Maternal Risk Factors for Peripartum Transfusion

Introduction

Maternal morbidity and mortality have declined in the United States in the past century, but rates continue to be disparately high for some racial and ethnic groups. ^1,2 Postpartum hemorrhage remains one of the most significant maternal complications of childbirth in the United States occurring in more than 4%–6% of all deliveries, and postpartum transfusion is the most commonly identified severe morbidity. ³ In spite of a national trend toward conservative and supportive management for patients with excessive blood loss, recent data suggest that up to 1.6% of all parturients receive a blood transfusion during the peripartum period, increasing morbidity, length of stay, and healthcare costs. ^{3 –7} Several recent large population-based studies conducted in developed countries suggested that rates of postpartum hemorrhage have increased over time. ^8,9

Multiple risk factors for postpartum hemorrhage have been identified and include maternal and obstetric factors, such as parity, duration of labor, and delivery method, and race and ethnicity. ¹⁰ It is well established that the rate of peripartum transfusion is higher among women undergoing a cesarean delivery; within this group, multiple other factors add to the risk of transfusion, including antenatal anemia, the use of general anesthesia, gestational hypertensive disorders, placental abnormalities, chorioamnionitis, a multiple gestation pregnancy, and black race or Hispanic ethnicity. ¹¹ Less is known about the risks of transfusion for women delivering vaginally.

To identify potentially modifiable risk factors for transfusion, we studied 9 years of clinical labor and delivery records of a nearly population-based cohort of women delivering at a large regional community obstetric facility. We included both vaginal and cesarean delivery to identify potential synergistic effects among risks.

Materials and Methods

We conducted a retrospective cohort study of all women delivering between January 2000 and July 2008 at a large regional community hospital. The hospital system provides obstetric care to approximately 85% of deliveries in the region and reflects the demographic diversity of the state. All women delivering at 20 or more completed gestational weeks with a birth weight of ≥350 were included. Cases were excluded if they had missing data for maternal race/ethnicity, parity, age, gestational age at delivery, birth weight, a complete blood count (CBC) within 7 days before delivery, or if the birth weight fell outside of the standard range for the gestational age, suggesting data entry error. ¹² Cases were excluded from the study if there was a medical diagnosis of thalassemia or sickle cell crisis or if the platelet count at presentation for delivery was <100,000/μL. This was done to ensure we had excluded women with medical comorbidities that might be independently associated with coagulopathy, anemia, or transfusion. Approval was obtained from the Christiana Care Institutional Review Board before initiation.

The main outcome of interest was perinatal transfusion of blood products. This was identified by linking the obstetric data file to the blood bank database, also stored in the institutional data warehouse. To validate the transfusion marker in the database, 80 medical records were reviewed by two of the investigators (O.S.C. and M.L.C). On chart review, all transfusions were confirmed with the exception of one, providing us with a positive predictive value (PPV) of 97%. The reason for transfusion could not be identified by medical records review for a significant fraction of cases; therefore, these data were not included in this analysis. The absence of documentation is consistent with prior studies showing poor documentation of transfusion indications in medical records at other institutions. ⁷

The medical, obstetric, and neonatal data were derived from the institutional electronic obstetric record. This electronic medical record serves as the hospital record for the patient's labor and delivery course and has been described previously. ¹³ Information was recorded through direct entry by nursing staff during the patients' hospitalizations. Covariates in the analysis included the maternal demographic characteristics, marital status, insurance type, and whether the patient received care from a private practice physician or service physician. Obstetric factors included best clinical estimate of gestational age at delivery and diagnoses of prepregnancy diabetes and gestational diabetes as recorded in the medical record by the obstetric provider. Obesity was defined using prepregnancy weight and height, both of which were self-reported, and used to calculate the prepregnancy body mass index (BMI) in kg/m². ¹⁴ Method of delivery, use of labor induction, and birth weight were recorded in the electronic obstetric record by the clinical care team at the time of delivery.

CBC, routinely drawn on admission to labor and delivery, was extracted from the institutional data warehouse and linked to the obstetric data using a common identifier. The CBCs were time and date-stamped in the data warehouse. Only results stamped as occurring in the 7 days immediately before delivery were included; >93% were drawn the day of delivery, >96% within a day of delivery. If multiple CBCs had been drawn, the test most proximate to delivery was used. Anemia was defined by hemoglobin (Hgb) measurement in grams/deciliter (g/dL) and categorized as not anemic (Hgb ≥10.5 g/dL), anemic (Hgb ≤10.5 but >9.5 g/dL), and severely anemic (Hgb ≤9.5 g/dL). The American College of Obstetricians and Gynecologists (ACOG) defines anemia at term as Hgb <11.0 g/dL and during the second trimester as Hgb <10.5 g/dL. ¹⁵ We chose to use the 10.5 g/dL cutoff, consistent with the ACOG definition of anemia during the second trimester, to maintain the same category for the entire cohort regardless of gestational age at delivery. A second category of severe anemia was added because we anticipated a nonlinear association of Hgb with the odds of transfusion.

