Abstract
Background:
Globally, obstetric hemorrhage (OH) remains a major contributor to maternal morbidity and mortality. A pilot study reported high rates of peripartum blood transfusion in a sample of South African hospitals.
Objectives:
To assess transfusion practices in a sample of Eastern Cape hospitals in South Africa in order to evaluate generalizability of preceding study findings.
Methods:
A cross-sectional, hospital chart review was conducted of all deliveries at 3 large regional hospitals from February to June 2013. Additional clinical data were collected on all patients who sustained OH and/or were transfused.
Results:
A total of 7,234 women were enrolled in the study; 1,988 (27.5%) were HIV positive. Of the 767 HIV positive women with a CD4 count < 350 /dL, 86.0% were on full antiretroviral therapy (ART), 9.9% received prevention of mother to child transmission. Only 1.2% did not receive any ART. Bivariate analysis showed a significant association between transfusion and positive HIV status; however in a multivariate analysis limited to one hospital this association [OR 1.45; 95% CI 0.78–2.71] was not significant after controlling for OH, age category, mode of delivery, gestational age, parity and birth weight. The overall transfusion rate was 3.2% (1.5%−4.6%) with significant variability among hospitals in number of RBC units per transfused patient as well as per delivery.
Conclusion:
The findings confirm high rates of peripartum blood transfusion in South Africa. While likely ascribed to variability in practice and patient profile, regional variation in care and temporal improvement in HIV treatment should also be considered.
Keywords: postpartum hemorrhage, blood transfusion, resource-poor setting, South Africa, epidemiology
Introduction:
In South Africa, as is the case globally, obstetric hemorrhage (OH) remains a leading contributor to maternal mortality and morbidity despite being the focus of related research and health interventions.1–3 Delayed recognition of risk factors for OH as well as failure to provide effective peripartum care, directly and adversely affect maternal outcomes. This is most notable in resource-poor settings, particularly where rural geography, ineffective referral and transportation systems compound risk associated with obstetric emergencies.4,5 Consequently, both OH and high rates of blood transfusion prescribed for OH are frequently encountered in resource poor settings; this is the case evident in South Africa.6,7
Shortfall in the provision of blood transfusion is estimated to contribute to a quarter of maternal deaths in Sub-Saharan Africa8 Specific to South Africa, postpartum hemorrhage (PPH) continued to contribute significantly to maternal deaths over the last reported triennium (2011 – 2013).3 While data exist for OH and maternal mortality, data pertaining to blood use in this population are lacking. For example, despite availability of published transfusion guidelines in South Africa in the obstetric setting, compliance with these guidelines is largely unknown.9 Obstetric blood transfusion places a significant burden on local blood transfusion services in South Africa whereby improved data are necessary for planning. More specifically, an understanding of risk factors for peripartum transfusion can identify deficiencies of care thus informing rational intervention. Finally, ~30% of pregnant women in South Africa are HIV positive10 where associated antenatal anemia may place patients at risk of peripartum blood transfusion.
Despite strained transfusion inventories in South Africa, a previous study demonstrated high rates of blood transfusion as compared to high resource settings, despite comparable rates of OH.11 We sought to investigate the risk factors and currently employed management strategies underlying peripartum transfusion in semi-rural South Africa, thereby complementing the earlier study, which had been conducted in urban South African hospitals.11 We also gathered contemporary data on the proportion of the obstetric population that is HIV positive as well as antiretroviral treatment regimens in-use to evaluate the level of compliance with standard guidelines.
Materials and Methods:
Study Design and Subjects:
We conducted a cross-sectional, hospital chart review of all deliveries at 3 large regional hospitals in the Eastern Cape Province of South Africa: Dora Nginza Hospital (DNH) in Port Elizabeth, Frere Hospital (FH) in East London and Cecilia Makiwane Hospital (CMH) in Mdantsane. The study was conducted over 4.5 months (February to June 2013). Ethical approval was obtained from the University of California, San Francisco (UCSF), the South African National Blood Service (SANBS), the Eastern Cape Provincial Department of Health and RTI International. The study was conducted under an approved waiver of consent for the collection of retrospective data without personal identifiers.
All three hospitals are part of the public sector health care system in South Africa, which provides medical services to low-income and uninsured, predominantly Black (African) and Colored (mixed population group) patients. Each of the hospitals provides a broad range of routine obstetric services; each hospital also serves as a referral center for the surrounding district hospitals and midwife obstetric units.
