Risk of recurrent red‐cell transfusion in delivery: A nationwide longitudinal study

To investigate the risk of recurrent maternal red‐cell transfusion in delivery.

worldwide, 5 and blood loss after delivery is also a common indication for maternal red-cell transfusion.7][8][9] Although the mechanisms driving recurrent haemorrhage can be difficult to disentangle, it may be explained by the persistence of both endogenous and environmental predisposing factors. 10,11revious studies have investigated risk factors for red-cell transfusion in delivery, 3,12,13 and have found that they are similar to those associated with postpartum haemorrhage.There are studies that show an increased risk of red-cell transfusion in subsequent deliveries when transfused in the first delivery. 14,15owever, transfusion recurrence is difficult to study because it is a rare event.There is also a need for comprehensive high-quality records, preferably linking large-scale nationwide medical birth registers with controlled administrative health data.In this long-standing nationwide study, with excellent coverage of both maternal characteristics and comorbidity, as well as transfusion events, we analyse risk factors for and the risk of recurrent maternal transfusion in delivery.

| Data sources
A unique personal identification number is assigned to all Swedish citizens, which provides the basis for the linking of health and population registers. 16We based our analyses on data from the Swedish portion of the Scandinavian Donations and Transfusions (SCANDAT-3S) database, which -in addition to holding information on all blood donations and transfusions -contains data on all persons in Sweden who have undergone ABO blood group testing, and therefore all women who give birth.Data are available from 1968, and the database has complete coverage in Sweden from the mid-1990s.Its creation has been described previously. 17CANDAT-3S is linked with the Swedish Medical Birth Register, providing maternal demographic and medical data on parturients and offspring in all deliveries in Sweden from 1973 onwards.From the combined database we extracted demographic and gestational data, reproductive history and diagnoses (with the latter coded according to the International Classification of Diseases, version 10, ICD-10, Swedish modification) and interventions associated with the first and, if available, second and third registered deliveries in women having a delivery in Sweden between the years 2000 and 2017.
The Swedish National Patient Register has attained complete coverage in Sweden since the mid-1980s and comprises data on all hospitalisations, dates, discharge diagnoses and, from 2001 onwards, it also contains data from outpatient specialist care. 18In addition to the Medical Birth Register, we used the Patient Register to retrieve prespecified diagnoses (ICD-10) from inpatient care or outpatient specialist care in the 9 months before a registered delivery.We treated chronic illnesses (e.g.non-gestational diabetes, non-gestational hypertension, haemoglobinopathies, coagulation factor deficiency, von Willebrand disease, systemic lupus erythematosus or kidney disease) as being persistent in any following deliveries.Conversely, curable conditions (e.g.iron-deficiency anaemia) had to be diagnosed no earlier than 9 months before the delivery date to be considered in further analyses.
Data on emigration, immigration, death and country of birth were retrieved from registers kept by Statistics Sweden.Data on first-degree relatives were retrieved from the Multi-Generation Register. 19here was no patient participation in this study.The study was approved by the regional ethics committee in Stockholm (2018/167-31).

