The study was designed as a case–control study with prospectively collected data. All women who gave birth before gestational week 37 during one year (2010) at Linköping University hospital were identified with the ICD-code for premature labour O60 in the computerized chart system Obstetrix®. All pregnancies and deliveries are registered in the chart system along with medical information and diagnosis during the antenatal period and admission to the delivery ward. The medical information of all study subjects were retrieved by manually inspecting the entries in Obstetrix® and transferring information to a spreadsheet. This was done by a senior medical student together with the first author (C L) in 2011. One hundred and sixty-eight women were diagnosed with preterm delivery and constitute the women from the index group. The control group consists of 172 women who gave birth after gestational week 37 at the same hospital during the same year. They were consecutively chosen from the chart system and matched for parity and age +/− 5 years. As for the study subjects who had delivered preterm, these individuals were retrieved by manually checking the entries in Obstetrix®.
The total study population comprised 332 women as eight women in the index group experienced an intrauterine fetal death (IUFD) before term and therefore were excluded.
Risk factors for preterm birth were scrutinized in each woman’s medical record. Age, marital status (married/cohabiting or single), sick leave at first visit with the midwife (yes or no), level of socioeconomics/education (higher education, lower education, unemployed/parental leave, student), BMI in the beginning of the pregnancy (−24.9, 25–29.9, 30-), use of tobacco products (yes or no), alcohol or drugs (yes or no), and parity (0,1, 2 or more previous deliveries). Earlier pregnancies were divided into groups with no previous preterm deliveries and one or more previous preterm delivery, previous and/or current diseases (yes or no) e.g., diabetes, asthma, uterus malformation, cervix insufficiency, conization, systemic lupus erythematosus, fibromyalgia, urinary tract infections/bacteriuria, pyelonephritis, genital infection, inflammatory bowel disease, hepatitis, hypothyroidism, hyperthyroidism, mediterranean fever, and congenital heart defects. SSRI use during pregnancy (yes or no), twin pregnancies (yes or no), amniotic fluid pathology examined with ultrasonography (divided into three groups: normal amount of amniotic fluid, polyhydramnios and oligohydramnios) clinical signs of chorioamnionitis (yes or no), reported or observed premature contractions (yes or no), observed cervical changes (yes or no), reported or observed vaginal bleeding (yes or no), anemia defined as Hemoglobin <100 (yes or no), observed preterm premature rupture of membranes (PPROM) (yes or no), small for gestational age (SGA) defined as a birth weight < −2 SD of the mean weight for the gestational length, observed congenital defects by a pediatrician (yes or no), observed ablation (yes or no), observed preeclampsia (yes or no), onset of labor (spontaneous or induced), mode of delivery (vaginal, vaccum extraction or cesarean section), and gestational week at birth were manually extracted from the records.
Maternal stress exposure was define as: having a psychiatric diagnosis according to medical records (depression, anxiety, fear of childbirth, psychosis, phobia and/or eating disorder) or if the patient’s medical record revealed self-reported stress. Self-reported maternal stress could be due to the pregnancy itself, previous traumatic experiences from pregnancy and childbirth, fear of congenital defects in the unborn child or injury. Lack of social support or a problematic relationship between the woman and her partner or with significant others as well as economic and work related problems were considered as stress exposure. These women were considered to have been exposed to stress during pregnancy.
Statistics
The statistical analyses were done with SPSS (version 18). The rejection of the null-hypothesis was set to 0.05 (two-sided) in all statistical analyses. Student t-test was used to test differences between quantitative variables. Pearson’s Chi-square test was used to for testing differences in frequencies between categories. The possible effects of confounders were estimated through multiple logistic regression analyses and the adjusted odds ratios (AOR) were presented with a 95 % confidence interval. In these models, preterm delivery was set as dependent variable and identified confounding factors were considered as independent variables. A variable theoretically causing both prematurity and maternal stress was considered a confounder. Of all the risk factors that were examined in this study; premature contractions, tobacco use, previous premature delivery, genital tract infection and twin pregnancy were considered as confounders. To verify these confounders correlations were computed. The correlations between stress and confounding factors were found to range between −0.062 and 0.268. Even though the correlations between stress and previous preterm delivery (ρ = 0.072), infection (ρ = 0.092), and twin pregnancy (ρ = −0.062) were low it was decided to keep this factors in the models based on previous theories on their importance with respect to preterm delivery [4].
Attributable risk (AR), was estimated in two different ways. AR1 = (AOR-1)/AOR, was used to estimate how many of the women exposed to stress during pregnancy had delivered preterm because of the exposure as an attributable risk factor. AR2 = AR1*case fraction, with case fraction standing for the proportion of women delivering preterm who were exposed to stress during pregnancy. AR2 estimates how many women who delivered preterm in our study population did so because of the exposure to stress during pregnancy as an attributable risk factor.
The present study was approved by the Ethical Review Board in Linköping, nr. 2011/183-31. Written informed consent for use of patient records in research is not required by the ethics committee standards.