RESULTS
There were 5243 deliveries in 2018-2019 with a documented quantitative
blood loss. Our patient cohort characteristics are described in Table 1
(Table 1).
From our primary ranked analysis, we found that while there was a
relatively stable low OBH-M frequency in deciles 1 through 8, there was
a sharp increase from a mean frequency of OBH-M of three percent in
decile 9 (895-1201cc QBL) to thirteen percent in decile 10 (1205-8325cc
QBL) (Figure 1). The difference between decile 9 and 10 was found to be
statistically significant (p<0.001).
As the range of significant inflection in OBH-M appeared to be contained
within deciles 9 and 10, we then compared OBH-M incidence of deliveries
with a QBL greater than 750cc based on increments of 250cc QBL (Table
2). We found that deliveries with a QBL from 750-1000cc had a 2.4
percent rate of OBH-M with only a small increase to 3.7 percent for QBL
ranging between 1001-1500cc (Table 2). There was a demonstrable
inflection where the rate of OBH-M roughly doubled with every 250cc QBL
interval: 7.2 percent OBH-M for QBL 1501-1750cc, 12.5 percent OBH-M for
QBL 1751-2000cc, 25.0 percent OBH-M for QBL 2001-2250cc and 84 percent
OBH-M for QBL >2250cc.
We then compared this to the accepted definition of 1000cc for
hemorrhage to the threshold of 1500cc. The PPV for 1500cc was 20.5%,
compared to 9.8% for 1000cc (Table 3). With a sample prevalence of
2.0% for OBH-M within our cohort, there was an increased post-test
probability of association of blood loss with OBH-M by 12% after
increasing the threshold for hemorrhage to 1500cc. While there was no
statistical significance when the AUC of each definition was compared
with the DeLong test (p=0.104) (Table 3), the “new” definition is
clearly associated with an improved specificity at the cost of poorer
sensitivity compared to the standard definition [Appendix S3].