| Study population, outcome and statistical analysis
Women aged 18-49 years who had a first delivery of an offspring of at least 22 weeks of gestation from January 2000 to December 2017 were included in the study.Before January 2008, stillbirths before 28 weeks of gestation were not recorded in the register and are thus not included in the analyses.All deliveries to women with missing personal identification number, missing delivery dates or with a date of emigration prior to the first delivery were excluded.The birth of multiples was treated as one delivery.Unless stated, we did not exclude subsequent deliveries to women that were younger than 18 years for the first delivery, nor did we exclude prior deliveries to women who also had deliveries after the age of 50 years.
The main outcome was maternal peripartum red-cell transfusion, defined as the recorded administration of at least one red-cell unit within 1 day before delivery or a maximum of 7 days thereafter.
We presented the proportion of deliveries in which maternal red-cell transfusion occurred in the first, second or third delivery (Table 1).Data were stratified according to maternal age at delivery (categorised as <25, 25-39 or ≥40 years of age), body mass index (BMI, categorised as <25, 25-34 or ≥35 kg/m 2 ), country of birth (categorised as Swedish or non-Swedish), AB0 blood group (0 or non-0), hospital category (university, regional or local) and according to year of delivery (categorised as 2000-2006,  2007-2012 or 2013-2017).We also considered maternal comorbidity (i.e.insulin-dependent and non-insulindependent diabetes, non-gestational hypertensive disorder, iron-deficiency anaemia and other anaemias, sickle cell disease, thalassaemia, idiopathic thrombocytopenic purpura, von Willebrand disease, systemic lupus erythematosus and kidney failure, stages 1-4).We also considered gestational factors and comorbidity (Table S1) that included factors often known before delivery (i.e.preeclampsia, gestational hypertension, gestational diabetes, intrauterine fetal death, antepartum haemorrhage, multiple gestation, placental abnormality and preterm or postterm delivery) and factors revealed in labour (i.e.dystocia, birthweight greater than 4500 g, ablatio placentae and uterine inversion or rupture).We classified the onset of labour as three categories: spontaneous, induction or caesarean before onset of labour.Similarly, we classified the mode of delivery as spontaneous vaginal, emergency caesarean, caesarean before onset of labour or instrumental vaginal delivery (with vacuum extraction or, rarely, forceps).In twin delivery, we allowed more than one mode of delivery (i.e.mixed delivery).
To estimate the association with red-cell transfusion at a second delivery, we fitted logistic regression models with maternal red-cell transfusion in the second delivery as the outcome among women with a recorded first and second delivery (Table 2).We considered predictors related to a diagnosis of postpartum haemorrhage (including subtypes, such as uterine atony, retained placenta, obstetric laceration, unspecified and delayed haemorrhage or excessive blood loss after caesarean) in a previous delivery, with or without concomitant red-cell transfusion.We investigated the effect of obstetric red-cell transfusions in female siblings (categorised as female sibling with transfusion in delivery, female sibling with no recorded peripartum red-cell transfusion or no female sibling with delivery).We also calculated the odds ratio of maternal red-cell transfusion in a third delivery in women with two previous deliveries, in relation to transfusion status in the first and second deliveries (Table 3).
We examined the risk of recurrent transfusion in a second or third delivery in women with singleton pregnancies and red-cell transfusion in the first delivery, in relation to maternal gestational and non-gestational comorbidity, gestational factors and events in the previous delivery (Table 4).
Analyses were adjusted for maternal age at delivery, BMI, year of delivery, country of birth, blood group and hospital level.
All processing and analyses were conducted in SAS 9.4 (SAS Institute, Cary, NC, USA).

No. of transfused (% of all)/not transfused aOR (95% CI) a
Events in first delivery T A B L E 3 Odds ratios for transfusion in a third delivery in relation to history of red-cell transfusion in the first and second deliveries.