The study adopted a similar approach to that used by the REDS-III Obstetric Hemorrhage Pilot Study.11 Limited data were collected on all peripartum obstetric patients who had an index hospitalization at a participating hospital. Comprehensive data were collected on the subset of patients who sustained OH and/or were transfused. Eligible patients were identified through daily review of the admission registers for the various wards and theatres coupled with direct communication with the local blood bank and ward staff. The data were collected using the Obstetric Hemorrhage Audit Tool (OHAT), which was adapted from the instrument used in the previous study.11 Volume of blood loss was not routinely recorded at CMH and FH. The data were subsequently transferred to RTI, the data-coordinating centre, for quality control and analysis.
Definitions.
We used the same definitions as those used in the previous study.11 “OH” was defined as any obstetric-related haemorrhage occurring in the peripartum period, the latter defined as 48 hours prior, during or post-delivery. In accordance with the World Health Organisation’s (WHO) definition, peripartum haemorrhage refers to blood loss ≥500mL for vaginal deliveries or ≥1000mL for Caesarean sections. All women who delivered at at least 26 weeks’ gestation were included in the study, independent of stillbirths or early neonatal deaths (pregnancy related haemorrhage prior to 26 weeks gestation is typically managed by gynaecology departments in South Africa). “Transfusion” was defined as having received a transfusion of any allogeneic blood product e.g. red blood cells (RBC), platelets, plasma and/or cryoprecipitate, however only RBC transfusions were used for calculating transfusion parameters. “Booked” refers to patients who had had access to antenatal care while “Unbooked” refers to those who had not accessed any antenatal care during the index pregnancy. The most recently measured hemoglobin prior to the delivery admission was recorded as the antenatal hemoglobin. Similarly, the pre-and post hemoglobin values refer to the last hemoglobin prior to and the first hemoglobin level after the transfusion, respectively. The delta hemoglobin is the difference between the post-and pre-hemoglobin levels and is used to calculate the hemoglobin increment per transfused RBC unit.
Both HIV status and HIV-related medication regimen were recorded. Prevention of mother to child transmission (PMTCT) refers to the use of limited antiretroviral mono-therapy during the antenatal period followed by additional ART peri-partum, administered specifically to mitigate against vertical HIV transmission. ART refers to the use of triple therapy anti-retroviral drugs.
Statistical Analysis.
Standard summary statistics were used to generate counts and percentages for categorical data and distributions for continuous variables. For categorical data, counts and percentages for single variables and combinations of variables were produced, using Chi-squared tests of significance. For the continuous data, distributions were examined individually and stratified by covariates using T-tests to test differences between means.
Mutlivariate modeling was limited to data from DNH and was conducted using logistic regression. A larger set of variables was initially considered. The model was refined using backwards elimination at the p=0.05 level to retain variables. Once a set of variables was identified, interactions were investigated. The models were subjected to the Hosmer-Lemeshow test for goodness of fit. SAS 9.3 (TS1M2) with enhanced analytic product SAS/STAT 12.1 (SAS Institute Inc 2011) was used for the data manipulation and analysis; R version 3.0.1 (2013–05-16)—“Good Sport” (R Core Team 2013) was used for the data visualization
Results:
A total of 7,234 women were enrolled in the study. (Table 1) The majority of the participants were aged 19 to 29 and 9.3% were 18 years or younger. The racial distribution was predominantly Black and Colored. The women were mostly gravida 1 or 2 and para 0 or 1. Over 95% of the women had received antenatal care. The majority of births were normal vaginal deliveries but there were high rates of caesarian section; 1.7% of births occurred before arrival at the hospital. Overall 27.5% of the patients were HIV positive and prevalence varied slightly by site.
Table 1.
Characteristics of the study population: all subjects and by hospital.
