- Email: [email protected]
Depressive symptoms in women in the six months after miscarriage
Depressive symptoms in women in the six months after miscarriage Richard Neugebauer, PhD, MPH,.·b Jennie Kline, PhD,b.d Patricia O'Connor, PhD,b.f Patrick Shrout, PhD,.. b.• Jim Johnson, PhD,b Andrew Skodol, MD,c Judith Wicks, BA,b and Mervyn Susser, MB, BCh, DPH b. d New York, New York This study, the first systematic investigation of the psychiatric impact of miscarriage, tests whether miscarriage markedly Increases depressive symptoms in the 6 months after loss. We interviewed 382 miscarrying women entering the study at 2 weeks, 6 weeks, or 6 months after loss and, for comparison, 283 pregnant women and 318 community women not recently pregnant. Among women interviewed 2 weeks after miscarriage the proportion highly symptomatic on the Center for Epidemiologic Studies-Depression scale was 3.4 times that of pregnant women (95% confidence limits 2.0 and 5.0) and 4.3 times that of community women (95% confidence limits 3.0 and 5.8). Among women first interviewed 6 weeks and 6 months after miscarriage the proportion highly symptomatic was three times that of the community women. Women reinterviewed at 6 weeks and 6 months did not have elevated symptom levels, a result attributed to the unintended therapeutic and test effects of study interviews. Interviews were fully structured, readily administered by telephone by nonmedical personnel. The possibility that such interviews afford miscarrying women substantial psychologic benefits merits future investigation. (AM J OSSTET GVNECOL 1992;166:104-9.)
Key words: Spontaneous abortion, depression, grief counseling, miscarriage, pregnancy Twelve percent to 15% of clinically recognized pregnancies end in spontaneous abortion. I For many women miscarriage represents death of a future child, dislocates reproductive plans, and may engender doubts about procreative competence. Nonetheless society lacks established rituals for mourning such loss!,3 In spite of repeated calls for studies of the psychiatric sequelae of reproductive loss2, 4·6 relevant investigations are rare, and methodologic shortcomings cloud their findings. Only three investigations focused solely on miscarriage'·9; an additional six included miscarriage (loss usually at >20 weeks' completed gestation) in investigations of perinatal death. 10·15 These studies lacked a comparison group unexposed to recent reproductive loss, thereby precluding estimation of its effect on depressive symptoms. Wide variation in the timing of subject assessment ranging within a given study from a few From the EpidemIOlogy of Developmental Bram Disorders Department, New York State Psychiatric Institute"; the Gertrude H. Sergtevsky Centei' and the Department of Psychiatry,' Faculty of MedZClne, and the DIVISIOns of Epzdemwlogyd and Biostatistics,' School of Public Health, Columbza UniverSity; and the Department of Psychology, Russell Sage College/ Supported in part by grants MH39581 and HD15509 from the NatIOnal Institutes of Health. Recewedfor publicatIOn December 11.1990; revISed May 23,1991; accepted May 29,1991, Reprint requests: Richard Neugebauer, PhD, MPH, Gertrude H. Sergievsky Center, Faculty of Medicine, Columbia Unzverszty, 630 West 168th St" New York, NY 10032. 6111J1440 104
weeks to months or even years after 10ss,1I, '2, '5 small samples, and absence of data on response rates ll , 13 further complicate evaluation and interpretation of research findings.
Material and methods Study design. The study comprised three cohorts: miscarrying women, pregnant women, and women in the community not pregnant in the preceding year. Levels of depressive symptoms in the miscarrying women were compared with those in these two other groups. Miscarriage cohort. Miscarriage was defined as the involuntary termination of a nonviable, intrauterine pregnancy before 28 completed weeks of gestation with the conceptus dead on expulsion. The miscarriage cohort was derived from the cases of an antecedent hospital-based case-control study of risk factors for miscarriage' 6 in which 77% of all women attending a New York City medical center for miscarriages from late 1984 to 1986 were interviewed. This antecedent case-control study interviewed approximately 75% of these cases within 2 weeks of miscarriage, 95% within 8 weeks, and the remainder by 6 months. Pregnant cohort. The pregnant cohort was derived from the controls of the case-control study and consisted of women registered for prenatal care before 22 weeks' gestation selected to be similar to the miscarrying women in age and whether they were receiving medical
Depression after miscarriage
Volume 166 Number 1, Part 1
105
Table I. Selected sociodemographic characteristics of women in miscarriage, pregnant, and community cohorts Cohort
Sociodemograph!c Characteristic
Miscarnage
= 382)
P1'egnant (n = 283)
Communtty (n = 318)
Age (yr, mean ± SD) Ethnicity (%) White Black Hispanic Other Interviewed in Spanish (%) Education (%) Less than high school High school graduate Some college College graduate plus Income (%)* < $9,000
29.2 ± 6.2
28.9 ± 5.7
30.0 ± 6.4
32.2 20.7 39.8 7.3 32.7
36.4 15.5 44.9 3.2 25.8
36.6 19.6 38.2 5.7 26.4
28.6 24.1 22.8 24.6
24.7 20.9 25.8 28.6
20.3 24.5 29.8 25.5
37.1 24.7
39.6 26.9
23 .9 22 .2
26.7 60.2 13.1
22.6 60.8 16.7
46.5 35.2 18.3
(n
*p < 0.01 for overall differences across cohorts. For categorical variables and age, differences among groups were evaluated by an overall X2 test and one-way analysis of variance, respectively.