Second delivery
No transfusion 1802 (1.6%)   first delivery, maternal red-cell transfusion was slightly more common among women with advanced age (40-49 years), with a BMI of ≥25 kg/m 2 or born outside Sweden than in the cohort overall.Transfusion was also more common in deliveries to women with prenatally diagnosed non-gestational comorbidity (i.e.non-gestational diabetes, non-gestational hypertension, haemoglobinopathies, factor deficiency, von Willebrand disease, systemic lupus erythematosus and kidney disease) compared with women without such comorbidities.The details of gestational comorbidity stratified by first, second or third delivery are presented in Table S1.
Table 2 describes maternal red-cell transfusion in women with (at least) a second delivery, in relation to transfusion and diagnoses of haemorrhage in the first delivery.Women with no red-cell transfusion and no previous diagnosis of obstetric haemorrhage in the first delivery were regarded as a reference group, and among them maternal transfusion occurred in 1.7% in a second delivery.For a parturient with previous postpartum haemorrhage but no transfusion, the corresponding figure was 4.1% (adjusted odds ratio aOR 2.4, 95% CI 2.2-2.6).Among women with previous transfusion, 8.7% were transfused in a second delivery, and a previous diagnosis of haemorrhage did not affect the odds ratio of transfusion recurrence, compared with the reference group (aOR 5.4, 95% CI 4.9-6.0,without previous diagnosis of haemorrhage; aOR 5.4, 95% CI 5.0-5.8, with diagnosis of haemorrhage).
The presence of a red-cell transfusion in the first delivery increased the odds of transfusion in a second delivery, regardless of haemorrhage subtype in the first delivery, but this was most pronounced for women with previous atonic haemorrhage with transfusion (aOR 6.7, 95% CI 6.1-7.4).Women with a sister who had been transfused in any delivery had a slightly higher risk of transfusion in a second delivery, compared with women with a sister with no transfusion in any delivery (aOR 1.8, 95% CI 1.5-2.1).
Table 3 shows the occurrence of maternal transfusion among women who contributed three deliveries in the data set.Transfusion in a third delivery occurred in 5.3% (aOR 3.6, 95% CI 3.0-4.2) of women with transfusion in the first delivery and 8.4% (aOR 5.9, 95% CI 5.0-7.1) of women with transfusion in the second delivery, compared with 1.6% among women who had never previously received a transfusion.Among women with a transfusion in their first two deliveries, 15.5% were transfused again in a third delivery, which is equivalent to a tenfold excess risk compared with women with no previous transfusion (aOR 11.5, 95% CI 7.9-16.6).
Table 4 describes pregnancy and delivery-related risk factors of subsequent transfusion in second and third deliveries among women with red-cell transfusion in the first delivery.Having a diagnosis of pre-eclampsia, antepartum haemorrhage, placental abnormalities, preterm birth, dystocia, or emergency caesarean was associated with a 2.0 or higher odds ratio for maternal transfusion in second delivery, compared to women without such diagnoses.A corresponding comparison in the third delivery revealed risk factors such as diagnoses of iron-deficiency anaemia, antepartum haemorrhage, placenta praevia, preterm delivery, induction, birthweight of 4500 g or more and emergency caesarean.The highest odds ratios for a new transfusion were seen for placental abnormality (aOR 7.5, 95% CI 3.6-17.6)in a second delivery and placenta praevia (aOR 4.8, 95% CI 2.2-10.2) in a third delivery.These odds ratios should be treated with caution, because of the small number of events.

| Main findings and interpretation
The key finding of this large, long-standing study on the risk of maternal peripartum red-cell transfusion is the striking risk of recurrence.Among women with red-cell transfusion in the first two deliveries, the risk of being transfused in a third delivery increased more than tenfold compared with non-transfused women.
Our findings are in line with a previous study by Patterson et al., who studied a cohort of 358 384 singleton births between 2003 and 2012 in Australia.They found a relative risk of 4.9 of transfusion recurrence in a second birth, when compared with women who did not receive transfusion in the first delivery. 15We found that the odds ratio of a new transfusion in women with a previous transfusion in delivery to be approximately the same.
Although haemorrhage is a key risk factor for transfusion, we found that only approximately seven out of ten transfused women had a diagnosis of postpartum haemorrhage.In a recent Swedish study, authors found the ICD codes for postpartum haemorrhage (corresponding to a blood loss of more than 1000 mL) to have a sensitivity of (88.5%) and a specificity of (99.4%), suggesting a moderate sensitivity and excellent specificity. 20Our findings suggest that these diagnoses will not fully capture the drivers of maternal transfusion, either because of incomplete coding practices or because of the non-negligible prevalence of other risk factors in our population.
A Danish study conducted by Thams et al. found that women with postpartum haemorrhage and transfusion in the first delivery had an almost sevenfold recurrence risk in a second delivery, compared with women without previous haemorrhage or transfusion. 21Our findings were similar.The risk was elevated for red-cell transfusion in a second delivery for women with a diagnosis of haemorrhage in the first delivery, but it was substantially higher in women who had a diagnosis of haemorrhage with concomitant red-cell transfusion.Although it was most pronounced in women with transfusion and uterine atony in the first delivery, we found the risk of new transfusion after previous transfusion to be higher, regardless of what was recorded as the main reason for haemorrhage (i.e.atony, placental retention or laceration).This is in line with the findings of Oberg et al., who found an overall increased risk of the recurrence of postpartum haemorrhage across all haemorrhage subtypes. 9However, it is in contrast with the findings of Linde et al., who found significant differences in the recurrence risk for postpartum haemorrhage, where the strongest risk for recurrence was found for dystocia, retained placenta and atony. 8It should be stated that many of these diagnoses are interconnected, and that medical coding may not fully ref lect the complexity of a postpartum haemorrhage.Our findings may ref lect an underlying coagulopathy evident in delivery, regardless of haemorrhage subtype.As such, it seems important to recognize that previous red-cell transfusion signals an increased risk of new transfusion, whatever the indication for the previous transfusion might have been.
We found that women with a sister who was transfused in delivery were at a moderately increased risk of transfusion in a second delivery (aOR 1.8), independent of other known risk factors, and it seems likely that the driver of the increased risk of transfusion is not only persistent but also at least partly heritable.Indeed, Oberg et al. found significant familial clustering in postpartum haemorrhage after vaginal births in Sweden. 10Similarly, a genetic susceptibility for recurrent postpartum haemorrhage was also found in a Norwegian study. 11Our findings suggest that shared genetic and environmental circumstances may also play a role in the recurrence of transfusion.
Among women with singleton deliveries and a transfusion in the first birth, we found no single factor that was both common and specific for the risk of new transfusions.The drivers of significant haemorrhage (i.e.placental abnormality) were associated with the highest odds ratio for new transfusion, which was unsurprising.Further, many of the factors considered are not easily modifiable, and the extent to which modification would affect the risks is beyond the scope of this study.3][24] It is reasonable to assume that the true prevalence of iron deficiency among Swedish women is not reflected in the available registers.However, an effort to actively investigate and manage iron status in pregnant women with previous transfusion in delivery could be of clinical value when planning for subsequent delivery, considering the risk of recurrent transfusion.

