Materials and Methods
We used a dataset with mental health and perinatal data, Michigan's Medicaid Eligibility and Paid Claims records, to test our hypothesis. The data included the completely processed fee-for-service records beginning January 1, 1994 and ending December 31, 1996 for 526,692 women born from 1950–1983. The data had been encrypted to protect confidentiality, and The University of Michigan institutional review board approved this secondary analysis. Variables for this analysis came from International Classification of Diseases, 9th Revision (ICD-9)13 diagnostic codes from hospital episodes and some procedure coding gathered by standardized billing forms and hospital chart reviews. Age, ethnicity, and program code variables came from eligibility information. Those data were converted from the multitable relational database to single-spreadsheet variables for analysis using SPSS version 7.5 (SPSS, Inc., Chicago, IL).
Cases of posttraumatic stress disorder were identified by presence of the ICD-9 diagnostic code 309.81. The American Psychiatric Association1 diagnostic criteria for this code require (A) presence of a qualifying trauma, and (B) one symptom of intrusive reexperiencing, (C) three symptoms of avoidance and numbing, and (D) two symptoms of arousal, with symptom duration longer than 1 month and significant functional impairment. Female recipients of Medicaid who had inpatient or outpatient ICD-9 code 309.81 were selected (n = 2219). An equal number of records were extracted by random-case selection process among female recipients who had no psychiatric diagnostic codes to form a comparison group. Among the 4438 women, 1093 had at least one diagnostic code relating to pregnancy (ICD-9 codes 630–676) and no code for multiple gestation because many of the hypothesized complications are associated with multiple gestation. Diagnostic bivariate and regression tests to assess the effect of including censored data showed no significant differences between groups.14 Given the cross-sectional nature of this preliminary study, in which no longitudinal analyses were planned, data from those cases were retained. Thus, the analysis included 455 women (41.6%) in the posttraumatic stress disorder group and 638 women (58.4%) in the comparison group.
Three categories of variables were developed. Demographic variables included ethnicity, age, and program code (ie, reason for eligibility) in effect at the last pregnancy-related service. Bivariate analysis showed that being eligible for Medicaid because of disability correlated strongly with posttraumatic stress disorder diagnosis (Table 1). We eliminated the disability demographic variable because it violated the noncorrelated independent variables assumption of regression analysis.15 From those data we could not tell whether the disabled women had posttraumatic stress disorder subsequent to a disabling injury or illness or whether the psychiatric disorder was the cause of disability. If the latter was true, those cases represent a level of severity of posttraumatic stress that might affect generalizability. Psychosocial screening variables included drug dependence complicating pregnancy (code 648.3), mental disorder complicating pregnancy (code 648.4), and a proxy variable for abuse and assault combining v-codes for emergency room observations after rape (code v71.5) or inflicted injury (code v71.6), and current domestic violence and counseling visits relating to abuse (codes v61.11 and v61.21). Obstetric variables show that diagnosis did or did not occur in the record.
Bivariate comparisons are presented for descriptive purposes. Comparisons that were significant after Bonferroni correction for multiple comparisons16 are indicated. Given the exploratory nature of this analysis, focusing on variables that are significant by the stringent Bonferroni criterion (P < .002) would likely result in a type II error. Therefore, all potential variables were included in the initial regression analysis.
The logistic regression model constructed for this analysis estimated association of multiple demographic, psychosocial, and obstetric variables with the single, dichotomous grouping variable (diagnosed with posttraumatic stress disorder or not). That model (Table 2) resulted from backward elimination of nonsignificant obstetric variables from among conditions we hypothesized were associated with posttraumatic stress disorder and any potentially confounding factors, such as preeclampsia accounting for induction and gestational diabetes accounting for excessive fetal weight. The variables were entered in a forward stepwise progression15 so that differences between groups on the basis of obstetric conditions have already taken into account the women's demographic and psychosocial factors. Model efficacy was evaluated using lambda-p15 (cases incorrectly predicted subtracted from cases correctly predicted and divided by cases correctly predicted) to determine the proportion reduction in error of prediction from using this model compared with chance.
Among 526,629 female Medicaid recipients with fee-for-service data from 1994–1996, 104,287 (19.8%) had at least one mental health diagnostic code from among the ICD-9 codes 290–347. Among those women, 2219 (2.1%) had code 309.81 for posttraumatic stress disorder. That is a prevalence of 0.4%, which likely represents an underreporting of the disorder.2
Bivariate tests on demographic variables (Table 1) showed no group differences in age. There were more white women in the diagnosed group, although it cannot be ascertained from the data if that was caused by differences in exposure to trauma, vulnerability, or likelihood of clinician assigning the posttraumatic stress disorder diagnosis. In the posttraumatic stress disorder group, 18.9% of women were in a disability program, versus 1.6% in the comparison group. Among women with posttraumatic stress disorder diagnoses, 17.4% had hospital episodes in which posttraumatic stress disorder was the primary diagnosis, and 22% of them also had substance abuse diagnoses.
