Assessing the effect of risk factors on rates of obstetric transfusion over time using two methodological approaches

There has been a global movement to reduce the usage of red blood cell transfusions and focus on patient blood management to prevent avoidable blood loss [1]. In Australia, the overall usage of red blood cells has been decreasing [2]. This decrease has been attributed to programmes designed to standardise the use of blood products, such as the implementation of patient blood management guidelines. However, rates of obstetric transfusion were still increasing up to 2010 [3]. These transfusions, given to mothers around childbirth, are primarily given as treatment following postpartum haemorrhage (PPH) [3] and make up 3.8% of all red blood cells issued [4].

In 2015, the National Blood Authority of Australia released the Patient Blood Management Guidelines module for Obstetrics and Maternity [5]. These guidelines detailed evidence- and experience-based best practice for blood management specifically within the obstetric setting, including transfusion, detection and management of anaemia, use of recombinant activated factor VII, tranexamic acid, cell salvage and interventional radiology. Prior to this, there were no specific guidelines for obstetric transfusions in place in Australia. However, evidence for patient blood management was already in development and guidelines for Critical Bleeding and Massive Transfusion were published in 2011 [6].

The use of red blood cells in the obstetric setting is influenced by patient indications as well as clinical practice [7]. There are numerous risk factors contributing to the risk of obstetric transfusion. These include demographics and baseline health of the mother (such as maternal age, parity, previous caesarean section) as well as conditions and treatments occurring during the antenatal and peripartum period (induction of labour, instrumental delivery) [3], and these risk factors are changing over time. PPH rates have also increased [8]. While transfusion rates have been explored up until 2010, more recent trends in obstetric transfusion, and the extent to which maternal, pregnancy and birth factors may be influencing the changing rate of transfusion, are unknown.

This study aimed to assess the recent trend in red blood cell transfusion rates for mothers giving birth in hospitals in NSW, Australia, and, using two methodological approaches, the extent to which this trend can be explained by available maternal, pregnancy and birth characteristics.

There were 939,470 births in NSW hospitals between July 2005 and June 2015. Of these, 3811 (0.4%) births were excluded due to missing values for one or more risk factors. Among the remaining 935,659 births included in the study, mothers were found to have received an obstetric red blood cell transfusion in 14,275 (15 per 1000 births) births. This rate increased by 38.7% from 13 per 1000 births in 2005 to a peak of 17 per 1000 births in 2011 (increasing trend, p < 0.001), but then remained mostly stable until 2015, remaining at 17 per 1000 births for each year apart from a drop to 16 per 1000 births in 2012 (Fig. 1). The rate of postpartum haemorrhage followed a similar trend to that of obstetric transfusion (Fig. 1) up to 2011, but continued to increase after 2011. Overall, the PPH rate increased by 54% from 74 in 2005 to 114 per 1000 births in 2015 (p < 0.0001). Of mothers receiving obstetric transfusions, 76.7% were found to have had PPH (Table 1). The proportions of mothers with several risk factors were found to have an increasing trend over the study period including mothers over 40 years of age, mothers born overseas, previous obstetric transfusions, diabetes, forceps delivery and caesarean sections, both with and without labour (all p < 0.0001) (Table 1).

In unadjusted analysis, women who received obstetric transfusion were more likely to be giving birth to their first child, giving birth pre-term (< 37 weeks gestation) or giving birth to multiples compared with those who did not have a transfusion (Table 2). Those who had a previous PPH or obstetric transfusion, an inherited or acquired haematological or platelet disorder, a third or fourth degree perineal tear, an episiotomy, an antepartum haemorrhage, a large-for-gestational-age infant, any placental abnormality, higher parity (5+), induced labour or hypertension, were at increased risk of transfusion. Women from lower socioeconomic status areas were more likely to receive a transfusion than those from higher socioeconomic status areas. There was no difference in risk of transfusion between mothers who had a previous caesarean or previous uterine surgery and those who did not (Table 2).