| All subjects N (%) | CMH N (%) | DNH N (%) | FH N (%) | |
|---|---|---|---|---|
| All subjects | 7,234 | 1,641 | 3,191 | 2,402 |
| Age | ||||
| <=18 | 676 (9.3) | 173 (10.5) | 281 (8.8) | 222 (9.2) |
| 19–24 | 2,298 (31.8) | 517 (31.5) | 996 (31.2) | 785 (32.7) |
| 25–29 | 1,763 (24.4) | 364 (22.2) | 792 (24.8) | 607 (25.3) |
| 30–34 | 1,358 (18.8) | 298 (18.2) | 622 (19.5) | 438 (18.2) |
| 35–39 | 797 (11.0) | 186 (11.3) | 355 (11.1) | 256 (10.7) |
| 40+ | 292 (4.0) | 88 (5.4) | 126 (3.9) | 78 (3.2) |
| Missing | 50 (0.7) | 15 (0.9) | 19 (0.6) | 16 (0.7) |
| Race | ||||
| Black | 6,149 (85.0) | 1,623 (98.9) | 2,392 (75.0) | 2,134 (88.8) |
| Colored | 867 (12.0) | 5 (0.3) | 712 (22.3) | 150 (6.2) |
| Asian | 19 (0.3) | 0 (0.0) | 10 (0.3) | 9 (0.4) |
| White | 80 (1.1) | 0 (0.0) | 19 (0.6) | 61 (2.5) |
| Missing | 119 (1.6) | 13 (0.8) | 58 (1.8) | 48 (2.0) |
| Gravida (incl. current pregnancy) | ||||
| 1 | 2,683 (37.1) | 653 (39.8) | 1,117 (35.0) | 913 (38.0) |
| 2 | 2,259 (31.2) | 488 (29.7) | 1,011 (31.7) | 760 (31.6) |
| 3 | 1,392 (19.2) | 306 (18.6) | 625 (19.6) | 461 (19.2) |
| 4+ | 851 (11.8) | 183 (11.2) | 414 (13.0) | 254 (10.6) |
| Missing | 49 (0.7) | 11 (0.7) | 24 (0.8) | 14 (0.6) |
| Para (before current delivery) | ||||
| 0 | 3,010 (41.6) | 717 (43.7) | 1,297 (40.6) | 996 (41.5) |
| 1 | 2,297 (31.8) | 487 (29.7) | 1,037 (32.5) | 773 (32.2) |
| 2 | 1,225 (16.9) | 274 (16.7) | 528 (16.5) | 423 (17.6) |
| 3 | 439 (6.1) | 108 (6.6) | 192 (6.0) | 139 (5.8) |
| 4+ | 210 (2.9) | 40 (2.4) | 113 (3.5) | 57 (2.4) |
| Missing | 53 (0.7) | 15 (0.9) | 24 (0.8) | 14 (0.6) |
| Prenatal visit | ||||
| Booked | 6,882 (95.1) | 1,554 (94.7) | 2,993 (93.8) | 2,335 (97.2) |
| Unbooked | 266 (3.7) | 72 (4.4) | 143 (4.5) | 51 (2.1) |
| Missing | 86 (1.2) | 15 (0.9) | 55 (1.7) | 16 (0.7) |
| Type of delivery | ||||
| Vaginal | 3,988 (55.1) | 756 (46.1) | 1,775 (55.6) | 1,457 (60.7) |
| C-section | 3,057 (42.3) | 811 (49.4) | 1,373 (43.0) | 873 (36.3) |
| Born Before Arrival (BBA) | 122 (1.7) | 64 (3.9) | 1 (0.0) | 57 (2.4) |
| Missing | 67 (0.9) | 10 (0.6) | 42 (1.3) | 15 (0.6) |
| HIV status | ||||
| Missing/unknown | 110 (1.5) | 25 (1.5) | 35 (1.1) | 50 (2.1) |
| Negative | 5,136 (71.0) | 1,109 (67.6) | 2,295 (71.9) | 1,732 (72.1) |
| Positive | 1,988 (27.5) | 507 (30.9) | 861 (27.0) | 620 (25.8) |
| CD4 = 350+ | 885 (12.2) | 241 (14.7) | 369 (11.6) | 275 (11.4) |
| CD4 = 200<350 | 463 (6.4) | 129 (7.9) | 197 (6.2) | 137 (5.7) |
| CD4 < 200 | 304 (4.2) | 84 (5.1) | 124 (3.9) | 96 (4.0) |
| CD4 missing/unknown | 336 (4.6) | 53 (3.2) | 171 (5.4) | 112 (4.7) |
Abbreviations:
CMH: Cecilia Makiwane Hospital
DNH: Dora Nginza Hospital
FH: Frere Hospital
A total of 234 (3.2%) women were transfused. Bivariate analysis demonstrated significant variability in transfusion incidence by hospital: FH (1.5%), DNH (3.8%) and CMH (4.6%) [p < 0.00001] (Table 2). Significantly increased transfusion rates were noted with increased parity as well as among unbooked patients (7.1%) versus booked patients (3.0%). Transfusion also varied significantly by birth weight; the highest rates occurred at the extremes of birth weight. Gestational age had a significant inverse association with transfusion.
Table 2.
