheless, based on these figures, together with a prevalence rate for major depression of 6.8% (which they calculate based on a review of prevalence studies), Gaynes and colleagues (2005) calculate an overall positive predictive value for the EPDS of 57% and a negative predictive value of 99%. The corresponding values for both major and minor depression (using a cut-off point of >10) are a positive predictive value of 30% and negative predictive value of 95%.
These data are difficult to translate into clinical practice. Specific difficulties identified with current practice in some areas are the use of cut-offs outside of the range of 13+ (>12) recommended for probable depression and 10+ (>9) for possible depression, such as use of cut-offs in the 15 to 20 range to indicate urgency of treatment or need to refer to mental health services without other information gathered by interview. Published reports of clinical practice with other cut-offs within the recommended range, typically 12+ (>11), fail to explain the rationale for the choice, usually trading off the risk of false positives and false negatives in the above two cut-offs.
Early research studies of acceptability suggested a very high degree of acceptability to the population in the context of research protocols and pre-selected populations (Murray & Carothers, 1990). Subsequent qualitative studies suggest that acceptability ratings are in fact much lower for some current clinical practices using the EPDS or variants of it: women feel compelled to complete the questionnaire but, owing to fears as to the possible consequences of being identified as symptomatic, may distort their responses accordingly (Shakespeare et al., 2003; Cubison & Munro, 2005).
5.4.3. Case finding with interview questions
Detection for depression in adult populations has moved away from the use of paper-and-pencil multi-item questionnaires (for example, NICE, 2004a). Studies indicate that two brief focused questions that address mood and interest are as likely to be effective as more elaborate methods and are more compatible with routine use in busy primary and secondary care settings (Whooley et al., 1997) (‘During the last month, have you often been bothered by feeling down, depressed or hopeless?’ and ‘During the last month have you often been bothered by having little interest or pleasure in doing things?’). The questions are based on the 2-item PHQ-9 (see above), although in the study by Whooley and colleagues (1997) the questions were not scored but simply required a yes or no answer.
Whooley and colleagues (1997) found that the two questions have a sensitivity of 96% (95% CI 90% to 99%) and a specificity of 57% (95% CI 53% to 62%), giving a positive likelihood ratio of 2.2. Arroll and colleagues (2005) have developed an extension to these two questions by adding the following question: ‘Is this something with which you would like help?’, with three possible responses: ‘No’, ‘Yes, but not today’ and ‘Yes’. The two Whooley questions plus the help question have been validated against a standardised psychiatric interview and the addition of the help question resulted in a sensitivity of 96% (95% CI 86% to 99%) and an improved specificity of 89% (95% CI 87% to 91%), with a positive likelihood ratio for the help question of 9.1. Based on calculations using data given in the paper, the three questions have a positive predictive value of 32% and a negative predictive value of 99%. It is not clear in the paper whether the diagnostic criteria used to validate the three questions (and the other instruments tested) included both minor and major depression or major depression alone.
5.4.4. Clinical summary for methods for detecting mental disorder in the postnatal period
Although a number of tools (essentially self-report questionnaires) have been developed for the detection of depression, only eight have been found with studies assessing their use in the postnatal period. Only one of these, the EPDS, has been the subject of a sufficient number studies to make a judgement of its usefulness. However, this dataset has a number of problems, including relatively high prevalence of depression amongst the included studies compared with that calculated by a high-quality review of prevalence studies, relatively small studies (the majority had fewer than 250 women, although one assessed around 1,500 women) and many studies undertaken in a research rather than a clinical setting. Given this, although the sensitivity of the test is reasonably high, the lower specificity means that the positive predictive value is poor and that, although the reliability of a negative test result is good, that for a positive test is poor, and would mean that nearly half of all women referred for further assessment would be referred unnecessarily, placing an increased and wasteful burden on resources. Similarly, the two Whooley questions plus the additional Arrol question have poor positive predictive value.
The NSC in its review of screening for depression in the postnatal period also found insufficient evidence to support a national screening programme based on any of the existing screening tools, including the EPDS (Shakespeare, 2001). It should be noted that the review’s criteria are much broader than an assessment of psychometric properties as they attempt to validate screening tools for use within a healthcare system rather than just whether the tool is effective in detecting a disorder; for example, ‘there must be evidence from high-quality RCTs that the screening programme is effective in reducing mortality or morbidity’. The two Whooley questions were not included in this review, although they probably would not meet the criteria either.
Current NICE guidelines for depression (NICE, 2004a) recommend the two questions and, although little specific evidence exists for their use in the perinatal period, their ease of use and reasonable sensitivity and specificity, particularly if combined with the additional help question from Arroll and colleagues (2005), suggest that their use in routine care may be practical and acceptable; for example, they do not require additional resources (such as copies of a questionnaire). The value of the questions lies in part in their brevity and the fact that they lend themselves to use both in the antenatal and postnatal periods.
