- Email: [email protected]
Screening women for high risk of postnatal depression
Journal ~fPsychomatic Research, Vol. 38, No. 6. pp. 539 545. 1994 Copyrighl ,c') 1994 Elsevier Science Ltd Printed in Great Britain, All rights reserved 0022-3999/94 $7.00 + 00
Pergamon 0022-3999(93)E0024-K
S C R E E N I N G W O M E N F O R H I G H RISK OF P O S T N A T A L DEPRESSION Louis APPLEBY,* ALAIN GREGOIRE,~ CHRISTINE PLATZ,~ MARTIN PRINCE§ a n d R. KUMAR§ (Received l0 May 1993; accepted in revisedj))rm 23 De~ember 1993)
Abstract A ten-item screening questionnaire was constructed from previous reports on risk factors for postnatal depression, and its ability to predict antenatally the development of postnatal depression was tested. Women attending an antenatal clinic at 36 wk gestation completed the questionnaire and, at 8 wk postpartum, were assessed for the presence of depression using the Edinburgh Postnatal Depression Scale (EPDS). Although antenatal questionnaire scores correlated significantlywith postnatal EPDS scores, this was largely because the questionnaire was able to identify correctly those who would not become depressed. Neither the questionnaire as a whole, nor groups of items, was able to discriminate well between women who later did or did not become depressed. However, women who reported previous or current treatment for depression were at a threefold greater risk of becoming or remaining depressed postnatally. Possible reasons for the negative results are discussed, including the heterogeneity of depression occurring in the postnatal period.
INTRODUCTION NON-PSYCHOTIC depression affects 10-15% of w o m e n who have recently given birth [1-4]. A l t h o u g h its aetiology is likely to be heterogeneous, its occurrence has been consistently shown to correlate with psychiatric a n d psychosocial features, m a n y of which can be detected a n t e n a t a l l y , e.g. previous history of p o s t n a t a l or other depression, p o o r marital a n d social support, anxiety d u r i n g pregnancy, thoughts of t e r m i n a t i o n , p o o r social a n d e c o n o m i c circumstances [3-6]. A p p l e b y et al. [7] used these risk factors clinically as the basis of a psychiatric liaison service to an obstetric unit, to identify a n t e n a t a l l y high-risk subjects who were then followed up into the p u e r p e r i u m to provide s u p p o r t a n d if necessary early t r e a t m e n t o f depression. The presence or absence of risk factors was recorded at a n t e n a t a l b o o k i n g b u t the highest rate of referral occurred much later, d u r i n g the first postnatal week, suggesting that m a n y high-risk individuals were being missed or failing to attend d u r i n g pregnancy. This study did n o t evaluate its clinical screening m e t h o d n o r relate a n t e n a t a l detection to outcome. H o l d e n et al. [8] d e m o n s t r a t e d that postnatal depression can be successfully treated
* Department of Psychiatry, University Hospital of South Manchester, West Didsbury, Manchester M20 8LR, U.K. tThe Old Manor Hospital, Wilton Road, Salisbury SP2 7EP, U.K. +Basingstoke District Hospital, Park Prewett, Basingstoke RG24 9EZ, U.K. §Department of Psychiatry, Institute of Psychiatry, De Crespigny Park, Denmark Hill, London SE5 8AF, U.K. 539
540
L. APPLEBY et al.
by psychological intervention once D e p r e s s i o n S c a l e [9]. T h e p r e s e n t correlates of postnatal depression women who are currently pregnant intervention.
it h a s b e e n d e t e c t e d b y t h e E d i n b u r g h P o s t n a t a l study tests the hypothesis that the psychosocial c a n b e u s e d t o p r e d i c t its l a t e r d e v e l o p m e n t in a n d c o u l d t h e r e f o r e f o r m t h e b a s i s o f a preventive
METHOD Subjects Women who were attending a weekly antenatal clinic at King's College Hospital, a large teaching hospital in south-east London, were asked to participate in the study. The sample consisted of a consecutive series of women attending the clinic at approximatley 36 wk gestation over a period of 6 months.