Bivariate analysis of risk factors for transfusion among all women (Chi square p-values).
| No Transfusion N (%) | Transfusion N (%) | P-Value | |
|---|---|---|---|
| All subjects | 7,000 (96.8) | 234 (3.2) | |
| Antenatal Hb | <0.0001 | ||
| Mean: | 10.95 | 10.17 | |
| Median: | 11.0 | 10.5 | |
| Hospital | <0.0001 | ||
| CMH | 1,565 (95.4) | 76 (4.6) | |
| DNH | 3,070 (96.2) | 121 (3.8) | |
| FH | 2,365 (98.5) | 37 (1.5) | |
| Age | 0.8855 | ||
| <=18 | 652 (96.4) | 24 (3.6) | |
| 19–24 | 2,221 (96.6) | 77 (3.4) | |
| 25–29 | 1,712 (97.1) | 51 (2.9) | |
| 30–34 | 1,313 (96.7) | 45 (3.3) | |
| 35–39 | 773 (97.0) | 24 (3.0) | |
| 40+ | 285 (97.6) | 7 (2.4) | |
| Missing | 44 (88.0) | 6 (12.0) | |
| Race | 0.2259 | ||
| Black | 5,955 (96.8) | 194 (3.2) | |
| Colored | 834 (96.2) | 33 (3.8) | |
| Asian | 19 (100.0) | 0 (0.0) | |
| White | 80 (100.0) | 0 (0.0) | |
| Missing | 112 (94.1) | 7 (5.9) | |
| Gravida | 0.0789 | ||
| 1 | 2,600 (96.9) | 83 (3.1) | |
| 2 | 2,192 (97.0) | 67 (3.0) | |
| 3 | 1,354 (97.3) | 38 (2.7) | |
| 4+ | 812 (95.4) | 39 (4.6) | |
| Missing | 42 (85.7) | 7 (14.3) | |
| Para (before current delivery) | 0.0493 | ||
| 0 | 2,917 (96.9) | 93 (3.1) | |
| 1 | 2,229 (97.0) | 68 (3.0) | |
| 2 | 1,190 (97.1) | 35 (2.9) | |
| 3 | 423 (96.4) | 16 (3.6) | |
| 4+ | 196 (93.3) | 14 (6.7) | |
| Missing | 45 (84.9) | 8 (15.1) | |
| Prenatal visit | 0.0002 | ||
| Booked | 6,675 (97.0) | 207 (3.0) | |
| Unbooked | 247 (92.9) | 19 (7.1) | |
| Missing | 78 (90.7) | 8 (9.3) | |
| Type of delivery | 0.5197 | ||
| Vaginal | 3,865 (96.9) | 123 (3.1) | |
| C-section | 2,961 (96.9) | 96 (3.1) | |
| BBA | 116 (95.1) | 6 (4.9) | |
| Missing | 58 (86.6) | 9 (13.4) | |
| HIV status | 0.0301 | ||
| Missing/unknown | 98 (89.1) | 12 (10.9) | |
| Negative | 4,996 (97.3) | 140 (2.7) | |
| Positive | 1,906 (95.9) | 82 (4.1) | |
| CD4 missing/unknown | 324 (96.4) | 12 (3.6) | |
| CD4 <200 | 290 (95.4) | 14 (4.6) | |
| CD4 200<350 | 441 (95.2) | 22 (4.8) | |
| CD4 350+ | 851 (96.2) | 34 (3.8) | |
| Birth Weight | <0.0001 | ||
| ≤2,000g | 690 (93.8) | 46 (6.3) | |
| 2,001–2,450g | 698 (95.7) | 31 (4.3) | |
| 2,451–2,700g | 677 (97.0) | 21 (3.0) | |
| 2,701–2,900g | 701 (98.0) | 14 (2.0) | |
| 2,901–3,060g | 728 (98.5) | 11 (15) | |
| 3,051–3,200g | 750 (97.8) | 17 (2.2) | |
| 3,201–3,350g | 641 (97.1) | 19 (2.9) | |
| 3,351–3,510g | 676 (97.4) | 18 (2.6) | |
| 3,511–3,770g | 697 (97.8) | 16 (2.2) | |
| 3,771g+ | 675 (95.9) | 29 (4.1) | |
| Missing | 67 (84.8) | 12 (15.2) | |
| Gestational Age | <0.0001 | ||
| <33 weeks | 697 (93.8) | 46 (6.2) | |
| 34–36 weeks | 1,000 (95.4) | 48 (4.6) | |
| 37 weeks | 514 (97.9) | 11 (2.1) | |
| 38 weeks | 1,198 (96.5) | 44 (3.5) | |
| 39 weeks | 859 (97.9) | 18 (2.1) | |
| 40 weeks | 2,220 (98.1) | 42 (1.9) | |
| 41+ weeks | 323 (98.5) | 5 (1.5) | |
| Missing | 189 (90.4) | 20 (9.6) | |