From the models which included year, if the risk factors could account for the change in rates of transfusion over time, then the odds ratios for each year in the adjusted model would be closer to 1 than the crude odds ratios. Controlling for pre-pregnancy characteristics did not change the effect estimates for year of birth meaningfully when compared with the crude estimates from 2006 to 2010, but slightly larger changes were seen in later years; crude odds ratio (OR) 1.35 (95% confidence interval (CI) 1.22, 1.49) to adjusted OR 1.30 (95% CI 1.17, 1.43) in 2014 and crude OR 1.39 (95% CI 1.25,1.56) to adjusted OR 1.34 (95% CI 1.20,1.50) in 2015 (Table 2). Additional inclusion of pregnancy and birth characteristics saw a further reduction in odds ratios for each of the years. The crude OR for obstetric transfusion comparing 2015 with 2005 was 1.39 (95% CI 1.25, 1.56). This dropped slightly to 1.34 (95% CI 1.20, 1.50) after adjusting for the maternal and pre-pregnancy factors, and further to 1.29 (95% CI 1.15, 1.45) when controlling for all available covariates (Table 2).

The same pattern is shown graphically in Fig. 2 which presents the predicted and observed rate ratios relative to 2005. Considering only pre-pregnancy risk factors, the rate of obstetric transfusion was only expected to rise by 4.1% from 2005 to 2015 (Fig. 2). When including pregnancy and birth risk factors, the expected increase in obstetric transfusion was predicted to be 10.3%. However, there was a 38.7% observed increase in rate over the same period (Fig. 2). The period during which the largest increases in transfusion rate were predicted, in each model, was 2011 to 2015. This is the same period in which the observed rate did not increase.

We found that, among births in NSW hospitals, the rate of obstetric transfusions increased by 38.7% between 2005 and 2011, from 13 to 17 per 1000 births, but remained relatively stable from 2011 to 2015. The two methodological approaches used both revealed that the change in rate of obstetric transfusions could only be partly explained by the available known maternal and pregnancy risk factors. Method 1, comparing crude and adjusted odds ratios, showed that the year effect could not be removed by adjusting for maternal and pregnancy factors. In Method 2, pre-pregnancy risk factors predicted a small, 4.1%, increase in the obstetric transfusion rate. Inclusion of pregnancy and birth factors predicted a larger, 10.3% increase in the rate, but not to the extent that was observed. The majority (76.7%) of obstetric transfusions in the study period were found to coincide with a PPH diagnosis, the rate of which increased throughout the study period, from 74 to 114 per 1000 births.

The increase in PPH and transfusion rates observed up to 2011 is consistent with previous studies in NSW [3, 8]. In the USA, the rate of transfusion also increased from 1998 to 2011, however PPH rates remained stable over the same period. [21] The PPH rate increased in the Netherlands from 40 per 1000 births in 2003 to 66/1000 in 2011 but appeared to plateau from 2011 to 64/1000 in 2013 [22] and in Canada the rate increased by 21% from 51/1000 in 2003–2004 to 6.1/1000 births in 2009–2010 [15].

This is the first study to report a largely stable rate of obstetric transfusions in NSW since 2011. Given that transfusion is often used as a marker of severity of haemorrhage, the stable transfusion rate in the context of increasing PPH is an encouraging sign. However, the obstetric transfusion rate in NSW is still higher than observed in the United States of America where there were 7 transfusions per 1000 births between 1998 and 2011 [21].

We found that pre-pregnancy characteristics, while being important risk factors for transfusion, did not contribute greatly to the change in obstetric transfusion rates over time. Accounting for all the pre-pregnancy, and pregnancy and birth factors that were ascertained, explained a larger degree of the change in transfusion rate but still only predicted a quarter of the increase observed from 2005 to 2015. Interestingly, the largest predicted increases in transfusion, with each model, were seen between 2010 and 2015 when the observed rate remained stable. This means that the rate of obstetric transfusion did not increase during this period despite an overall increase in maternal risk factors for obstetric transfusion. Our findings indicate that changes in rates of obstetric transfusion are largely influenced by other unmeasured factors which may include clinical decision making, policy changes, and the unpredictable nature of PPH as it is understood that many women who experience PPH do not possess any identifiable risk factors [23]. The diagnosis of PPH, based on an estimation of blood loss, and the decision to transfuse are both dependent on perceptions and experience of the attending clinicians as well as the hospital practice and blood supply [7, 11]. Furthermore, the combination of factors prompting intervention, relationships between risk factors or changing management of factors such as induction of labour may have played a role in obstetric transfusion trends.

