Database

This study linked 2 nationwide population-based datasets. The first was the Taiwan National Health Insurance Research Dataset (NHIRD), which is available to scientists in Taiwan for research purposes. As of 2007, the NHIRD includes all medical claims data and registries for contracted medical facilities, board-certified specialists, and 22.6 million beneficiaries, which represent >98.4% of the island’s population. The NHIRD is 1 of the largest and most comprehensive nationwide population-based datasets available in the world, which offers a unique opportunity to examine how different types of anesthetic treatment for women who undergo CS affect the risk of PPH.

The second database was the national birth certificate registry that is supervised by the Ministry of the Interior in Taiwan. The data on birth certificates consists of both infant and parent birth dates, gestational week at birth, birthweight, sex, parity, place of birth, parental educational level, and maternal marital status. The mother’s unique personal identification numbers provided links between the NHIRD and birth certificate data, with assistance from the Bureau of Health Promotion, Department of Health, Taiwan. Because the NHIRD consists of deidentified secondary data that have been released to the public for research purposes, this study was exempt from full review by the Institutional Review Board.

Study sample

We identified 204,610 women who were pregnant and used prenatal care services between January 1, 2005, and December 31, 2005. Of these, 69,533 women had live singleton births by CS and were included in our study. We excluded those patients (n = 1609) who were diagnosed with antepartum hemorrhage (ICD-9-CM codes 641, 641.3, 641.8, and 641.9) or hemorrhage from placenta previa (ICD-9-CM code 641.1) to better examine the relationship between types of anesthetic management and PPH. We also excluded patients who took medications to stop contractions (n = 10 women) or to induce labor (exclusively includes patients who had no signs of labor and were over term or for whom there were some special medical reasons for induction; n = 586). Ultimately, 67,328 women were included in this study: 2433 women received general anesthesia, and the other 64,895 women received spinal/epidural anesthesia.

Variables of interest

In this study, the independent variable of interest was treated as a dichotomous category, according to whether a woman had received general anesthesia during delivery. The outcome variable was whether a woman had hemorrhage within 24 hours of delivery (ICD-9-CM codes 666.1, 666.10, 666.12, and 666.14). We did not include patients who had delayed (defined as hemorrhage that occurred >24 hours after giving birth) and secondary PPH (defined as excessive vaginal bleeding between 24 hours and 6 weeks after delivery; ICD-9-CM codes 666.2, 666.20, 666.22, and 666.24).

To assess the independent effect of different modes of anesthetic management on PPH, we also adjusted for a number of potential confounders that had been identified in previous studies in the regression modeling and included mother’s age and fetal parity. Parity consisted of 3 categories: 1, 2, and ≥3. We also took the following antepartum or intrapartum complications that may be related to PPH into consideration in the regression modeling: placenta previa without hemorrhage (ICD-9-CM code 641.01), placenta abruption (ICD-9-CM code 641.2), overdistended uterus (ICD-9-CM code 653.5), pregnancy-induced hypertension (ICD-9-CM code 642), prolonged labor (ICD-9-CM code 662), previous myomectomy (ICD-9-CM code 654.91), chorioamnionitis (ICD-9-CM code 658.4), placenta accreta (ICD-9-CM code 667.0), previous CS (ICD-9-CM code 654.21), and emergent CS.

Statistical analysis

We used the SAS statistical package (version 8.2; SAS System for Windows; SAS Institute, Cary, NC) to perform all analyses in this study. Pearson χ ² tests were used to examine differences between women who received general anesthesia and those who received spinal/epidural anesthesia. Therefore, multivariate logistic regression analyses were carried out to explore the risk of PPH among women who undergo CS. Crude and adjusted odds ratios (ORs) are presented with 95% CI. Variables that were significant at the level of < .1 in the bivariate analyses were recruited in the multivariate logistic regression models. A 2-sided probability value of < .05 was considered statistically significant for this study.