In a multivariable analysis of data from DNH, the only hospital with recorded blood loss, OH was significantly associated with cesarean section [OR = 3.17; 95% CI 2.16 – 4.67], positive HIV status (OR = 1.64; 95% CI 1.12 – 2.39), gestational age <=34 weeks [OR = 2.63; 95% CI 1.24 – 5.55] and high birth weight (>3,770g) [OR = 3.09; 95% CI 1.35 – 7.06]. In a separate model of transfusion, which was also limited to DNH, only OH [OR = 62.14; 95% CI 35.90 – 107.54], and to a lesser degree, increased parity [4+ vs. 0, OR = 3.31; 95% CI 1.08 – 10.16] and decreased gestational age [34–36 weeks vs. 41+ weeks, OR = 4.32, 95% CI 1.96 – 9.49; 37 weeks vs. 41+ weeks, OR = 2.58, 95% CI 0.99 – 6.74; 38 vs. 41+ weeks, OR = 3.30, 95% CI 1.62 – 6.72] were associated with receipt of a blood transfusion. Cesarean section patients were less like to receive a transfusion after controlling for OH [OR 0.44; 95% CI 0.26 – 0.74]. Although bivariate analysis showed higher transfusion rates among HIV positive patients, this association [OR 1.45; 95% CI 0.78–2.71] was not significant after controlling for OH, age category, mode of delivery, birth weight, parity and gestational age.
Mean antenatal hemoglobin differed significantly (p < 0.001) between the respective hospitals. (Figure 1) Although there was not a significant difference in the mean pre-or post transfusion hemoglobin levels across hospitals (Table 3), there was a significant mean difference (p = 0.0302) in delta hemoglobin levels with DNH (1.08 g/dL), FH (1.45 g/dL) and CMH (1.99 g/dL). The mean number of RBC units both per transfused patient as well as per delivery differed significantly by hospital; transfused patients received an average of 1.93 units at DNH, 2.05 units at FH, and 2.61 units at CMH (p = 0.0079). CMH had the highest mean number of units transfused per delivery (0.12) followed by DNH (0.07) and then FH (0.03) (p < 0.0001).
Figure 1.
Haemoglobin distribution by hospital.
Table 3.
RBC transfusions and haemoglobin (Hb) values by hospital.
| CMH | DNH | FH | P-value | ||
|---|---|---|---|---|---|
| Antenatal Hb | N | 1,391 | 2,889 | 1,724 | |
| Mean (Median) | 10.89 (10.9) | 10.86 (10.8) | 11.05 (11.0) | <0.0001 | |
| Pre-transfusion Hb* | N | 57 | 118 | 32 | |
| Mean (Median) | 6.97 (6.6) | 7.55 (7.3) | 7.12 (6.7) | 0.1036 | |
| Post-transfusion Hb* | N | 57 | 118 | 32 | |
| Mean (Median) | 8.96 (8.7) | 8.64 (8.5) | 8.57 (8.4) | 0.3463 | |
| Delta Hb* (Post-Pre) | N | 57 | 118 | 32 | |
| Mean (Median) | 1.99 (2.2) | 1.08 (1.3) | 1.45 (1.7) | 0.0302 | |
| RBC units / transfused patient* | N | 76 | 121 | 37 | |
| Mean (Median) | 2.61 (2) | 1.93 (2) | 2.05 (2) | 0.0079 | |
| RBC units transfused / delivery | N | 1,641 | 3,191 | 2,402 | |
| Mean (Median) | 0.12 (0) | 0.07 (0) | 0.03 (0) | <0.0001 | |
Transfused patients only.
Among HIV-positive women, the CD4 count distribution did not differ between the hospitals [p = 0.97]. However, the hospitals differed in the proportion of women who received ART (p-value < 0.0001): of those patients with a CD4<350 (old guidelines) FH had the highest ART coverage (91.0% full ART; 6.4% PMTCT) followed by DNH (86.6% ART; 8.7% PMTCT) and CMH (79.8% ART; 15.5% PMTCT). (Table 4) In contrast, for women with CD4>=350, DNH (62.1%) had the highest ART coverage followed by CMH (45.2%) and FH (10.2%; p < 0.0001). HIV-negative women had a mean Hb of 11.03 g/dL as compared to 10.67 g/dL in HIV-positive women [p-value < 0.001] (Figure 2). Of the 234 women who were transfused, 82 (35.0%) were HIV-positive.