Results

We identified 60,916 deliveries: 35 cases were excluded for a birth weight outside the standard range, and 1,188 (1.9%) were excluded because of missing blood count within 7 days before delivery. In addition, 411 women with a diagnosis of sickle cell crisis or thalassemia or a platelet count <100,000 were excluded. This provided 59,282 (97.3%) cases for inclusion in the analysis.

Table 1 shows the characteristics of the women in the cohort as well as the bivariate association of social and demographic, obstetric, and medical risk factors with postpartum transfusion. A total of 614 women were identified as having received a transfusion, placing the overall incidence of transfusion for the cohort at 10.4/1,000 deliveries. Transfusion rates were significantly higher for black (14.1/1,000 deliveries) and lowest for white (8.4/1,000 deliveries) women. Maternal anemia at entry for delivery and cesarean delivery were each significantly associated with transfusion. Other factors associated with increased odds of transfusion were age ≥35 years, being unmarried, and not being privately insured. Also associated with transfusion were multiparity, placenta previa, placental abruption, fetal macrosomia, a multiple gestation pregnancy, prepregnancy or gestational diabetes, and having a hypertensive disorder. Although women with placenta previa had a lower mean Hgb than the group overall (11.3±1.3 g/dL vs. 12.0±1.2 g/dL, p<0.05), women with either placenta previa or placental abruption noted at delivery were not more likely to be anemic at entry.

Table 2 shows the results of the multivariate logistic regression stratified by delivery method and adjusted for the effects of maternal demographic characteristics, obstetric factors, and comorbidities while controlling for gestational age at delivery. For women with a vaginal delivery, in addition to anemia, factors that remained significantly associated with transfusion were race/ethnicity, insurance type, diabetes, primiparity, a hypertensive disorder, and a multiple gestation pregnancy. The increased risk associated with insurance type and black race was no longer significant; the increased risk associated with Asian race and Hispanic ethnicity remained.

Analysis of the interaction between anemia and cesarean delivery, within the full multivariable model, revealed heterogeneity of the effect of anemia by delivery method (Table 2), with a greater effect of anemia for women undergoing cesarean delivery. In Table 3 we show evidence of a joint effect between anemia and a cesarean delivery on the risk of a transfusion. This was a synergistic effect of the co-occurrence of anemia and cesarean delivery, which had an expected multiplicative joint effect of OR of 10.2 but an observed RR of 15.08 (11.84-19.21). The joint effect was unchanged when women with placenta previa or abruption were excluded from the analysis.

The PAF of transfusions associated with an Hgb <10.5 g/dL for women with a vaginal delivery was determined to be 22.7%, for women with a cesarean delivery 36.6%, and for the overall cohort 31.7%. This provides an estimate of the fraction of transfusions that might have been avoided if there were no women with anemia.

Discussion

Estimates of the incidence of postpartum hemorrhage are difficult to obtain using information contained in medical records or administrative databases; however, the transfusion of blood products is a validated and available marker of significant maternal morbidity. ¹⁷ We found the rate of perinatal transfusion therapy for a nearly population-based cohort of women at a large obstetric delivery hospital, all without known medical diagnoses associated with anemia or a coagulopathy, to be 10.4/1,000 deliveries. The maternal risk factors most strongly associated with transfusion were a cesarean delivery and the presence of anemia at the time of hospital admission for labor and delivery. These two factors behaved synergistically: the 30% of women undergoing a cesarean delivery who had an Hgb <10.5 g/dL at the time of delivery had a 15-fold increase in the risk of transfusion when compared to the risk for women without anemia who delivered vaginally.

A number of social and demographic characteristics were also associated with receipt of transfusion, although most were attenuated or eliminated after adjustment for confounding factors. The excess risk of transfusion observed for black women was able to be attributed to the higher rates of anemia in that group.

Our findings from data derived from clinical records are consistent with those from studies using administrative and clinical research data sources. Using the US National Inpatient Sample, Bateman et al. ⁸ identified age <20 or ≥40 years, cesarean delivery, gestational hypertensive disorders, polyhydramnios, chorioamnionitis, a multiple gestation pregnancy, retained placenta, and antepartum hemorrhage as significant risk factors for postpartum hemorrhage. An observational study by Rouse et al. ¹¹ found a transfusion rate of 32/1,000 deliveries and identified multiple factors, including black race, Hispanic ethnicity, multiple gestation pregnancy, gestational hypertensive disorders, chorioamnionitis, placental abruption, general anesthesia, prematurity, and anemia, to be significantly associated with odds of transfusion. Notably, they found 36% of women with a preoperative hematocrit <25% required transfusion after a primary cesarean delivery.