Measurements At 36wk all subjects completed a screening questionnaire (Table I) and the Edinburgh Postnatal Depression Scale (EPDS). The screening questionnaire was constructed for the present study. Its ten items were taken from research findings on the psychosocial risk factors for postnatal depression [3], and covered psychiatric history, antenatal worries, thoughts of termination, social stress and confiding support. Each item was followed by a choice of responses, in most cases giving a score of between 0 and 3. In two items, the choice was 'no' or 'yes', giving a score of 0 or 1, Approximately 8 wk after delivery, the EPDS was sent to each subject and completed again. If necessary, one reminder letter was sent and telephone contact made. The use of the EPDS in research has been wider than its original purpose of detecting possible cases of postnatal depression. | n this study it was given both to identify psychiatric morbidity postnatally and also to test its value as an antenatal screening instrument. It is a ten-item questionnaire enquiring about depressive symptoms, each item scoring 0 3 according to severity. The threshold total score for significant morbidity is usually taken to be in the region of 10 12. Statistical analysis Statistical analysis was carried out using SPSS-PC +. The internal consistency of both questionnaires was calculated using Cronbach's standardized item alpha correlation coefficient. The relationship between antenatal screening scores and postnatal EPDS scores was analysed by multiple regression, Discriminant function analysis tested the ability of antenatal screening scores to predict 'cases' identified by the postnatal EPDS. RESULTS O f 178 w o m e n r e c r u i t e d t o t h e s t u d y , 165 c o m p l e t e d a n t e n a t a l q u e s t i o n n a i r e s , a n d o f t h e s e i 26 ( 7 7 % ) r e t u r n e d t h e p o s t n a t a l q u e s t i o n n a i r e . S i x t e e n s u b j e c t s s c o r e d 12 o r a b o v e o n t h e p o s t n a t a l E P D S , a r a t e o f 1 3 % , a f i g u r e in k e e p i n g w i t h o t h e r s t u d i e s o f
TABLE |.--ANTENATAL SCREENINGQUESTIONNAIRE 1. Do you have any worries about your own or your baby's health in this pregnancy? No--not at all Yes--worry occasionally Yes--worry a lot Yes can't stop worrying (Score 0--3) 2. Are you currently receiving any treatment for psychological problems'? No No--but such treatment might be helpful Yes from a GP or other therapist or as a psychiatric out-patient Yes as a psychiatric in-patient (Score 0-3)
Screening for postnatal depression TABLE I.
541
Continued
3. Have you ever suffered from postnatal depression? N o - - n o t applicable, having first baby No never Y e s - - b u t not treated Yes--treated by G P or other therapist or as a psychiatric out-patient Yes--treated as a psychiatric in-patient
(Score O, O, 1, 2, 3) 4. Have you ever suffered from psychological problems that were not related to childbirth? No never Yes but not treated Yes treated by G P or other therapist or as a psychiatric out-patient Yes--treated as a psychiatric in-patient
(Score 0 3) 5. Has any close relative (grandparent, parent, brother, sister) had a course of psychiatric treatment either as an out-patient or as an in-patient? No Yes
(Score 0-1) 6. Have you at any time seriously considered not going through with this pregnancy (have you thought of asking for an abortion)? No Yes
(Score 0-1) 7. How is the relationship between you and your partner? Very good Reasonable Poor Very poor No current partner
(Score O, 1, 2, 3, 3) 8. Do you have a relative or friend in whom you can confide and who can give you emotional and practical support? Yes definitely Yes up to a point N o - - n o t really N o - - n o t at all
(Score 0 3 ) 9. All in all, is your day to day life comfortable and settled from a practical and financial point of view? Yes definitely Yes reasonably so N o - - t h e r e are some big problems N o - - t h e problems are enormous
(Score 0-3) 10. Do you think, after your baby is born, that you stand any risk of suffering from postnatal depression? No--definitely not No--unlikely Yes quite possible Yes--highly likely
(Score 0-3)