Table 4.
Characteristics of HIV-positive patients by hospital.
| CMH N (%) | DNH N (%) | FH N (%) | P-Value | |
|---|---|---|---|---|
| HIV positive patients | 507 | 861 | 620 | |
| CD4 lymphocyte count (cell/mm3): | 0.9748 | |||
| Missing | 53 (10.5) | 171 (19.9) | 112 (18.1) | |
| <200 | 84 (16.6) | 124 (14.4) | 96 (15.5) | |
| 200–349 | 129 (25.4) | 197 (22.9) | 137 (22.1) | |
| 350+ | 241 (47.5) | 369 (42.9) | 275 (44.4) | |
| HIV Treatment: (CD4 <350) | 0.0021 | |||
| Missing | 5 (2.3) | 11 (3.4) | 6 (2.5) | |
| ART | 170 (79.8) | 278 (86.6) | 212 (91.0) | |
| None | 5 (2.3) | 4 (1.2) | 0 (0.0) | |
| PMTCT | 33 (15.5) | 28 (8.7) | 15 (6.4) | |
|
HIV Treatment: (CD4 >=350) <0.0001 | ||||
| Missing | 7 (2.9) | 20 (5.4) | 4 (1.4) | |
| ART | 109 (45.2) | 229 (62.1) | 28 (10.2) | |
| None | 0 (0.0) | 9 (2.4) | 0 (0.0) | |
| PMTCT | 125 (51.9) | 111 (30.1) | 243 (88.4) | |
Figure 2.
Haemoglobin distribution by HIV status.
Discussion:
This study describes the incidence of peripartum transfusion at three hospitals in the Eastern Cape Province of South Africa. In contrast to the four hospitals that participated in the earlier OH pilot study, the Eastern Cape hospitals are comparatively under-resourced, have less academic support and are located in a more rural setting, lending greater generalizability to the findings offered by the preceding study.12 Estimated blood loss was incompletely documented at 2 of the 3 hospitals, precluding detailed analysis of OH. However, at DNH where OH was routinely recorded, OH incidence was 4.6% (data not shown) which is higher than that reported in both the USA as well as the first pilot study.11,13,14 Similarly, the overall transfusion rate was 3.2%, exceeding that reported in the previous study11 and markedly higher than that reported for the USA and other high resource settings.13,14 Furthermore, there was significant variation (1.5% to 4.6%) in the transfusion rates among the participating hospitals. While this could reflect differences in the patient populations, variability in clinical practice (individual and/or institutional) cannot be excluded. Variability in practice might also explain differences in HIV and ART coverage.
The variability in transfusion practice among the participating hospitals is likely multifactorial. As observed in other developing countries15, this variability may be ascribed in part due to differences in clinical practice both by individual physician as well as by institution. Clinical transfusion practice is influenced by the presence of transfusion policy and/or guidelines in addition to the extent to which they are mandated in the hospital. Current South African obstetric transfusion guidelines recommend antenatal transfusion for those patients with symptomatic anemia and/or with hemoglobin values below 7g/dL or those who are anemic at greater than 36 weeks gestation, independent of whether symptomatic and/or associated with higher hemoglobin values In addition, the guidelines recommend maintaining a hemoglobin above 8 g/dL in patients with OH.9 Importantly, the implementation of, and adherence to, transfusion guidelines can serve to reduce the number of transfusions without a concomitant increase in morbidity or mortality.16 Differences in the hospital population may also contribute to the variability in blood utilization. For example, a higher gravidity, parity and proportion of unbooked cases were noted at DNH and CMH as compared to FH. Similar variation in blood utilization by population has previously been reported in other settings.11 The impact of blood utilization and transfusion practice on patient outcomes in the obstetric setting is an active research focus.16
Several authors have suggested that the global increase in OH is due to an increased incidence of uterine atony. The latter is associated with increasing parity, high birth weight and cesarean sections14,17 all of which were notably also associated with increased rates of transfusion in our study. However, the association between increasing age, parity/gravidity and OH has not been consistently demonstrated14,17 as was the case in our study where increased transfusion rates were noted with increasing parity but not increasing age. Age and parity are themselves associated, which may account for the variable findings in regards to the association between increasing age, parity and OH noted in various publications.14,17 Lack of, or poor access to (or failure to access) antenatal care, in this instance denoted by “booked” vs. “unbooked” cases, is a known predictor of poor maternal and fetal outcomes18 and is reflected in our findings by the higher transfusion rate among “unbooked” patients. This further underscores the importance of antenatal care in the prevention of obstetric complications.