care privately or in a public, clinic setting. Eighty-five percent of identified controls were interviewed for the case-control stud y. Of identified cases and controls, participants and nonparticipants in the case-control study were similarly distributed on sociodemographic and reproductive characteristics. 16 At the conclusion of the case-control interview all participants who were 18 years of age or older, English or Spanish speaking, and accessible by telephone were invited to participate in our study. For pregnant women we also required that the case-control interview occur before 28 weeks' completed gestation. We aimed to assess miscarrying women at three time points: the second week after miscarriage (2 weeks), weeks 6 through 8 (6 weeks), and weeks 26 through 35 (6 months). Overall, 73% (n = 382) of eligible miscarrying women were evaluated at least once; 232 were first interviewed at 2 weeks, 114 at 6 weeks, and 36 at 6 months. This staggered recruitment arises largely from the timing of the antecedent case-control study interview and logistic difficulties encountered by the current study in scheduling women for assessments within narrow, fixed time intervals after loss. Seventy-two percent (n = 283) of eligible pregnant women were interviewed on a single occasion. Community cohort. Recruits for the community cohort were located by randomly dialing telephone numbers based on the telephone area codes and exchanges of the interviewed miscarrying women. The eligibility of household residents was established for 92% of
working telephone numbers. Community women were frequency matched to the women in the miscarriage cohort on language of interview, age (in 5-year intervals), education (less than high school, high school graduate, college graduate, postgraduate training), and season of interview. Of (known) eligible community women, 82% (n = 318) were interviewed on a single occasion. Comparability of interviewed and uninterviewed subjects. Within each cohort interviewed and uninterviewed women were similar in age, ethnicity, education, language, and season of interview. Interviewed and uninterviewed miscarrying and pregnant women also did not differ on private-public status, marital status, parity, or prior reproductive loss. For miscarrying women mean length of gestation at loss was slightly greater (8 to 9 days) for interviewed women compared with those not interviewed. Mean scores on a measure of depressed mood l7 did not differ between interviewed and uninterviewed women.
Study measures The presence and duration of depressive symptoms in the week before the interview were evaluated with the 20-item Center for Epidemiologic Studies-Depression (CES-D) scale. IS This scale measures affective, cognitive, and somatic components of depression. Each item is scored from 0 to 3 depending on the frequency of the symptom and is summed to produce an overall index. The CES-D correlates well with other self-report
106 Neugebauer et al.
January 1992
Am J Obstet Gynecol
Table II. Selected reproductive history characteristics of women in miscarriage, pregnant, and community cohorts Cohort Characteristic
Miscarriage (n = 382)
Pregnant (n = 283)
34.0
39.2
36.6
34.6
50.5
38.0 29.9 18.3 13.8
35.3 37.8 18.4 8.5
50.5 15.7 19.7 14.1
66.0 22.3 11.7
71.0 20.8 8.2
82.5 12.9 4.6
Private payment of hospital bill (%) Reproductivet Nulliparous (%) Living children (%) 0 I 2 3+ Prior reproductive loss (%)* 0 1 2+ Weeks' gestation (%)t <8 8-11 12-15 16-19 20-23 24-27 28+
10.5 38.4 24.3 15.2 8.4 3.2
Community (n = 318)
0.7 5.3 16.6 33.9 36.8 6.7
*Includes spontaneous abortions (89%), fetal deaths (5%), ectopic pregnancies (3%), and neonatal deaths (3%). tFor the miscarriage cohort, the interval refers to time from last menstrual period to date of expulsion of conceptus; for the pregnant cohort, the interval refers to time from last menstrual period to miscarriage study interview. tp Value < 0.01 for overall differences across cohorts. Differences among groups were evaluated by one-way analysis of variance.
Table III. Comparison of CES-D means within miscarriage cohort across time points and by interview status Women first interviewed at 2 wk
Women first interviewed at 6wkor6mo
Adjusted difference In means (95% CL)*
Time point
No.
Observed CES-D (mean ± SEM)
No.
Observed CES-D (mean ± SEM)
Across tIme pOInts: Women first intervIewed at2wkvs women first interviewed later
2 wk 6wk 6mo
232 196 266
23.9 ± 0.8 13.7 ± 0.8 13.8 ± 0.7
114 36
18.8 ± 1.2 20.6 ± 2.3
4.0 (1.1, 6.8) 3.3 (-l.l, 7.7)
Within time POints: Relnterviewed women vs first intervIewed women
6.1 (3.4, 8.8) 6.3 (2.1, 10.5)
CL, Confidence limits. *Means are adjusted by ordinary least-squares multiple regression analysis.
measures of depressive symptoms, distinguishes clinically depressed patients from the general population, and has been used previously with samples of pregnant women. 19 A CES-D mean of 24.4 has been reported for heterogeneous samples of psychiatric inpatients. IS The CES-D was administered by telephone 20 and preceded an extensive interview covering the woman's sociodemographic characteristics, social supports, attitude toward the pregnancy, and reproductive history, including history of the current pregnancy and miscarriage. Interviewers were not required to have clinical training or experience.
Sample characteristics. The three cohorts were similarly distributed on the majority of sociodemographic variables (Table I). Overall mean age was 30 years; 40% were Hispanic; a quarter had not graduated from high school; approximately 30% had incomes of <$10,000. As expected, the two hospital-based cohorts contained fewer single and childless women and had more reproductive loss (Table II). Half of index miscarriages occurred at < 12 weeks' gestation; the majority of pregnant women were interviewed after the first trimester. Analytic strategy. A first set of analyses tested differences in symptom levels within the miscarriage co-
Depression after miscarriage
Volume 166 Number 1, Part I
107
Table IV. Comparison of proportions of subjects scoring 2:30 on the CES-D: Miscarriage cohort at three time points after loss (stratified by interview status) with pregnant and community cohorts at one time point Adjusted ratw of proporl!ons* Cohort (n)
Pregnant (283) Community (318) Miscarriage, 2 wk First interview (232) Miscarriage, 6 wk First interview (114) Remterview (196) Miscarriage, 6 mo First interview (36) Reinterview (266)
Observed percent scoring '?30 (mean:!: SEM)
Miscarriage to pregnant cohorts (95% CL)
!