542 TABLE 1I.
Component
k. APPLEBY et al. SCREENINGQUESTIONNAIRECOMPONENTS AND CORRELATIONSWITH POSTNATAL E P D S SCORES
Items from questionnaire
l l, 2, 4, 6, l0 2 7, 8, 9 3 3, 5 Total Screening questionnaire Antenatal EPDS
Correlation with postnatal EPDS
l-tailed significance
0.36 0.02 0.10 0.24 0.32
p < 0.001 NS NS p < 0.01 p < 0.001
postnatal depression [1-4], although EPDS identifies possible rather than confirmed cases. There was a non-significant tendency for those who did not return the postnatal EPDS to have higher antenatal EPDS scores--mean 9.08 (SD 6.08) compared to mean 7.10 (so 4.39). The standardized item alpha correlation coefficients for each of the scales were: antenatal EPDS 0,82; screening questionnaire 0.65; postnatal EPDS 0.88. Principal components analysis of the screening questionnaire identified three components. These are described in Table II. Table II also shows the correlations between the postnatal EPDS and these components, and with the total scores on the screening questionnaire and the antenatal EPDS. There were significant correlations between certain screening questionnaire scores and subsequent depression, women who were either currently receiving treatment for depression or who had received such treatment in the past (treatment from general practitioner at least on screening questionnaire items 2, 3 and 4) were three times more likely to be depressed postnatally (EPDS score 12 or above) than women who had not received any such treatment (30.8% compared to 10.6%, Z2 = 4.27, p < 0.05). Further analysis showed the predictive ability of the screening questionnaire to be severely limited. Multiple regression analyses with postnatal EPDS score as the dependent variable showed a weak relationship to antenatal scores, the total score on the screening questionnaire accounting for only 6% of the variance. The most successful result was achieved by component 1 of the screening questionnaire, composed of five items (numbers 1, 2, 4, 6, 10), which accounted for 15% of the variance. By an equivalent analysis, the antenatal EPDS score was shown to account for 11% of the variance. Discriminant function analysis led to similarly disappointing results. Entering both the antenatal EPDS and the screening questionnaire scores in a stepwise fashion allowed correct classification of subjects into "cases' (postnatal EPDS = 12 or above) and 'non-cases' in 62% of cases. Of the sixteen who were depressed postnatally, only seven (44%) were predicted. The best overall prediction was achieved when the individual items from both questionnaires were entered in a stepwise fashion, giving a figure of 89% correctly classified. However, this was explained by a high percentage of true negatives (97%), i.e. the successful prediction of those who would n o t become depressed. Of the sixteen who were depressed postnatally, thirteen (81%) were incorrectly predicted as non-cases. The figures were not improved by using alternative scores on the postnatal EPDS as the threshold for "caseness'.
Screening for postnatal depression
543
DISCUSSION This study was an attempt to formalize part of the clinical activity described by Appleby et al. [7] who in an obstetric liaison service identified women thought to be at risk of developing postnatal depression, based on the presence of risk factors reported in the psychiatric literature. The screening questionnaire used here was constructed from the same risk factors and its ability to predict cases of postnatal depression was tested. The principal result was negative. Although significant correlations were found between the postnatal EPDS score and both the total score on the screening questionnaire and the score on a group of items identified by principal components analysis, the predictive ability of the questionnaire was unsatisfactory, even when taken in combination with scores on the antenatal EPDS. The three items covering past or present depression were together able to identify a group of women with a 30% risk of depression but the predictive ability of individual items was also poor. Both EPDS and the screening questionnaires taken together could successfully predict those who did not become depressed postnatally, in that there were few false positives, but the rate of false negatives was unacceptably high. The screening questionnaire could not be used either in clinical practice or in research into preventive intervention on the basis of this evidence. Similarly, these results do not support the use of the EPDS in antenatal screening for risk of postnatal depression. What are the possible reasons for the poor prediction by a questionnaire based on well established risk factors? Explanations may lie in the questionnaire itself, in the sampling method, or in the assumptions about postnatal depression that underlie the study. The screening questionnaire was not assessed for test retest reliability, but internal reliability was satisfactory. Being composed of direct questions concerning accepted risk factors, it carried reasonable face validity. Although all items were given the same weighting in the overall score, its predictive ability was not improved when groups of items obtained from principal components analysis were entered into the analysis separately or in a stepwise fashion. Therefore, despite some potential flaws in its construction, it would be hard simply to attribute the failure in prediction to the form of the questionnaire. The study design required subjects to be attending for hospital-based antenatal care at 36 wk gestation and to return the postal EPDS 8 wk after delivery. It could be argued that some women at high risk may have been omitted because of nonattendance antenatally or inability to respond at 8 wk resulting from depression itself. As already described, there was a slight tendency for those who did not respond postnatally to have scored more highly on the antenatal EPDS. Nevertheless, the overall response rate postnatally (77%) was satisfactory and the rate of postnatal depression identified by the EPDS was in the expected range at 13%. A third possibility is that the assumptions in this study about the applicability of risk factors to different populations are wrong. The risk factors on which the questionnaire was based came largely from K u m a r and Robson [3] whose subjects were women in their first pregnancy attending an antenatal clinic in a north London hospital serving a predominantly middle class area. Potential risk factors were recorded during the first trimester. Subjects in our study were both primiparous and multiparous, were assessed towards the end of pregnancy and were attending King's