While the new Patient Blood Management guidelines for Obstetrics and Maternity were only released in 2015 [5], the general movement away from transfusion-based management to patient-based management had been in progress for a number of years prior. The first module of the guidelines for patient blood management in the context of Critical Bleeding and Massive Transfusion was released in 2011 and included some recommendations for maternal care in the area of massive transfusion [6]. This is the same year that the rates of transfusion in the study appeared to peak. The guidelines included treatment paths to reduce transfusion for any critically bleeding or trauma hospital patient and reflected the movement towards reduced use of blood products to reduce unnecessary risk associated with this exposure. Specific recommendations regarding pharmacotherapy include the use of recombinant activated Factor VII [6].

Alongside new guidelines, other treatments for PPH and anaemia have been approved by the Australian Therapeutic Goods Authority (ATGA) and came into clinical practice during the study period, which may have contributed to the stable obstetric transfusion rate. These include tranexamic acid, recombinant activated Factor VII and ferric carboxymaltose (Ferinject). Intravenous tranexamic acid, an antifibrinolytic agent, was approved by ATGA in 2010 to reduce blood loss in cardiac and orthopaedic surgical patients [24]. However, studies since 2001 suggest that tranexamic acid has a role in reducing bleeding and risk of transfusion in postpartum haemorrhage [25‐28], such that WHO updated its guideline for management of postpartum haemorrhage to include tranexamic acid in 2012 [29, 30]. Furthermore, the publication of the WOMBAT trial in 2017 and subsequent updates in PPH guidelines for tranexamic acid to be part of standard management, will potentially lead to a reduction in PPH and transfusion rates in the future [31]. Recombinant activated Factor VII is used in Australia to prevent, slow or stop bleeding [32, 33]. Ferric carboxymaltose, introduced in Australia in 2011, is an intravenous infusion given to treat iron deficiency that has advantages over previously available iron treatments including reduced side effects and easier administration [34, 35]. The increased use of such treatments could potentially have had an impact on the rate of transfusions administered to mothers in response to anaemia and haemorrhage; however, this information is not collected in routinely collected birth or hospital data. Further research into specific blood management treatments and variation in practice by hospital may better explain the changes in trends in transfusion over time and point to the potential for improvements in practice to encourage further safe reductions in the use of blood products.

The main strengths of our study were the size and completeness of the dataset as well as the high-quality linked data from both the perinatal and hospital data. All births in NSW hospitals were included in the study with only 0.4% being excluded from the analysis due to missing data. This allowed us to investigate rare risk factors and treatments. While PPH and transfusion codes have been found to be reliable, changes in ascertainment may play a small part in changes in reported rates. This study also used two approaches to answer the study question which provided similar results, strengthening the conclusions. There were, however, some key covariates that are not well reported in either dataset, including body mass index, a change in which over time may have contributed to the changing rate of transfusions. The hospital data did not include laboratory results such as haemoglobin, which may have helped indicate the appropriateness of red blood cell transfusion. We were also limited by our inability to include prior obstetric history of mothers before the availability of linked data in July 2001.While we have taken into account risk factors related to transfusion, particularly in the predicted versus observed analysis, we have assumed the relationship between these factors and risk of transfusion has not changed over time. However, induction of labour (for example) may be more or less likely to result in PPH and transfusion at the beginning compared to the end of our study period.

The rate of obstetric transfusion among NSW hospital-based births increased over the period of 2005 to 2011 but has since stabilised through to 2015. Meanwhile, the rate of postpartum haemorrhage continued to rise from 2005 through to 2015. The changes in rate of transfusion could only be partly explained by changes in maternal and pregnancy risk factors using each of the methods employed in this study. Further monitoring and investigation, especially following the implementation of specific patient blood management guidelines for obstetrics in 2015, will give deeper insight into these trends to assist policy makers and clinicians to improve best-practice administration of blood products in an obstetric setting.