MlScarnage to communzty cohorts (95 % CL)
12.4 :!: 2.0 10.1 :!: 1.7 36.2 :!: 3.2
3.4 (2.0, 5.0):j:
4 .3 (3.0, 5.8)
20.2 :!: 3.8 11.7 :!: 2.3
2.6 (1.5, 4.4) 1.4 (0.8, 2.4)
30.6 :!: 7.7 12.8 :!: 2.0
3.0 (1.2, 5.9) 1.5 (0.9, 2.5)
CL. Confidence lImits. *Proportions are adjusted with maximum likelihood logistic regression analysis. tOdds ratios (95% CL) for miscarriage-community comparisons in descending-column order: 6.2 (3.6, 10.5); 3. 1 (1.5,6.2); 1.4 (0.7, 2.7); 3.6 (1.3, 10.2); 1.6 (0.9, 2.9). :j:Odds ratio (95 % CL) for this comparison is 4.9 (2.4, 10.0).
hort across time points. A second set tested differences in symptom levels between the miscarriage cohort and the two comparison groups. At 2 weeks after loss the miscarriage cohort was compared with both the pregnant and community cohorts. Thereafter, because the pregnant cohort offered an estimate of symptom levels among miscarrying women immediately before loss, it lacked pertinence beyond the 2-week mark. Depressive symptom levels did not vary with length of gestation at loss. Consequently early and late miscarriages were combined in these analyses. CES-D symptom levels were treated as continuous and categorical variables. The categorical approach tested whether findings based on the entire range of CES-D scores hold for extreme symptom levels. Subjects were dichotomized so that women receiving CESD scores of 2:30, that is, those who reported at least half of the 20 symptoms in the CES-D for at least 1 week, were categorized as highly symptomatic. Approximately two thirds of these individuals would be expected to meet diagnostic criteria for major depressive disorder. 2 ! Analyses were conducted with ordinary least-squares multiple regression and maximum likelihood logistic regression to permit adjustment for potentially confounding variables. Comparisons of miscarrying with community women controlled for maternal age, ethnicity, marital and socioeconomic status, number of living children, and prior reproductive loss. Comparisons with pregnant women, as well as those within the miscarriage cohort, also controlled for private-public status, length of gestation (at time of miscarriage at interview), and pregnancy attitude. Results from logistic regression analyses are presented in terms
of ratios of proportions scoring 2:30 and odds ratios, the latter reserved for notes to the relevant tables. Statistical significance was assessed at p < 0.05. Results
Comparisons within the miscarriage cohort. In the miscarriage cohort at 2 weeks after loss the observed symptom mean is 23.9 (Table III), with 36.2% of the subjects being highly symptomatic (Table IV). At 6 weeks and 6 months after loss, depressive symptom levels are lower than at 2 weeks but substantially so only among reinterviewed women. Between miscarrying women first interviewed at 2 weeks and those first interviewed at later time points, adjusted differences in means at 6 weeks are 4.0 points and at 6 months 3.3 (Table III). Between women first interviewed and those reinterviewed at the same times, the differences are considerable. At 6 weeks their adjusted symptom means differ by 6.1 points and at 6 months by 6.3 points. The same pattern of results holds in analyses based on proportions highly symptomatic. Comparisons between study cohorts. In the miscarriage cohort at 2 weeks after loss the adjusted symptom mean exceeds those in the pregnant and community cohorts by 7.5 to 9.1 points, respectively (Table V). Among miscarrying women receiving their first interviews at 6 weeks and 6 months symptom means were still significantly greater than that in the community cohort (5.2 and 5.9, respectively). Similarly, at 2 weeks after loss the proportion of highly symptomatic subjects in the miscarriage cohort is 3.4 times that in the pregnant cohort and 4.3 times that in the community cohort (Table IV). Among miscarrying women first inter-
108 Neugebauer et al.
January 1992 Am J Obstet Gyneco1
Table V. Comparison of CES-D means: Miscarriage cohort at three time points after loss (stratified by interview status) with pregnant and community cohorts at one time point Adjusted difference in means*
Cohort (n)
Observed CES-D (mean ± SEM)
Pregnant (283) Community (318) Miscarriage, 2 wk First interview (232) Miscarriage, 6 wk First interview (114) Reinterview (196) Miscarriage, 6 mo First interview (36) Reinterview (266)
15.2 ± 0.7 14.6 ± 0.8 23.9 ± 0.8
Between miscarriage and pregnant cohorts (95% CL)
Between miscarriage and community cohorts (95% CL)
7.5 (4.5, 10.5)
9.1 (6.9, 11.2)
18.8 ± 1.2 13.7 ± 0.8
5.2 (2.6, 7.8) -1.0 (-3.1,1.1)
20.6 ± 2.3 13.8 ± 0.7
5.9 (2.0, 9.8) - 0.6 ( - 2.6, 1.5)
CL, Confidence limits. *Means are adjusted by ordinary least-squares multiple regression analysis.
viewed at 6 weeks and 6 months, the proportions that are highly symptomatic are 2.6 and 3 times that in the community cohort, respectively. Among women reinterviewed at 6 weeks or 6 months, symptom levels do not differ from those in the community cohort. The pattern of findings described persisted after controlling for the contribution of somatic symptoms to the overall CES-D score. Hence symptom levels reported by study subjects do not reflect simply somatic complaints arising from the physical sequelae of the miscarriage (or of the pregnancy). In addition, while one third of miscarrying women were hospitalized overnight and another third for 2:2 nights, hospital stay was not associated with symptom levels. Women who had dilatation and curettage procedures (approximately 85%) and those who did not also did not differ on level of depressive symptoms. Finally, women with early and late miscarriages reported similarly elevated symptom levels. For example, 34.1 % of women with first-trimester miscarriages (12 completed weeks' gestation) were highly symptomatic, as were 39.8% of women with later loss (p < 0.40).