544
L. APPLEBY et al.
College Hospital in London, which serves a predominantly poor inner city area. It is possible that the putative risk factors did not transfer well from one population to the other. If this was so, it would support a clinical impression that depression found postnatally is not a homogeneous phenomenon. Although there is no clear dividing line between subtypes of the disorder, there may be groups of patients who differ according to the relative importance of cognitive, social and biological risk factors. In the predominantly social risk group, multiple social problems such as poor housing may be made more stressful by the additional demands of a new baby. In the predominantly cognitive risk group, perfectionist and self-critical attitudes may be amplified by the responsibility of child care. Although similar risk factors have been reported by several studies, differences have also been found such as in the relationship between antenatal mood disturbance and postnatal depression [3, 4]. This may reflect the fact that any socially vulnerable group may also be depressed during pregnancy when social adversity will have almost the same impact it does postnatally, unlike any cognitively vulnerable group whose vulnerability is more likely to be expressed after delivery. Heterogeneity of subject samples, with varying proportions of these groups in different studies, may have led to some inconsistencies in the findings on risk factors and may have contributed to our disappointing result. It is possible that some items of our questionnaire, being derived from previous studies, were less likely to predict depression in our inner city subject sample with its predominantly social vulnerability. Although mainly negative, the results of this study should not deter other researchers from further attempts at developing clinically useful ways of predicting this common and important illness, which can have a long-term impact on maternal health [3, 10] and infant development [11-13]. However, the possible reasons for relatively poor prediction should be remembered in the construction of any future screening questionnaire. Acknowledgement--The authors would like to thank Dr Maureen Marks for her comments on an earlier draft of this paper. Dr Graham Dunn provided statistical advice.
REFERENCES 1. PITT B. Atypical depression following childbirth. Br J Psychiatry 1968; 114" 1225 1235. 2. Cox JL, CONNOR Y, KENDELL RE. Prospective study of the psychiatric disorders of childbirth. Br J Psychiatry 1982; 140:111 117. 3. KUMAR R, ROBSON KM. A prospective study of emotional disorders in childbearing women. Br J Psychiatry 1984; 144: 35-47. 4. WATSON JP, ELLIOTT SA, RUGG AJ, BROUGH D1. Psychiatric disorder in pregnancy and the first postnatal year. Br J Psychiatry 1984; 144: 453-462. 5. STEIN A, COOPER PJ, CAMPBELL EA, DAY A, ALTHAM PME. Social adversity and perinatal complications: their relation to postnatal depression. Br M e d J 1989; 288: 1073-1074. 6. O'HARA MW and ZE~OSKI EM. Postpartum depression: a comprehensive review. In Motherhood and Mental Illness 2. Causes and Consequences (Edited by KUMAR R, BROC~INGTON IF). London: Wright, 1988. 7. APPLEBY L, FOX H, SHAW M, KUMAR R. The psychiatrist in the obstetric unit: establishing a liaison service. Br J Psychiatry 1989; 154: 510-515. 8. HOLDENJM, SAGOVSKYR, COX JL. Counselling in a general practice setting: controlled study of health visitor intervention in treatment of postnatal depression. Br M e d J 1989; 208: 223-226. 9. Cox JL, HOLDEN JM, SAGOVSKYR. Detection of postnatal depression: development of the Edinburgh postnatal depression scale. Br J Psychiatry 1987; 150:782 786.
Screening for postnatal depression
545
10. NOTT PN. Extent, timing and persistence of emotional disorders following childbirth. Br J Psychiatry 1987; 151: 523-527. 11. WRATERM, ROONEYAC, Tr~OMASPF, Cox JL. Postnatal depression and child development: a 3-year follow-up study. Br J Psychiatry 1985; 146: 622-627. 12. COGILLSR, CAPLANHL. ALEXANDRAH, ROBSONKM. Impact of maternal depression on the cognitive development of young children. Br M e d J 1986; 292:1165 1167. 13. STEIN A, GATH DH, BUCHER J, BOND A, DAY A, COOPER PJ. The relationship between postnatal depression and mother-child interaction. Br J Psychiatry 1991; 158: 4(~52.