For the vast majority of women in developing and developed countries, anemia during pregnancy is the result of iron deficiency. ¹⁸ In the United States, 21.55 of every 1,000 women were found to have an Hgb <10 g/dL. ¹⁹ The prevalence of anemia varies significantly by race and ethnicity, socioeconomic status, and age and varies throughout the course of the pregnancy because of the effects of hemodilution during the second trimester. Non-Hispanic black women are twice as likely to be anemic as non-Hispanic white women; teens have the highest rates of anemia for all races. ¹⁹ A 2005 study of a low-income, mostly minority population found that rates of iron deficiency anemia increased through the course of pregnancy, with 1.8% in the first trimester, 8.2% in the second trimester, and 27.4% in the third trimester. ²⁰ The World Health Organization reported 12% of women and 18% of pregnant women from industrialized countries are anemic. ²¹

Iron deficiency anemia is an easily identified and treatable risk factor. ^18,22 Hgb levels are obtained routinely early in the prenatal period to screen for anemia, and when it is recognized before conception or early in the prenatal period, there is adequate time to intervene. ¹⁵ The United States Preventive Services Task Force recommends routine screening for anemia during pregnancy, and in 2011, the Institute of Medicine recommended assessment for anemia each trimester. ²³ Treatment of anemia can be through multiple approaches, most often starting with oral iron and vitamin C supplementation in combination with counseling about adequate dietary intake of iron. ²² An adequate iron balance during pregnancy requires body iron reserves of ≥500 mg at conception. Requirements for iron in the second half of gestation generally cannot be met solely through dietary iron, and supplementation is routinely recommended. ^24,25 Women who have anemia either before conception or early in pregnancy might benefit from follow-up to assess adherence and verify response. Women may stop taking iron because of discomfort from side effects or may not respond sufficiently to the dose given. If anemia persists, further steps are warranted: counseling about adherence, iron dose adjustment, or further evaluation for other causes of anemia. ²⁶ The use of new well-tolerated parenteral iron preparations to replenish stores in the rare patient who cannot tolerate oral iron should be considered. ²⁷ Moreover, follow-up during the postpartum period to assure resolution of anemia is needed. Women with short interpregnancy interval pregnancies and higher parity appear to be particularly at risk for iron deficiency. ²⁸

There are several important aspects of this study that potentially limit the validity of our findings and their applicability to other settings. It is possible that we were unable to identify a fraction of women who had placenta previa or placental abruption, which could have caused vaginal bleeding before admission and also provide an indication for cesarean delivery, thereby confounding the relation among anemia, cesarean delivery, and transfusion. However, these are uncommon complications at term, and we believe this is likely to have only a small impact on the overall study findings. Moreover, the data represent the experience of a single institution and may not reflect practices in other centers. For example, the threshold for transfusion may vary from one institution to another. Whereas there is no evidence-based precise guide, transfusion most frequently is triggered in our institution by a combination of estimated blood loss, whether the blood loss is ongoing, Hgb <7–8 g/dL, maternal vital signs, and urinary output. This is consistent with guidelines put forth by the American Society of Anesthesiologists and ACOG. ^15,29,30 Finally, finding of association does not prove causality and could be the result of confounding by other factors.

There are several important strengths of our study design that suggest these findings may be broadly applicable. The use of clinical obstetric records allowed the exclusion of women with confounding medical diagnoses that might themselves lead to transfusion of blood products. Additionally, the transfusion outcome was obtained from records provided by the institutional blood bank and was validated through chart review, enabling a reliable outcome. Finally, the cohort studied is a large community population reflecting the diversity of the obstetric population in the United States, increasing the generalizability of the findings.

In summary, we found multiple risk factors to be associated with maternal transfusion. Those with the greatest effect were cesarean delivery and antepartum anemia, which worked synergistically to increase the odds of transfusion. Cesarean delivery is a well-established risk factor for peripartum transfusion and of particular concern, given the rising number of primary and repeat cesarean deliveries. Early recognition of anemia with appropriate treatment may be increasingly important in this context. Greater attention to anemia may reduce the incidence of blood transfusion, thereby limiting exposure to transfusion-associated risks, increased length of stay, and cost of care. Addressing anemia during the preconception and interconception periods may be an ideal time to establish adequate iron stores before subsequent pregnancies. Future research is needed to determine optimal strategies to improve the management of anemia during the preconception, prenatal, and interconception periods and to assess its effectiveness in reducing rates of peripartum transfusion.