Comment Among miscarrying women depressive symptoms are elevated fourfold over those in the community women in the second week after miscarriage and 3.4-fold over those in pregnant women. These results document for the first time the magnitude of the early effect of miscarriage on maternal psychologic status. Levels remain high at 6 weeks and at 6 months among women first interviewed at these time points but not among the reinterviewed. This difference in symptom levels between women first interviewed at 6 weeks and 6 months and those being reinterviewed suggests a strong ther-
apeutic effect of study interviews. However, alternative explanations for this finding need to be considered first. Selection bias in the form of selective departure or selective deferment of participation by more depressed study subjects is unlikely to contribute substantially to these results. First, as to selective departure, women interviewed at 2 weeks or at 6 weeks after loss who thereafter left the study did not differ on symptom levels from women who remained. Second, as for selective deferment, adjusted scores on the depressed mood scale administered at the time of the case-control interview did not differ among women first interviewed at 2 weeks, 6 weeks, and 6 months. Furthermore, women first interviewed at 2 weeks and 6 weeks had nearly identical symptom levels at 6 months (13.6 and 14.1, respectively). Similarly, in a large community survey of depressive symptoms, subjects interviewed in the calendar week targeted for study assessment and those deferring to a later time point did not differ on CESD symptom levels. 22 The lower scores of reinterviewed women are more easily explained as a consequence of unintended benefits of study participation, together with a possible test effect arising from scale repetition. Our initial interview afforded each subject an opportunity to discuss her pregnancy, the circumstances of the miscarriage, and its psychologic aftermath. In this respect it inadvertently resembled grief counseling. 7• 23 The efficacy of such counseling, often propounded, has been rigorously demonstrated in at least one randomized, controlled trial. 23 After a perinatal death women in the treatment group received a modal number of two counseling sessions. At 6 months after loss, the rate of psychiatric disorder in the treated group was one quarter that in the untreated group (12.5% and 52.6%, re-
Volume 166 Number I, Part 1
spectively). The first interview with miscarrying women in our study lasted from 1.5 to 3 hours and was generally given in two parts on separate days. Our estimates of a twofold to threefold elevation in depressive symptoms at 6 weeks and 6 months after loss were based on miscarrying women receiving their first interview; these estimates accord well with the previous controlled investigation of reproductive loss. Clarke and Williams" compared depressive symptoms in women experiencing perinatal death with depressive symptoms in women having live births. (The study protocol did not involve extended personal interviews with subjects.) At 6 weeks and 6 months thereafter the proportions of depressed women were 2.9 and 2.4 times those of women with live births, respectively. The results from the study of grief counseling cited above 2' also argue for at least a fourfold elevation in psychiatric symptoms at 6 months after loss in the absence of intervention. Given the preceding evidence based on both our own data and those of these other reports, we take symptom levels at first interview at each time to represent a better estimate of the true magnitude of the effect of miscarriage than levels at reinterview. Previous authors have suggested that the "conspiracy of silence'" surrounding reproductive loss on the part of hospital staff, family, and friends exacerbates depressive reactions. It would not be at all surprising therefore if our interview, with its direct focus on the miscarriage, proved to be psychotherapeutic. Although we had anticipated some drop in the subjects' reported symptoms as a result of study interviews, the magnitude of the observed decline was unexpected. The efficacy of this proposed method of intervention requires rigorous a priori testing. Nonetheless , on the basis of these initial findings, structured telephone interviews administered by lay personnel should have strong clinical and budgetary appeal for medical institutions unable to offer formal bereavement counseling. These results should also prompt obstetricians to meet with a miscarrying patient soon after loss simply to afford her a potentially therapeutic opportunity to review the circumstances of the pregnancy and the miscarriage. REFERENCES 1. Kline], Stein Z, Susser M. Conception to birth: epidemiology of prenatal development. New York: Oxford University Press, 1989:43-68. 2. Phipps S. Mourning response and intervention in stillbirth: an alternative genetic counseling approach. Soc Bioi 1981 ;28:1-13.
Depression after miscarriage
109
3. Stirtzinger R, Robinson GE. The psychologic effects of spontaneous abortion. Can Med Assoc] 1989; 140:799801,805. 4. Morris D. Parental reactions to perinatal death. Proc R Soc Med 1976;69:33-4. 5. Lewis E. The management of stillbirth: coping with an unreality. Lancet 1976;2:619-20. 6. Elliott BA, Hein HA. Neonatal death; reRections for physicians. Pediatrics 1978;62:96-100. 7. Leppert PC, Pahlka BS. Grieving characteristics after spontaneous abortion: a management approach. Obstet Gynecol 1984;64: 119-22. 8. Seibel M, Graves WL. The psychological implications of spontaneous abortions.] Reprod Med 1980;25:161-5. 9. Friedman T, Gath D. The psychiatric consequences of spontaneous abortion. Br] Psychiatry 1989;155:810-3. 10. Toedter L], Lasker ]N, Alhadeff]M. The Perinatal Grief Scale: development and initial validation. Am ] Orthopsychiatry 1988;58:435-49. 11. Peppers LG, Knapp RJ. Maternal reactions to involuntary fetal/infant death. Psychiatry 1980;43:155-9. 12. Nicol MT, Tompkins]R, Campbell NA, Syme GJ. Maternal grieving response after perinatal death. Med] Aust 1986;144:287-9. 13. laRoche C, Lalinec-Michaud M, Engelsmann F, et al. Grief reactions to perinatal death-a follow-up study. Can ] Psychiatry 1984;29:14-9. 14. Graham MA, Thompson SC, Estrada M, Yonekura ML. Factors affecting psychological adjustment to a fetal death. AM] OSSTET GVNECOL 1987;157:254-7. 15. Rowe ], Clyman R, Green C, Mikkelsen C, Haight ], Ataide L. Follow-up families who experience a perinatal death. Pediatrics 1978;62:166-70. 16. Kline], Stein Z, Susser M, Warburton D. Induced abortion and the chromosomal characteristics of subsequent miscarriages (spontaneous abortions). Am ] Epidemiol 1986;123 : 1066-79. 17. Folstein MF, Luria R. Reliability, validity and clinical application of the Visual Analogue Mood Scale. Psychol Med 1973;3:479-86. 18. Radloff LS. The CES-D Scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1977;1:385-401. 19. Zuckerman B, Amaro H, Bauchner H, Cabral H. Depressive symptoms during pregnancy: relationship to poor health behaviors. AM ] OSSTET GVNECOL 1989; 160: 1107-11. 20. Aneshensel CS, Frerichs RR, Clark VA, Yokopenic PA. Measuring depression in the community: a comparison of telephone and personal interviews. Public Opinion Q 1982;46: 110-21. 21 . Boyd]H, Weissman MN, Thompson D, Myers]K. Screening for depression in a commumty sample. Arch Gen Psychiatry 1982;39: 1195-200. 22. Mebrane 1. On time and late respondents. Technical report, Center for Epidemiologic Studies, National Institute of Mental Health, 1973. 23. Forrest GC, Standish E, Baum ]D. Support after perinatal death : a study of support and counselling after perinatal bereavement. BM] 1982;285: 1475-9. 24. Clarke M, Williams AJ. Depression in women after perinatal death. Lancet 1979;2:916-7.