Pergamon 0022-3999(93)E0024-K
S C R E E N I N G W O M E N F O R H I G H RISK OF P O S T N A T A L DEPRESSION Louis APPLEBY,* ALAIN GREGOIRE,~ CHRISTINE PLATZ,~ MARTIN PRINCE§ a n d R. KUMAR§ (Received l0 May 1993; accepted in revisedj))rm 23 De~ember 1993)
Abstract A ten-item screening questionnaire was constructed from previous reports on risk factors for postnatal depression, and its ability to predict antenatally the development of postnatal depression was tested. Women attending an antenatal clinic at 36 wk gestation completed the questionnaire and, at 8 wk postpartum, were assessed for the presence of depression using the Edinburgh Postnatal Depression Scale (EPDS). Although antenatal questionnaire scores correlated significantlywith postnatal EPDS scores, this was largely because the questionnaire was able to identify correctly those who would not become depressed. Neither the questionnaire as a whole, nor groups of items, was able to discriminate well between women who later did or did not become depressed. However, women who reported previous or current treatment for depression were at a threefold greater risk of becoming or remaining depressed postnatally. Possible reasons for the negative results are discussed, including the heterogeneity of depression occurring in the postnatal period.
INTRODUCTION NON-PSYCHOTIC depression affects 10-15% of w o m e n who have recently given birth [1-4]. A l t h o u g h its aetiology is likely to be heterogeneous, its occurrence has been consistently shown to correlate with psychiatric a n d psychosocial features, m a n y of which can be detected a n t e n a t a l l y , e.g. previous history of p o s t n a t a l or other depression, p o o r marital a n d social support, anxiety d u r i n g pregnancy, thoughts of t e r m i n a t i o n , p o o r social a n d e c o n o m i c circumstances [3-6]. A p p l e b y et al. [7] used these risk factors clinically as the basis of a psychiatric liaison service to an obstetric unit, to identify a n t e n a t a l l y high-risk subjects who were then followed up into the p u e r p e r i u m to provide s u p p o r t a n d if necessary early t r e a t m e n t o f depression. The presence or absence of risk factors was recorded at a n t e n a t a l b o o k i n g b u t the highest rate of referral occurred much later, d u r i n g the first postnatal week, suggesting that m a n y high-risk individuals were being missed or failing to attend d u r i n g pregnancy. This study did n o t evaluate its clinical screening m e t h o d n o r relate a n t e n a t a l detection to outcome. H o l d e n et al. [8] d e m o n s t r a t e d that postnatal depression can be successfully treated
* Department of Psychiatry, University Hospital of South Manchester, West Didsbury, Manchester M20 8LR, U.K. tThe Old Manor Hospital, Wilton Road, Salisbury SP2 7EP, U.K. +Basingstoke District Hospital, Park Prewett, Basingstoke RG24 9EZ, U.K. §Department of Psychiatry, Institute of Psychiatry, De Crespigny Park, Denmark Hill, London SE5 8AF, U.K. 539
540
L. APPLEBY et al.
by psychological intervention once D e p r e s s i o n S c a l e [9]. T h e p r e s e n t correlates of postnatal depression women who are currently pregnant intervention.
it h a s b e e n d e t e c t e d b y t h e E d i n b u r g h P o s t n a t a l study tests the hypothesis that the psychosocial c a n b e u s e d t o p r e d i c t its l a t e r d e v e l o p m e n t in a n d c o u l d t h e r e f o r e f o r m t h e b a s i s o f a preventive
METHOD Subjects Women who were attending a weekly antenatal clinic at King's College Hospital, a large teaching hospital in south-east London, were asked to participate in the study. The sample consisted of a consecutive series of women attending the clinic at approximatley 36 wk gestation over a period of 6 months.