Groups differed significantly on all psychosocial proxies. Emergency room observations after alleged rape and inflicted injury, or counseling visits related to abuse had been recorded for 6.4% of the posttraumatic stress disorder group and 0.8% of the comparison group. Half of the 29 women with victim service codes had prior posttraumatic stress disorder codes. Maternity care providers noted that pregnancy was complicated by drug dependence (1.3% overall) or mental disorder (2.8% overall) more often among women in the posttraumatic stress disorder group, but relatively infrequently given the proportion of women in the analysis who had mental health diagnoses for posttraumatic stress disorder (41.6%) and substance abuse (22%) during the 3-year period.
In bivariate testing of the pregnancy variables (Table 3), diagnostic codes for hospital episodes of spontaneous abortion or excessive vomiting were significantly associated with posttraumatic stress disorder diagnosis and remained significant after Bonferroni correction (set at P < .002). At a marginal level of significance, ectopic pregnancy, preterm contractions, excessive fetal growth, and poor fetal growth all were more frequent in the posttraumatic stress disorder group. Pregnancy problems that were hypothesized to be unrelated to posttraumatic stress disorder, including gestational diabetes and preeclampsia, did not occur at significantly different rates between groups. There were no significant bivariate differences between groups for some conditions that were hypothesized to be related to oxytocin dysregulation, including postdate gestation and dysfunctional labor diagnostic and treatment codes, such as uterine inertia or cesarean delivery.
The logistic regression model (Table 2) found five of the hypothesized complications to be significantly associated with posttraumatic stress disorder, after controlling for demographic and psychosocial factors. Women who had excessive vomiting were most likely to have been diagnosed with posttraumatic stress disorder. Those who had ectopic pregnancies, hospital treatment for miscarriage, a code that indicated concern about macrosomia, or had hospital episodes for preterm contractions and eventually gave birth at term were also more likely to have been diagnosed with posttraumatic stress disorder.
Assessment of the final regression model's efficacy by lambda-p15 showed 11% improvement over chance when using the demographic and psychosocial screening variables alone to predict which group each case belonged in. When obstetric variables were added, proportional reduction in error of prediction improved to 19%.
Our findings suggest that women with posttraumatic stress disorder might be at higher risk for some physical pregnancy problems, which were predicted based on known behavioral and neuroendocrine sequelae of traumatic stress, including ectopic pregnancy, spontaneous abortion, hyperemesis, preterm contractions, and excessive fetal growth. Hypothesized labor differences were not confirmed in those data. No differences were found in complication rates that were not believed to be related to traumatic stress, such as preeclampsia or gestational diabetes.
A number of limitations of our study exist. First, results of this study cannot be generalized to insured women, although they might be relevant to the United States Medicaid population.17 Second, these existing service usage data appear to underreport posttraumatic stress disorder, which might cause underestimation of associations studied because undiagnosed women with posttraumatic stress disorder were included in the comparison group. It is also possible that only the most severely affected women were diagnosed. Third, service usage data provide no information about antecedent trauma. Women with such trauma exposures as car accidents or house fires might be less likely to be triggered by and become symptomatic during pregnancy, making it less likely that their childbearing would be disrupted by posttraumatic stress. Fourth, women with the posttraumatic stress disorder diagnostic code likely receive some treatment for the disorder that might decrease its effect on their pregnancies. Epidemiologic data suggest that it is likely that most disorder-diagnosed women were first exposed to abuse and affected by traumatic stress symptoms before pregnancy.3–7 However, given the 3-year time limit of those data, it is impossible to determine chronology with certainty. It is also possible that posttraumatic stress disorder developed in women subsequent to pregnancy. Generally, that source of error would introduce a conservative bias by diluting differences between groups because the disorder acquired after pregnancy could not logically cause complications. However, a condition such as ectopic pregnancy could be life-threatening and thus be a traumatic stressor causing a new incidence of the disorder. Post hoc determination of dates for posttraumatic stress disorder codes and ectopic pregnancy codes was done for the 45 women in the posttraumatic stress disorder group who had ectopic pregnancies. The posttraumatic stress disorder code predated the ectopic code for 32 women (71%). Posttraumatic stress disorder is a chronic recurrent condition,18 so it is likely that few, if any, of the remaining cases had new onset after ectopic pregnancy diagnoses, and thus are unlikely to significantly affect results. The question of whether traumatic stress results from some obstetric experiences is a separate question that also warrants study.
This preliminary study from existing data expands current thinking about the effect of stress in pregnancy by considering traumatic stress rather than life-event stressors, daily stress, or anxiety.19 It also differs from prior studies by considering specifically the complications that could be mediated by behavioral and neuroendocrine mechanisms among women with posttraumatic stress disorder, particularly those mediated by oxytocin.
Future research is needed to corroborate and extend our findings, examine causality, study biologic mechanisms, and identify posttraumatic stress disorder–related interventions. Prospective studies should measure trauma exposure and posttraumatic stress disorder with instruments validated for use with women.20,21 Posttraumatic stress disorder and childbearing are both complex phenomena with many likely mediating or moderating factors, such as current battering, socioeconomic stressors, life events, social support, and comorbid obstetric, medical, and psychiatric conditions. Additional moderators such as women's coping strategies, prior or current mental health care, and client-provider alliance in maternity care should also be considered.