Key words: Spontaneous abortion, depression, grief counseling, miscarriage, pregnancy Twelve percent to 15% of clinically recognized pregnancies end in spontaneous abortion. I For many women miscarriage represents death of a future child, dislocates reproductive plans, and may engender doubts about procreative competence. Nonetheless society lacks established rituals for mourning such loss!,3 In spite of repeated calls for studies of the psychiatric sequelae of reproductive loss2, 4·6 relevant investigations are rare, and methodologic shortcomings cloud their findings. Only three investigations focused solely on miscarriage'·9; an additional six included miscarriage (loss usually at >20 weeks' completed gestation) in investigations of perinatal death. 10·15 These studies lacked a comparison group unexposed to recent reproductive loss, thereby precluding estimation of its effect on depressive symptoms. Wide variation in the timing of subject assessment ranging within a given study from a few From the EpidemIOlogy of Developmental Bram Disorders Department, New York State Psychiatric Institute"; the Gertrude H. Sergtevsky Centei' and the Department of Psychiatry,' Faculty of MedZClne, and the DIVISIOns of Epzdemwlogyd and Biostatistics,' School of Public Health, Columbza UniverSity; and the Department of Psychology, Russell Sage College/ Supported in part by grants MH39581 and HD15509 from the NatIOnal Institutes of Health. Recewedfor publicatIOn December 11.1990; revISed May 23,1991; accepted May 29,1991, Reprint requests: Richard Neugebauer, PhD, MPH, Gertrude H. Sergievsky Center, Faculty of Medicine, Columbia Unzverszty, 630 West 168th St" New York, NY 10032. 6111J1440 104
weeks to months or even years after 10ss,1I, '2, '5 small samples, and absence of data on response rates ll , 13 further complicate evaluation and interpretation of research findings.
Material and methods Study design. The study comprised three cohorts: miscarrying women, pregnant women, and women in the community not pregnant in the preceding year. Levels of depressive symptoms in the miscarrying women were compared with those in these two other groups. Miscarriage cohort. Miscarriage was defined as the involuntary termination of a nonviable, intrauterine pregnancy before 28 completed weeks of gestation with the conceptus dead on expulsion. The miscarriage cohort was derived from the cases of an antecedent hospital-based case-control study of risk factors for miscarriage' 6 in which 77% of all women attending a New York City medical center for miscarriages from late 1984 to 1986 were interviewed. This antecedent case-control study interviewed approximately 75% of these cases within 2 weeks of miscarriage, 95% within 8 weeks, and the remainder by 6 months. Pregnant cohort. The pregnant cohort was derived from the controls of the case-control study and consisted of women registered for prenatal care before 22 weeks' gestation selected to be similar to the miscarrying women in age and whether they were receiving medical
Depression after miscarriage
Volume 166 Number 1, Part 1
105
Table I. Selected sociodemographic characteristics of women in miscarriage, pregnant, and community cohorts Cohort
Sociodemograph!c Characteristic
Miscarnage
= 382)
P1'egnant (n = 283)
Communtty (n = 318)
Age (yr, mean ± SD) Ethnicity (%) White Black Hispanic Other Interviewed in Spanish (%) Education (%) Less than high school High school graduate Some college College graduate plus Income (%)* < $9,000
29.2 ± 6.2
28.9 ± 5.7
30.0 ± 6.4
32.2 20.7 39.8 7.3 32.7
36.4 15.5 44.9 3.2 25.8
36.6 19.6 38.2 5.7 26.4
28.6 24.1 22.8 24.6
24.7 20.9 25.8 28.6
20.3 24.5 29.8 25.5
37.1 24.7
39.6 26.9
23 .9 22 .2
26.7 60.2 13.1
22.6 60.8 16.7
46.5 35.2 18.3
(n
*p < 0.01 for overall differences across cohorts. For categorical variables and age, differences among groups were evaluated by an overall X2 test and one-way analysis of variance, respectively.
care privately or in a public, clinic setting. Eighty-five percent of identified controls were interviewed for the case-control stud y. Of identified cases and controls, participants and nonparticipants in the case-control study were similarly distributed on sociodemographic and reproductive characteristics. 16 At the conclusion of the case-control interview all participants who were 18 years of age or older, English or Spanish speaking, and accessible by telephone were invited to participate in our study. For pregnant women we also required that the case-control interview occur before 28 weeks' completed gestation. We aimed to assess miscarrying women at three time points: the second week after miscarriage (2 weeks), weeks 6 through 8 (6 weeks), and weeks 26 through 35 (6 months). Overall, 73% (n = 382) of eligible miscarrying women were evaluated at least once; 232 were first interviewed at 2 weeks, 114 at 6 weeks, and 36 at 6 months. This staggered recruitment arises largely from the timing of the antecedent case-control study interview and logistic difficulties encountered by the current study in scheduling women for assessments within narrow, fixed time intervals after loss. Seventy-two percent (n = 283) of eligible pregnant women were interviewed on a single occasion. Community cohort. Recruits for the community cohort were located by randomly dialing telephone numbers based on the telephone area codes and exchanges of the interviewed miscarrying women. The eligibility of household residents was established for 92% of
working telephone numbers. Community women were frequency matched to the women in the miscarriage cohort on language of interview, age (in 5-year intervals), education (less than high school, high school graduate, college graduate, postgraduate training), and season of interview. Of (known) eligible community women, 82% (n = 318) were interviewed on a single occasion. Comparability of interviewed and uninterviewed subjects. Within each cohort interviewed and uninterviewed women were similar in age, ethnicity, education, language, and season of interview. Interviewed and uninterviewed miscarrying and pregnant women also did not differ on private-public status, marital status, parity, or prior reproductive loss. For miscarrying women mean length of gestation at loss was slightly greater (8 to 9 days) for interviewed women compared with those not interviewed. Mean scores on a measure of depressed mood l7 did not differ between interviewed and uninterviewed women.