Measurements At 36wk all subjects completed a screening questionnaire (Table I) and the Edinburgh Postnatal Depression Scale (EPDS). The screening questionnaire was constructed for the present study. Its ten items were taken from research findings on the psychosocial risk factors for postnatal depression [3], and covered psychiatric history, antenatal worries, thoughts of termination, social stress and confiding support. Each item was followed by a choice of responses, in most cases giving a score of between 0 and 3. In two items, the choice was 'no' or 'yes', giving a score of 0 or 1, Approximately 8 wk after delivery, the EPDS was sent to each subject and completed again. If necessary, one reminder letter was sent and telephone contact made. The use of the EPDS in research has been wider than its original purpose of detecting possible cases of postnatal depression. | n this study it was given both to identify psychiatric morbidity postnatally and also to test its value as an antenatal screening instrument. It is a ten-item questionnaire enquiring about depressive symptoms, each item scoring 0 3 according to severity. The threshold total score for significant morbidity is usually taken to be in the region of 10 12. Statistical analysis Statistical analysis was carried out using SPSS-PC +. The internal consistency of both questionnaires was calculated using Cronbach's standardized item alpha correlation coefficient. The relationship between antenatal screening scores and postnatal EPDS scores was analysed by multiple regression, Discriminant function analysis tested the ability of antenatal screening scores to predict 'cases' identified by the postnatal EPDS. RESULTS O f 178 w o m e n r e c r u i t e d t o t h e s t u d y , 165 c o m p l e t e d a n t e n a t a l q u e s t i o n n a i r e s , a n d o f t h e s e i 26 ( 7 7 % ) r e t u r n e d t h e p o s t n a t a l q u e s t i o n n a i r e . S i x t e e n s u b j e c t s s c o r e d 12 o r a b o v e o n t h e p o s t n a t a l E P D S , a r a t e o f 1 3 % , a f i g u r e in k e e p i n g w i t h o t h e r s t u d i e s o f
TABLE |.--ANTENATAL SCREENINGQUESTIONNAIRE 1. Do you have any worries about your own or your baby's health in this pregnancy? No--not at all Yes--worry occasionally Yes--worry a lot Yes can't stop worrying (Score 0--3) 2. Are you currently receiving any treatment for psychological problems'? No No--but such treatment might be helpful Yes from a GP or other therapist or as a psychiatric out-patient Yes as a psychiatric in-patient (Score 0-3)
Screening for postnatal depression TABLE I.
541
Continued
3. Have you ever suffered from postnatal depression? N o - - n o t applicable, having first baby No never Y e s - - b u t not treated Yes--treated by G P or other therapist or as a psychiatric out-patient Yes--treated as a psychiatric in-patient
(Score O, O, 1, 2, 3) 4. Have you ever suffered from psychological problems that were not related to childbirth? No never Yes but not treated Yes treated by G P or other therapist or as a psychiatric out-patient Yes--treated as a psychiatric in-patient
(Score 0 3) 5. Has any close relative (grandparent, parent, brother, sister) had a course of psychiatric treatment either as an out-patient or as an in-patient? No Yes
(Score 0-1) 6. Have you at any time seriously considered not going through with this pregnancy (have you thought of asking for an abortion)? No Yes
(Score 0-1) 7. How is the relationship between you and your partner? Very good Reasonable Poor Very poor No current partner
(Score O, 1, 2, 3, 3) 8. Do you have a relative or friend in whom you can confide and who can give you emotional and practical support? Yes definitely Yes up to a point N o - - n o t really N o - - n o t at all
(Score 0 3 ) 9. All in all, is your day to day life comfortable and settled from a practical and financial point of view? Yes definitely Yes reasonably so N o - - t h e r e are some big problems N o - - t h e problems are enormous
(Score 0-3) 10. Do you think, after your baby is born, that you stand any risk of suffering from postnatal depression? No--definitely not No--unlikely Yes quite possible Yes--highly likely
(Score 0-3)