The study also showed that HIV remains highly prevalent among pregnant women in South Africa. Almost a third of our study patients were HIV-positive, of whom less than half (45%) had CD4 counts above 350 cells/mm3 and approximately 15% had CD4+ counts below 200 cells/mm3, reflecting a high burden of advanced HIV disease. The South African guidelines for the PMTCT changed during the course of the study, recommending full ART for all pregnant women, independent of viral load or CD4 count. Germane to our results, the majority of the patients (63%) were on full ART coverage with a larger proportion of the patients on full ART coverage as compared to the preceding pilot study, suggesting a rapid adoption of the new guidelines. At DNH, multivariable analysis showed that caesarean section, positive HIV status and high birth weight were independently associated with OH. However, after controlling for OH, the odds of transfusion in HIV positive patients –while similar to that of the preceding pilot study-did not attain statistical significance. This discrepancy could reflect the change to the South African guidelines; specifically, ART is associated in improvement of anemia which may blunt the effect of HIV on risk of transfusion. In addition, the relatively small number of participants included in the single hospital analysis may have limited power to detect a significant HIV effect.
In a meta-analysis of HIV and the risk of obstetric complications, Calvert et al, found that HIV-positive women had twice the odds of antepartum hemorrhage, but no association with postpartum hemorrhage.19 Early in the epidemic, cesarean section was recommended in patients who were HIV positive, potentially confounding an association given increased risk of bleeding with operative care. The significantly lower hemoglobin values observed in HIV-positive patients (also evident in the preceding study), reflects unaddressed antenatal anemia in this population. Even if not sufficiently severe to warrant transfusion, the low hemoglobin values, noted in patients at all three hospitals, confer increased risk of peripartum transfusion given that anemia renders those patients less able to tolerate normal blood loss associated with delivery.
The study had limitations. Foremost was the lack of routine recording of the estimated blood loss during delivery at two of the participating hospitals. Second, despite selecting three comparatively rural and less academic hospitals, this study still fails to provide insight into the rates of obstetric hemorrhage and associated transfusion practice at primary level hospitals. It is conceivable that OH rates may differ significantly in primary level hospitals, given existing barriers to antenatal and high-risk obstetric care.4,5 However, it would be logistically challenging to replicate this study at the primary care level given the smaller number of deliveries and therefore OH and transfusions at any one hospital. Because high-risk pregnancies are referred upward, the majority of transfusions in South Africa occur in secondary and tertiary hospitals, supporting our approach.
Despite the limitations, this study offers insight into variability in contemporary obstetric transfusion practice. It has contributed to the identification of antecedent risk factors for and the obstetric and current transfusion management of OH, enabling the implementation of systematic improvement in regional transfusion practice. The study also provides further validation of the methods and OHAT audit form such that similar studies might be undertaken in other settings. Finally, while the HIV clinical data confirms enduring, high rates of HIV in this population, it also suggests the successful implementation of the new HIV treatment guidelines.
In conclusion, this study further demonstrates the high rates of peripartum blood transfusion in South Africa despite rates of OH that approximate those reported in high-income countries. Blood inventories in South Africa are strained given the clinical demand (average blood reserve of 4.4 days; T. Brits, SANBS, personal communication, 2016) and the findings suggest that variability in transfusion practice is contributing to high rates of transfusion. Published national obstetric transfusion guidelines are available in South Africa; our findings suggest the need both to promote those guidelines (e.g. using focused training) as well as to review compliance. Knowledge of blood use is imperative to the South African transfusion services to ensure an adequate blood supply thus directly impacting a key Millennium Development Goal to reduce maternal mortality by 75%.
Acknowledgments:
The authors wish to thank Ms. Rita Solwandle and the staff of the Effective Care Research Unit for their invaluable contribution to the study. We are also grateful to the medical and nursing personnel at Dora Nginza Hospital, Cecilia Makiwane Hospital and Frere Hospital for their support.
Funding: National Heart Lung and Blood Institute Recipient Epidemiology and Donor Evaluation Study (REDS-III) research contract HHSN268201100009I.