Study measures The presence and duration of depressive symptoms in the week before the interview were evaluated with the 20-item Center for Epidemiologic Studies-Depression (CES-D) scale. IS This scale measures affective, cognitive, and somatic components of depression. Each item is scored from 0 to 3 depending on the frequency of the symptom and is summed to produce an overall index. The CES-D correlates well with other self-report
106 Neugebauer et al.
January 1992
Am J Obstet Gynecol
Table II. Selected reproductive history characteristics of women in miscarriage, pregnant, and community cohorts Cohort Characteristic
Miscarriage (n = 382)
Pregnant (n = 283)
34.0
39.2
36.6
34.6
50.5
38.0 29.9 18.3 13.8
35.3 37.8 18.4 8.5
50.5 15.7 19.7 14.1
66.0 22.3 11.7
71.0 20.8 8.2
82.5 12.9 4.6
Private payment of hospital bill (%) Reproductivet Nulliparous (%) Living children (%) 0 I 2 3+ Prior reproductive loss (%)* 0 1 2+ Weeks' gestation (%)t <8 8-11 12-15 16-19 20-23 24-27 28+
10.5 38.4 24.3 15.2 8.4 3.2
Community (n = 318)
0.7 5.3 16.6 33.9 36.8 6.7
*Includes spontaneous abortions (89%), fetal deaths (5%), ectopic pregnancies (3%), and neonatal deaths (3%). tFor the miscarriage cohort, the interval refers to time from last menstrual period to date of expulsion of conceptus; for the pregnant cohort, the interval refers to time from last menstrual period to miscarriage study interview. tp Value < 0.01 for overall differences across cohorts. Differences among groups were evaluated by one-way analysis of variance.
Table III. Comparison of CES-D means within miscarriage cohort across time points and by interview status Women first interviewed at 2 wk
Women first interviewed at 6wkor6mo
Adjusted difference In means (95% CL)*
Time point
No.
Observed CES-D (mean ± SEM)
No.
Observed CES-D (mean ± SEM)
Across tIme pOInts: Women first intervIewed at2wkvs women first interviewed later
2 wk 6wk 6mo
232 196 266
23.9 ± 0.8 13.7 ± 0.8 13.8 ± 0.7
114 36
18.8 ± 1.2 20.6 ± 2.3
4.0 (1.1, 6.8) 3.3 (-l.l, 7.7)
Within time POints: Relnterviewed women vs first intervIewed women
6.1 (3.4, 8.8) 6.3 (2.1, 10.5)
CL, Confidence limits. *Means are adjusted by ordinary least-squares multiple regression analysis.
measures of depressive symptoms, distinguishes clinically depressed patients from the general population, and has been used previously with samples of pregnant women. 19 A CES-D mean of 24.4 has been reported for heterogeneous samples of psychiatric inpatients. IS The CES-D was administered by telephone 20 and preceded an extensive interview covering the woman's sociodemographic characteristics, social supports, attitude toward the pregnancy, and reproductive history, including history of the current pregnancy and miscarriage. Interviewers were not required to have clinical training or experience.
Sample characteristics. The three cohorts were similarly distributed on the majority of sociodemographic variables (Table I). Overall mean age was 30 years; 40% were Hispanic; a quarter had not graduated from high school; approximately 30% had incomes of <$10,000. As expected, the two hospital-based cohorts contained fewer single and childless women and had more reproductive loss (Table II). Half of index miscarriages occurred at < 12 weeks' gestation; the majority of pregnant women were interviewed after the first trimester. Analytic strategy. A first set of analyses tested differences in symptom levels within the miscarriage co-
Depression after miscarriage
Volume 166 Number 1, Part I
107
Table IV. Comparison of proportions of subjects scoring 2:30 on the CES-D: Miscarriage cohort at three time points after loss (stratified by interview status) with pregnant and community cohorts at one time point Adjusted ratw of proporl!ons* Cohort (n)
Pregnant (283) Community (318) Miscarriage, 2 wk First interview (232) Miscarriage, 6 wk First interview (114) Remterview (196) Miscarriage, 6 mo First interview (36) Reinterview (266)
Observed percent scoring '?30 (mean:!: SEM)
Miscarriage to pregnant cohorts (95% CL)
!
MlScarnage to communzty cohorts (95 % CL)
12.4 :!: 2.0 10.1 :!: 1.7 36.2 :!: 3.2
3.4 (2.0, 5.0):j:
4 .3 (3.0, 5.8)
20.2 :!: 3.8 11.7 :!: 2.3
2.6 (1.5, 4.4) 1.4 (0.8, 2.4)
30.6 :!: 7.7 12.8 :!: 2.0
3.0 (1.2, 5.9) 1.5 (0.9, 2.5)
CL. Confidence lImits. *Proportions are adjusted with maximum likelihood logistic regression analysis. tOdds ratios (95% CL) for miscarriage-community comparisons in descending-column order: 6.2 (3.6, 10.5); 3. 1 (1.5,6.2); 1.4 (0.7, 2.7); 3.6 (1.3, 10.2); 1.6 (0.9, 2.9). :j:Odds ratio (95 % CL) for this comparison is 4.9 (2.4, 10.0).