542 TABLE 1I.
Component
k. APPLEBY et al. SCREENINGQUESTIONNAIRECOMPONENTS AND CORRELATIONSWITH POSTNATAL E P D S SCORES
Items from questionnaire
l l, 2, 4, 6, l0 2 7, 8, 9 3 3, 5 Total Screening questionnaire Antenatal EPDS
Correlation with postnatal EPDS
l-tailed significance
0.36 0.02 0.10 0.24 0.32
p < 0.001 NS NS p < 0.01 p < 0.001
postnatal depression [1-4], although EPDS identifies possible rather than confirmed cases. There was a non-significant tendency for those who did not return the postnatal EPDS to have higher antenatal EPDS scores--mean 9.08 (SD 6.08) compared to mean 7.10 (so 4.39). The standardized item alpha correlation coefficients for each of the scales were: antenatal EPDS 0,82; screening questionnaire 0.65; postnatal EPDS 0.88. Principal components analysis of the screening questionnaire identified three components. These are described in Table II. Table II also shows the correlations between the postnatal EPDS and these components, and with the total scores on the screening questionnaire and the antenatal EPDS. There were significant correlations between certain screening questionnaire scores and subsequent depression, women who were either currently receiving treatment for depression or who had received such treatment in the past (treatment from general practitioner at least on screening questionnaire items 2, 3 and 4) were three times more likely to be depressed postnatally (EPDS score 12 or above) than women who had not received any such treatment (30.8% compared to 10.6%, Z2 = 4.27, p < 0.05). Further analysis showed the predictive ability of the screening questionnaire to be severely limited. Multiple regression analyses with postnatal EPDS score as the dependent variable showed a weak relationship to antenatal scores, the total score on the screening questionnaire accounting for only 6% of the variance. The most successful result was achieved by component 1 of the screening questionnaire, composed of five items (numbers 1, 2, 4, 6, 10), which accounted for 15% of the variance. By an equivalent analysis, the antenatal EPDS score was shown to account for 11% of the variance. Discriminant function analysis led to similarly disappointing results. Entering both the antenatal EPDS and the screening questionnaire scores in a stepwise fashion allowed correct classification of subjects into "cases' (postnatal EPDS = 12 or above) and 'non-cases' in 62% of cases. Of the sixteen who were depressed postnatally, only seven (44%) were predicted. The best overall prediction was achieved when the individual items from both questionnaires were entered in a stepwise fashion, giving a figure of 89% correctly classified. However, this was explained by a high percentage of true negatives (97%), i.e. the successful prediction of those who would n o t become depressed. Of the sixteen who were depressed postnatally, thirteen (81%) were incorrectly predicted as non-cases. The figures were not improved by using alternative scores on the postnatal EPDS as the threshold for "caseness'.
Screening for postnatal depression
543
DISCUSSION This study was an attempt to formalize part of the clinical activity described by Appleby et al. [7] who in an obstetric liaison service identified women thought to be at risk of developing postnatal depression, based on the presence of risk factors reported in the psychiatric literature. The screening questionnaire used here was constructed from the same risk factors and its ability to predict cases of postnatal depression was tested. The principal result was negative. Although significant correlations were found between the postnatal EPDS score and both the total score on the screening questionnaire and the score on a group of items identified by principal components analysis, the predictive ability of the questionnaire was unsatisfactory, even when taken in combination with scores on the antenatal EPDS. The three items covering past or present depression were together able to identify a group of women with a 30% risk of depression but the predictive ability of individual items was also poor. Both EPDS and the screening questionnaires taken together could successfully predict those who did not become depressed postnatally, in that there were few false positives, but the rate of false negatives was unacceptably high. The screening questionnaire could not be used either in clinical practice or in research into preventive intervention on the basis of this evidence. Similarly, these results do not support the use of the EPDS in antenatal screening for risk of postnatal depression. What are the possible reasons for the poor prediction by a questionnaire based on well established risk factors? Explanations may lie in the questionnaire itself, in the sampling method, or in the assumptions about postnatal depression that underlie the study. The screening questionnaire was not assessed for test retest reliability, but internal reliability was satisfactory. Being composed of direct questions concerning accepted risk factors, it carried reasonable face validity. Although all items were given the same weighting in the overall score, its predictive ability was not improved when groups of items obtained from principal components analysis were entered into the analysis separately or in a stepwise fashion. Therefore, despite some potential flaws in its construction, it would be hard simply to attribute the failure in prediction to the form of the questionnaire. The study design required subjects to be attending for hospital-based antenatal care at 36 wk gestation and to return the postal EPDS 8 wk after delivery. It could be argued that some women at high risk may have been omitted because of nonattendance antenatally or inability to respond at 8 wk resulting from depression itself. As already described, there was a slight tendency for those who did not respond postnatally to have scored more highly on the antenatal EPDS. Nevertheless, the overall response rate postnatally (77%) was satisfactory and the rate of postnatal depression identified by the EPDS was in the expected range at 13%. A third possibility is that the assumptions in this study about the applicability of risk factors to different populations are wrong. The risk factors on which the questionnaire was based came largely from K u m a r and Robson [3] whose subjects were women in their first pregnancy attending an antenatal clinic in a north London hospital serving a predominantly middle class area. Potential risk factors were recorded during the first trimester. Subjects in our study were both primiparous and multiparous, were assessed towards the end of pregnancy and were attending King's