The preliminary data presented here point toward potentially important clinical implications. Should women's health care providers screen for posttraumatic stress symptoms and the associated problems with eating, substance abuse, and high-risk sex, especially among women with histories of abuse or assault? There are treatments available for posttraumatic stress disorder.22 Treatment early enough to precede conception might lead to prevention of some risk behaviors and some childbearing morbidity. For gravidas with ideopathic hyperemesis or preterm contractions, posttraumatic stress disorder could be an additional differential diagnosis. Psychopharmacologic and psychologic interventions for posttraumatic stress might bring quick relief. For those women, an oxytocin antagonist might be the more effective tocolytic. Last but not least, getting mental health treatment for the mother before birth also might improve the well-being of her infant.
1. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed. Washington, DC: American Psychiatric Association, 1994.
2. Resnick HS, Kilpatrick DG, Dansky BS, Saunders BE, Best CL. Prevalence of civilian trauma and posttraumatic stress disorder in a representative national sample of women. J Consult Clin Psychol 1993;61:984–91.
3. Kessler RC, Sonnega A, Bromet E, Hughes M, Nelson C. Posttraumatic stress disorder in the National Comorbidity Survey. Arch Gen Psychiatry 1995;52:1048–60.
4. Breslau N, Davis GC, Andreski P, Peterson E. Traumatic events and posttraumatic stress disorder in an urban population of young adults. Arch Gen Psychiatry 1991;48:216–22.
5. Breslau N, Kessler RC, Chilcoat HD, Schultz LR, Davis GC, Andreski P. Trauma and posttraumatic stress disorder in the community: The 1996 Detroit Area Survey of Trauma. Arch Gen Psychiatry 1998;55:626–32.
6. Koss MP, Gidycz CA, Wisniewski N. The scope of rape: Incidence and prevalence of sexual aggression and victimization in a national sample of higher education students. J Consult Clin Psychol 1987;55:162–70.
7. Kilpatrick DG, Aciernro R, Saunders B, Resnick HS, Best CL, Schnurr PP. Risk factors for adolescent substance abuse and dependence: Data from a national sample. J Consult Clin Psychol 2000;68:19–30.
8. Bohn D, Holz K. Sequelae of abuse: Health effects of childhood sexual abuse, domestic battering, and rape. J Nurse Midwifery 1996;41:442–55.
9. Herman JL. Trauma and recovery: The aftermath of violence—from domestic abuse to political terror. New York: Basic Books, 1992.
10. Yehuda R. Sensitization of the hypothalamic-pituitary-adrenal axis in posttraumatic stress disorder. In: Yehuda R, McFarlane AC, eds. Psychobiology of posttraumatic stress disorder. Vol. 821. New York: The New York Academy of Sciences, 1997:57–75.
11. Liberzon I, Young EA. Effects of stress and glucocorticoids on CNS oxytocin receptor binding. Psychoneuroendocrinology 1997;22:411–22.
12. Bullock LF, McFarlane J. The birth-weight/battering connection. Am J Nursing 1990;89:1153–5.
13. Practice Management Information Corporation. International Classification of Diseases, 9th Revision, Clinical Modification. 5th ed. Los Angeles: Practice Management Information Corporation, 1998.
14. Little RJA, Rubin DB. Statistical analysis with missing data. New York: John Wiley & Sons, 1987.
15. Menard S. Applied logistic regression analysis. Thousand Oaks, California: Sage Publications, 1995.
16. Pedhazur EJ, Schmelkin LP. Measurement, design, and analysis:An integrated approach. Hillsdale, NJ: Lawrence Erlbaum Associates, 1991.
17. Winterbottom C, Liska DW, Obermaier KM. State-level databook on health care access and financing. 2nd ed & Suppl. Washington, DC: The Urban Institute, 1995.
18. Ronis DL, Bates EW, Garfein AJ, Built BK, Falcon SP, Liberzon I. Longitudinal patterns of care for patients with posttraumatic stress disorder. J Trauma Stress 1996;9:763–81.
19. Paarlberg KM, Vingerhoets AJJM, Passchier J, Dekker GA, Van Geijn HP. Psychosocial factors and pregnancy outcome: A review with emphasis on methodological issues. J Psychosom Res 1995; 39:563–95.
20. Wolfe J, Kimerling R. Gender issues in the assessment of posttraumatic stress disorder. In: Wilson JP, Keane TM, eds. Assessing psychological trauma and PTSD. New York: Guilford Press, 1997: 192–238.
21. Pelcovitz D, van der Kolk B, Roth S, Mandel F, Kaplan S, Resick P. Development of a criteria set and a structured interview for disorders of extreme stress (SIDES). J Trauma Stress 1997;10:3–16.
22. Foa EB, Keane TM, Friedman MJ, eds. Effective treatments for PTSD: Practice guidelines from the International Society for Traumatic Stress Studies. New York: Guilford, 2000.