| Strengths and limitations
The strength of our study is that it is a long-standing population-based contemporary cohort with excellent coverage of most covariates, such as diagnosis of haemorrhage, and complete coverage of red-cell transfusions. 17To our knowledge, this is the largest study examining the risk of recurrence of red-cell transfusion in up to three consecutive deliveries.With this said, this study also has weaknesses.There is likely to be a certain extent of misclassification of the covariates included in our analyses, arising both from coding practices and from underdiagnosis.In addition, the recorded diagnoses, even in high-quality registers, cannot fully explain the causal mechanisms behind maternal red-cell transfusion in delivery, which is why we do not explore the specific mechanisms further.Lastly, although the transfusion records from the SCANDAT-3S database are likely to be very accurate, there is no specific information on the indication for each transfusion, which means that we cannot formally rule out that the recorded transfusions may in some instances have been given for reasons that were unrelated to the delivery, but rather for other indications.

| CONCLUSION
In this study, we found that the risk of peripartum redcell transfusion is substantially higher in women with previous red-cell transfusion in delivery, compared with parturients without previous transfusion.The elevated risk in a second delivery is also present in women with a diagnosis of haemorrhage in the first delivery only, but is more pronounced when there is a concomitant red-cell transfusion.Physicians should consider that a previous transfusion in delivery, irrespective of the cause, warrants prelabour planning and active investigation into any potentially modifiable risk factors, such as iron deficiency or iron-deficiency anaemia.

AU T HOR C ON T R I BU T ION S
AB was responsible for the outline, planning, performance and analyses of the study, and writing the article.GE was responsible for the outline, planning and analyses of the study, and took part in writing the article.AS was responsible for the outline and analyses of the study, and took part writing the article.

AC K NO W L E D GE M E N T S
None.

F U N DI NG I N FOR M AT ION
The conduct of this study and the creation of the SCANDAT database was made possible by a grant to Dr Edgren from the Swedish Research Council (2017-01954).Dr Edgren is further supported by Region Stockholm (a clinical research appointment).Dr Anna Sandström is supported by Region Stockholm (a clinical research appointment).

C ON F L IC T OF I N T E R E S T S TAT E M E N T
The authors declare no conflict of interests.

DATA AVA I L A BI L I T Y S TAT E M E N T
The data that support the findings of this study are available, upon reasonable request, from the corresponding author.The data are not publicly available because of privacy and ethical restrictions.

T A B L E 2
Risk of red-cell transfusion in a second delivery among women with at least a first and second delivery, in relation to previous transfusion, postpartum haemorrhage diagnosis and obstetric transfusions in their sisters.

T A B L E 4
Risk of recurrent red-cell transfusion at delivery in subsequent singelton deliveries in women with red-cell transfusion in the first singleton delivery, in relation to gestational comorbidity and events.

Second delivery Third delivery No. of transfused deliveries (%) aOR (95% CI) a No. of transfused deliveries (%) aOR (95% CI) a
bVariables are binary and non-cases act as reference.c Mutually exclusive variable.