hort across time points. A second set tested differences in symptom levels between the miscarriage cohort and the two comparison groups. At 2 weeks after loss the miscarriage cohort was compared with both the pregnant and community cohorts. Thereafter, because the pregnant cohort offered an estimate of symptom levels among miscarrying women immediately before loss, it lacked pertinence beyond the 2-week mark. Depressive symptom levels did not vary with length of gestation at loss. Consequently early and late miscarriages were combined in these analyses. CES-D symptom levels were treated as continuous and categorical variables. The categorical approach tested whether findings based on the entire range of CES-D scores hold for extreme symptom levels. Subjects were dichotomized so that women receiving CESD scores of 2:30, that is, those who reported at least half of the 20 symptoms in the CES-D for at least 1 week, were categorized as highly symptomatic. Approximately two thirds of these individuals would be expected to meet diagnostic criteria for major depressive disorder. 2 ! Analyses were conducted with ordinary least-squares multiple regression and maximum likelihood logistic regression to permit adjustment for potentially confounding variables. Comparisons of miscarrying with community women controlled for maternal age, ethnicity, marital and socioeconomic status, number of living children, and prior reproductive loss. Comparisons with pregnant women, as well as those within the miscarriage cohort, also controlled for private-public status, length of gestation (at time of miscarriage at interview), and pregnancy attitude. Results from logistic regression analyses are presented in terms
of ratios of proportions scoring 2:30 and odds ratios, the latter reserved for notes to the relevant tables. Statistical significance was assessed at p < 0.05. Results
Comparisons within the miscarriage cohort. In the miscarriage cohort at 2 weeks after loss the observed symptom mean is 23.9 (Table III), with 36.2% of the subjects being highly symptomatic (Table IV). At 6 weeks and 6 months after loss, depressive symptom levels are lower than at 2 weeks but substantially so only among reinterviewed women. Between miscarrying women first interviewed at 2 weeks and those first interviewed at later time points, adjusted differences in means at 6 weeks are 4.0 points and at 6 months 3.3 (Table III). Between women first interviewed and those reinterviewed at the same times, the differences are considerable. At 6 weeks their adjusted symptom means differ by 6.1 points and at 6 months by 6.3 points. The same pattern of results holds in analyses based on proportions highly symptomatic. Comparisons between study cohorts. In the miscarriage cohort at 2 weeks after loss the adjusted symptom mean exceeds those in the pregnant and community cohorts by 7.5 to 9.1 points, respectively (Table V). Among miscarrying women receiving their first interviews at 6 weeks and 6 months symptom means were still significantly greater than that in the community cohort (5.2 and 5.9, respectively). Similarly, at 2 weeks after loss the proportion of highly symptomatic subjects in the miscarriage cohort is 3.4 times that in the pregnant cohort and 4.3 times that in the community cohort (Table IV). Among miscarrying women first inter-
108 Neugebauer et al.
January 1992 Am J Obstet Gyneco1
Table V. Comparison of CES-D means: Miscarriage cohort at three time points after loss (stratified by interview status) with pregnant and community cohorts at one time point Adjusted difference in means*
Cohort (n)
Observed CES-D (mean ± SEM)
Pregnant (283) Community (318) Miscarriage, 2 wk First interview (232) Miscarriage, 6 wk First interview (114) Reinterview (196) Miscarriage, 6 mo First interview (36) Reinterview (266)
15.2 ± 0.7 14.6 ± 0.8 23.9 ± 0.8
Between miscarriage and pregnant cohorts (95% CL)
Between miscarriage and community cohorts (95% CL)
7.5 (4.5, 10.5)
9.1 (6.9, 11.2)
18.8 ± 1.2 13.7 ± 0.8
5.2 (2.6, 7.8) -1.0 (-3.1,1.1)
20.6 ± 2.3 13.8 ± 0.7
5.9 (2.0, 9.8) - 0.6 ( - 2.6, 1.5)
CL, Confidence limits. *Means are adjusted by ordinary least-squares multiple regression analysis.
viewed at 6 weeks and 6 months, the proportions that are highly symptomatic are 2.6 and 3 times that in the community cohort, respectively. Among women reinterviewed at 6 weeks or 6 months, symptom levels do not differ from those in the community cohort. The pattern of findings described persisted after controlling for the contribution of somatic symptoms to the overall CES-D score. Hence symptom levels reported by study subjects do not reflect simply somatic complaints arising from the physical sequelae of the miscarriage (or of the pregnancy). In addition, while one third of miscarrying women were hospitalized overnight and another third for 2:2 nights, hospital stay was not associated with symptom levels. Women who had dilatation and curettage procedures (approximately 85%) and those who did not also did not differ on level of depressive symptoms. Finally, women with early and late miscarriages reported similarly elevated symptom levels. For example, 34.1 % of women with first-trimester miscarriages (12 completed weeks' gestation) were highly symptomatic, as were 39.8% of women with later loss (p < 0.40).