5.4.5. Clinical practice recommendations
- 5.4.5.1.
At a woman’s first contact with primary care, at her booking visit and post-natally (usually at 4 to 6 weeks and 3 to 4 months), healthcare professionals (including midwives, obstetricians, health visitors and GPs) should ask two questions to identify possible depression.
During the past month have you often been bothered by feeling down, depressed or hopeless?
During the past month, have you often been bothered by having little interest or pleasure in doing things?
A third question should be considered if the woman answers ‘yes’ to either of the initial questions:
Is this something you feel you need or want help with?
- 5.4.5.2.
Healthcare professionals may consider the use of self-report measures such as the EPDS, HADS or PHQ-9 as part of subsequent assessment or for the routine monitoring of outcomes.
- 5.4.5.3.
If a woman has a current mental disorder or a history of severe mental illness, she should be asked about her mental health at all subsequent contacts.
5.4.6. Research recommendation
Case finding for depression
A validation study should be undertaken of the ‘Whooley questions’ (‘During the past month, have you often been bothered by feeling down, depressed or hopeless?’ ‘During the past month, have you often been bothered by having little interest or pleasure in doing things?’) in women in the first postnatal year, examining the questions’ effectiveness when used by midwives and health visitors compared with a psychiatric interview.
Why this is important
Depression in the first postnatal year is relatively common and may have a lasting impact on the woman, her baby and other family members. Case finding is most conveniently undertaken by healthcare professionals in regular contact with women, but they do not traditionally have training in mental health. The Whooley questions appear to offer a relatively quick and convenient way of case finding for healthcare professionals who are not specialists in mental health.
5.5. REFERRAL PATHWAYS
If initial detection of a disorder is to be useful, clear referral pathways supported by effective treatment options need to be available. All healthcare professionals involved in detecting mental illness in women in the antenatal and postnatal periods must be aware of the appropriate care and referral options, so that effective assessment and treatment is available to those who are identified as requiring further assessment. The nature of the assessment and any subsequent treatment will vary between primary and secondary care services and will reflect the nature of the potential disorder identified.
For the large majority of women with identified or suspected mental disorder, the major source of effective assessment and treatment will take place in primary care, coordinated by the GP. Where a women is identified within primary care as suffering from a common mental disorder, the further treatment and assessment of the disorder will be managed and coordinated in primary care according to the protocols in that practice. As a minimum, it is expected that a record of the disorder should be entered in the woman’s notes and further appropriate assessment or monitoring undertaken. The extent of this assessment and monitoring will vary with the severity of the disorder and the presence of any potential risk factors.
If the initial detection of a common mental disorder occurs in secondary care maternity services, again the minimum should be an entry in the notes and a communication with the GP of the initial outcome of the assessment. Depending on the severity of the disorder, healthcare professionals should consider further assessment of the woman’s psychological state, where appropriate seeking advice from colleagues on the details of the further assessment to be undertaken. Where the methods for the detection of mental disorder identify a history of, or the presence of a severe mental disorder, healthcare professionals in primary care or maternity services should refer the woman for a specialist mental health assessment, given the high risk of relapse or exacerbation of symptoms for some women with a history of severe mental illness and the risks associated with the presence of these conditions.
5.5.1. Clinical practice recommendations
- 5.5.1.1.
A written care plan covering pregnancy, delivery and the postnatal period should be developed for pregnant women with a current or past history of severe mental illness, usually in the first trimester. It should:
be developed in collaboration with the woman and her partner, family and carers, and relevant healthcare professionals
include increased contact with specialist mental health services (including, if appropriate, specialist perinatal mental health services)
be recorded in all versions of the woman’s notes (her own records and maternity, primary care and mental health notes) and communicated to the woman and all relevant healthcare professionals.
- 5.5.1.2.
In all communications (including initial referral) with maternity services, healthcare professionals should include information on any relevant history of mental disorder.
- 5.5.1.3.
After identifying a possible mental disorder in a woman during pregnancy or the postnatal period, further assessment should be considered, in consultation with colleagues if necessary.
If the healthcare professional or the woman has significant concerns, the woman should normally be referred for further assessment to her GP.
If the woman has, or is suspected to have, a severe mental illness (for example, bipolar disorder or schizophrenia), she should be referred to a specialist mental health service, including, if appropriate, a specialist perinatal mental health service. This should be discussed with the woman and preferably with her GP.
The woman’s GP should be informed in all cases in which a possible current mental disorder or a history of significant mental disorder is detected, even if no further assessment or referral is made.
- 5.5.1.4.
Managers and senior healthcare professionals responsible for perinatal mental health services (including those working in maternity and primary care services) should ensure that:
there are clearly specified care pathways so that all primary and secondary healthcare professionals involved in the care of women during pregnancy and the postnatal period know how to access assessment and treatment
staff have supervision and training, covering mental disorders, assessment methods and referral routes, to allow them to follow the care pathways.