Footnotes
Conflicts of Interests: None to disclose
References
- 1.Bates I, Chapotera GK, McKew S, van den Broek N. Maternal mortality in sub-Saharan Africa: the contribution of ineffective blood transfusion services. BJOG: An International Journal Of Obstetrics And Gynaecology. 2008;115(11):1331–1339. [DOI] [PubMed] [Google Scholar]
- 2.Homer C, Clements V, McDonnell N, Peek M, Sullivan E. Maternal mortality: what can we learn from stories of postpartum haemorrhage? Women And Birth: Journal Of The Australian College Of Midwives. 2009;22(3):97–104. [DOI] [PubMed] [Google Scholar]
- 3.National Committee for Confidential Enquiry into Maternal Deaths. Saving Mother 2011–2013: Sixth report on the Confidential Enquiries into Maternal Deaths in South Africa. Pretoria: South African National Department of Health;2015. [Google Scholar]
- 4.Kon ZR, Lackan N. Ethnic disparities in access to care in post-apartheid South Africa. American Journal of Public Health. 2008;98(12):2272–2277. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.McLaren ZM, Ardington C, Leibbrandt M. Distance decay and persistent health care disparities in South Africa. BMC Health Services Research. 2014;14(1):762–776. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Gandhi MN, Welz T, Ronsmans C. Severe acute maternal morbidity in rural South Africa. International Journal Of Gynaecology And Obstetrics: The Official Organ Of The International Federation Of Gynaecology And Obstetrics. 2004;87(2):180–187. [DOI] [PubMed] [Google Scholar]
- 7.Fawcus SR, van Coeverden de Groot HA, Isaacs S. A 50-year audit of maternal mortality in the Peninsula Maternal and Neonatal Service, Cape Town (1953–2002). BJOG: An International Journal Of Obstetrics And Gynaecology. 2005;112(9):1257–1263. [DOI] [PubMed] [Google Scholar]
- 8.Bates I, Chapotera GK, McKew S, van den Broek N. Maternal mortality in sub-Saharan Africa: the contribution of ineffective blood transfusion services. BJOG : an international journal of obstetrics and gynaecology. October 2008;115(11):1331–1339. [DOI] [PubMed] [Google Scholar]
- 9.Fawcus S, Moodley J, for the National Committee for Confidential Enquiry into Maternal Deaths. A Monograph of the Management of Postpartum Haemorrhage. Pretoria: South African National Department of Health;2010. [Google Scholar]
- 10.DOH RoSA. The 2011 National Antenatal Sentinel HIV & Syphilis Prevalence Survey in South Africa. 2012.
- 11.Bloch EM, Crookes RL, Hull J, et al. The impact of human immunodeficiency virus infection on obstetric hemorrhage and blood transfusion in South Africa. Transfusion. 2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Bloch EM, Crookes RL, Hull J, et al. The impact of human immunodeficiency virus infection on obstetric hemorrhage and blood transfusion in South Africa. Transfusion. March 16 2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Callaghan WM, Kuklina EV, Berg CJ. Trends in postpartum hemorrhage: United States, 1994–2006. Am J Obstet Gynecol. 2010;202(4):353.e351–356. doi: 310.1016/j.ajog.2010.1001.1011. [DOI] [PubMed] [Google Scholar]
- 14.Bateman BT, Berman MF, Riley LE, Leffert LR. The epidemiology of postpartum hemorrhage in a large, nationwide sample of deliveries. Anesthesia And Analgesia. 2010;110(5):1368–1373. [DOI] [PubMed] [Google Scholar]
- 15.Anorlu RI, Orakwe CO, Abudu OO, Akanmu AS. Uses and misuse of blood transfusion in obstetrics in Lagos, Nigeria. West African Journal Of Medicine. 2003;22(2):124–127. [DOI] [PubMed] [Google Scholar]
- 16.Prick BW, Jansen AJG, Steegers EAP, et al. Transfusion policy after severe postpartum haemorrhage: a randomised non-inferiority trial. BJOG: An International Journal Of Obstetrics And Gynaecology. 2014;121(8):1005–1014. [DOI] [PubMed] [Google Scholar]
- 17.Kramer MS, Berg C, Abenhaim H, et al. Incidence, risk factors, and temporal trends in severe postpartum hemorrhage. American journal of obstetrics and gynecology. 2013;209(5):449.e441–447. [DOI] [PubMed] [Google Scholar]
- 18.Talip Q, Theron G, Steyn W, Hall D. Total perinatally related losses at Tygerberg Hospital a comparison between 1986, 1993 and 2006. South African Medical Journal = Suid-Afrikaanse Tydskrif Vir Geneeskunde. 2010;100(4):250–253. [DOI] [PubMed] [Google Scholar]
- 19.Calvert C, Ronsmans C. HIV and the risk of direct obstetric complications: a systematic review and meta-analysis. PloS one. 2013;8(10):e74848. [DOI] [PMC free article] [PubMed] [Google Scholar]