Comment Among miscarrying women depressive symptoms are elevated fourfold over those in the community women in the second week after miscarriage and 3.4-fold over those in pregnant women. These results document for the first time the magnitude of the early effect of miscarriage on maternal psychologic status. Levels remain high at 6 weeks and at 6 months among women first interviewed at these time points but not among the reinterviewed. This difference in symptom levels between women first interviewed at 6 weeks and 6 months and those being reinterviewed suggests a strong ther-
apeutic effect of study interviews. However, alternative explanations for this finding need to be considered first. Selection bias in the form of selective departure or selective deferment of participation by more depressed study subjects is unlikely to contribute substantially to these results. First, as to selective departure, women interviewed at 2 weeks or at 6 weeks after loss who thereafter left the study did not differ on symptom levels from women who remained. Second, as for selective deferment, adjusted scores on the depressed mood scale administered at the time of the case-control interview did not differ among women first interviewed at 2 weeks, 6 weeks, and 6 months. Furthermore, women first interviewed at 2 weeks and 6 weeks had nearly identical symptom levels at 6 months (13.6 and 14.1, respectively). Similarly, in a large community survey of depressive symptoms, subjects interviewed in the calendar week targeted for study assessment and those deferring to a later time point did not differ on CESD symptom levels. 22 The lower scores of reinterviewed women are more easily explained as a consequence of unintended benefits of study participation, together with a possible test effect arising from scale repetition. Our initial interview afforded each subject an opportunity to discuss her pregnancy, the circumstances of the miscarriage, and its psychologic aftermath. In this respect it inadvertently resembled grief counseling. 7• 23 The efficacy of such counseling, often propounded, has been rigorously demonstrated in at least one randomized, controlled trial. 23 After a perinatal death women in the treatment group received a modal number of two counseling sessions. At 6 months after loss, the rate of psychiatric disorder in the treated group was one quarter that in the untreated group (12.5% and 52.6%, re-
Volume 166 Number I, Part 1
spectively). The first interview with miscarrying women in our study lasted from 1.5 to 3 hours and was generally given in two parts on separate days. Our estimates of a twofold to threefold elevation in depressive symptoms at 6 weeks and 6 months after loss were based on miscarrying women receiving their first interview; these estimates accord well with the previous controlled investigation of reproductive loss. Clarke and Williams" compared depressive symptoms in women experiencing perinatal death with depressive symptoms in women having live births. (The study protocol did not involve extended personal interviews with subjects.) At 6 weeks and 6 months thereafter the proportions of depressed women were 2.9 and 2.4 times those of women with live births, respectively. The results from the study of grief counseling cited above 2' also argue for at least a fourfold elevation in psychiatric symptoms at 6 months after loss in the absence of intervention. Given the preceding evidence based on both our own data and those of these other reports, we take symptom levels at first interview at each time to represent a better estimate of the true magnitude of the effect of miscarriage than levels at reinterview. Previous authors have suggested that the "conspiracy of silence'" surrounding reproductive loss on the part of hospital staff, family, and friends exacerbates depressive reactions. It would not be at all surprising therefore if our interview, with its direct focus on the miscarriage, proved to be psychotherapeutic. Although we had anticipated some drop in the subjects' reported symptoms as a result of study interviews, the magnitude of the observed decline was unexpected. The efficacy of this proposed method of intervention requires rigorous a priori testing. Nonetheless , on the basis of these initial findings, structured telephone interviews administered by lay personnel should have strong clinical and budgetary appeal for medical institutions unable to offer formal bereavement counseling. These results should also prompt obstetricians to meet with a miscarrying patient soon after loss simply to afford her a potentially therapeutic opportunity to review the circumstances of the pregnancy and the miscarriage. REFERENCES 1. Kline], Stein Z, Susser M. Conception to birth: epidemiology of prenatal development. New York: Oxford University Press, 1989:43-68. 2. Phipps S. Mourning response and intervention in stillbirth: an alternative genetic counseling approach. Soc Bioi 1981 ;28:1-13.
Depression after miscarriage
109
3. Stirtzinger R, Robinson GE. The psychologic effects of spontaneous abortion. Can Med Assoc] 1989; 140:799801,805. 4. Morris D. Parental reactions to perinatal death. Proc R Soc Med 1976;69:33-4. 5. Lewis E. The management of stillbirth: coping with an unreality. Lancet 1976;2:619-20. 6. Elliott BA, Hein HA. Neonatal death; reRections for physicians. Pediatrics 1978;62:96-100. 7. Leppert PC, Pahlka BS. Grieving characteristics after spontaneous abortion: a management approach. Obstet Gynecol 1984;64: 119-22. 8. Seibel M, Graves WL. The psychological implications of spontaneous abortions.] Reprod Med 1980;25:161-5. 9. Friedman T, Gath D. The psychiatric consequences of spontaneous abortion. Br] Psychiatry 1989;155:810-3. 10. Toedter L], Lasker ]N, Alhadeff]M. The Perinatal Grief Scale: development and initial validation. Am ] Orthopsychiatry 1988;58:435-49. 11. Peppers LG, Knapp RJ. Maternal reactions to involuntary fetal/infant death. Psychiatry 1980;43:155-9. 12. Nicol MT, Tompkins]R, Campbell NA, Syme GJ. Maternal grieving response after perinatal death. Med] Aust 1986;144:287-9. 13. laRoche C, Lalinec-Michaud M, Engelsmann F, et al. Grief reactions to perinatal death-a follow-up study. Can ] Psychiatry 1984;29:14-9. 14. Graham MA, Thompson SC, Estrada M, Yonekura ML. Factors affecting psychological adjustment to a fetal death. AM] OSSTET GVNECOL 1987;157:254-7. 15. Rowe ], Clyman R, Green C, Mikkelsen C, Haight ], Ataide L. Follow-up families who experience a perinatal death. Pediatrics 1978;62:166-70. 16. Kline], Stein Z, Susser M, Warburton D. Induced abortion and the chromosomal characteristics of subsequent miscarriages (spontaneous abortions). Am ] Epidemiol 1986;123 : 1066-79. 17. Folstein MF, Luria R. Reliability, validity and clinical application of the Visual Analogue Mood Scale. Psychol Med 1973;3:479-86. 18. Radloff LS. The CES-D Scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1977;1:385-401. 19. Zuckerman B, Amaro H, Bauchner H, Cabral H. Depressive symptoms during pregnancy: relationship to poor health behaviors. AM ] OSSTET GVNECOL 1989; 160: 1107-11. 20. Aneshensel CS, Frerichs RR, Clark VA, Yokopenic PA. Measuring depression in the community: a comparison of telephone and personal interviews. Public Opinion Q 1982;46: 110-21. 21 . Boyd]H, Weissman MN, Thompson D, Myers]K. Screening for depression in a commumty sample. Arch Gen Psychiatry 1982;39: 1195-200. 22. Mebrane 1. On time and late respondents. Technical report, Center for Epidemiologic Studies, National Institute of Mental Health, 1973. 23. Forrest GC, Standish E, Baum ]D. Support after perinatal death : a study of support and counselling after perinatal bereavement. BM] 1982;285: 1475-9. 24. Clarke M, Williams AJ. Depression in women after perinatal death. Lancet 1979;2:916-7.