544
L. APPLEBY et al.
College Hospital in London, which serves a predominantly poor inner city area. It is possible that the putative risk factors did not transfer well from one population to the other. If this was so, it would support a clinical impression that depression found postnatally is not a homogeneous phenomenon. Although there is no clear dividing line between subtypes of the disorder, there may be groups of patients who differ according to the relative importance of cognitive, social and biological risk factors. In the predominantly social risk group, multiple social problems such as poor housing may be made more stressful by the additional demands of a new baby. In the predominantly cognitive risk group, perfectionist and self-critical attitudes may be amplified by the responsibility of child care. Although similar risk factors have been reported by several studies, differences have also been found such as in the relationship between antenatal mood disturbance and postnatal depression [3, 4]. This may reflect the fact that any socially vulnerable group may also be depressed during pregnancy when social adversity will have almost the same impact it does postnatally, unlike any cognitively vulnerable group whose vulnerability is more likely to be expressed after delivery. Heterogeneity of subject samples, with varying proportions of these groups in different studies, may have led to some inconsistencies in the findings on risk factors and may have contributed to our disappointing result. It is possible that some items of our questionnaire, being derived from previous studies, were less likely to predict depression in our inner city subject sample with its predominantly social vulnerability. Although mainly negative, the results of this study should not deter other researchers from further attempts at developing clinically useful ways of predicting this common and important illness, which can have a long-term impact on maternal health [3, 10] and infant development [11-13]. However, the possible reasons for relatively poor prediction should be remembered in the construction of any future screening questionnaire. Acknowledgement--The authors would like to thank Dr Maureen Marks for her comments on an earlier draft of this paper. Dr Graham Dunn provided statistical advice.
REFERENCES 1. PITT B. Atypical depression following childbirth. Br J Psychiatry 1968; 114" 1225 1235. 2. Cox JL, CONNOR Y, KENDELL RE. Prospective study of the psychiatric disorders of childbirth. Br J Psychiatry 1982; 140:111 117. 3. KUMAR R, ROBSON KM. A prospective study of emotional disorders in childbearing women. Br J Psychiatry 1984; 144: 35-47. 4. WATSON JP, ELLIOTT SA, RUGG AJ, BROUGH D1. Psychiatric disorder in pregnancy and the first postnatal year. Br J Psychiatry 1984; 144: 453-462. 5. STEIN A, COOPER PJ, CAMPBELL EA, DAY A, ALTHAM PME. Social adversity and perinatal complications: their relation to postnatal depression. Br M e d J 1989; 288: 1073-1074. 6. O'HARA MW and ZE~OSKI EM. Postpartum depression: a comprehensive review. In Motherhood and Mental Illness 2. Causes and Consequences (Edited by KUMAR R, BROC~INGTON IF). London: Wright, 1988. 7. APPLEBY L, FOX H, SHAW M, KUMAR R. The psychiatrist in the obstetric unit: establishing a liaison service. Br J Psychiatry 1989; 154: 510-515. 8. HOLDENJM, SAGOVSKYR, COX JL. Counselling in a general practice setting: controlled study of health visitor intervention in treatment of postnatal depression. Br M e d J 1989; 208: 223-226. 9. Cox JL, HOLDEN JM, SAGOVSKYR. Detection of postnatal depression: development of the Edinburgh postnatal depression scale. Br J Psychiatry 1987; 150:782 786.
Screening for postnatal depression
545
10. NOTT PN. Extent, timing and persistence of emotional disorders following childbirth. Br J Psychiatry 1987; 151: 523-527. 11. WRATERM, ROONEYAC, Tr~OMASPF, Cox JL. Postnatal depression and child development: a 3-year follow-up study. Br J Psychiatry 1985; 146: 622-627. 12. COGILLSR, CAPLANHL. ALEXANDRAH, ROBSONKM. Impact of maternal depression on the cognitive development of young children. Br M e d J 1986; 292:1165 1167. 13. STEIN A, GATH DH, BUCHER J, BOND A, DAY A, COOPER PJ. The relationship between postnatal depression and mother-child interaction. Br J Psychiatry 1991; 158: 4(~52.