Secondary Logo

Journal Logo

Featured Articles: Original Clinical Research Report

Introduction of a Novel System for Quantitating Blood Loss After Vaginal Delivery: A Retrospective Interrupted Time Series Analysis With Concurrent Control Group

Lumbreras-Marquez, Mario I. MBBS, MMSc*,†; Reale, Sharon C. MD; Carusi, Daniela A. MD, MS*; Robinson, Julian N. MD*; Scharf, Nora RN, MS; Fields, Kara G. MS; Farber, Michaela K. MD, MS

Author Information
doi: 10.1213/ANE.0000000000004560

Abstract

KEY POINTS

  • Question: Does quantitation of gravimetric and volumetric estimation of blood loss (tBL) increase the odds of postpartum hemorrhage (PPH) detection after vaginal delivery (VD) compared to visual estimation of blood loss (vBL)?
  • Findings: tBL implementation was associated with an increase in the odds of PPH detection in VDs, but this increase was not statistically greater than the increase observed in the cesarean delivery (CD) control series.
  • Meaning: Enhanced detection of PPH was achieved with the use of tBL, and evaluating the impact of earlier PPH detection is warranted in future studies.

See Article, p 854

Postpartum hemorrhage (PPH) is the most preventable cause of maternal mortality worldwide.1,2 The incidence of PPH after vaginal delivery (VD) ranges from 0.8% to 7.9%,3–6 and a recognized challenge of resuscitation is that detection of PPH can occur too late.7,8 More timely detection of abnormal bleeding prompts earlier recognition and subsequent treatment of PPH. Therefore, systems that promptly identify PPH after VD are warranted. The National Partnership for Maternal Safety (NPMS) published a Consensus Bundle on Obstetric Hemorrhage in 2015.9 A key NPMS bundle element for enhanced recognition is to adopt quantitative measurement of cumulative blood loss, given that blood loss is underestimated by 30%–50% using visual estimation, and the degree of underestimation increases with the magnitude of hemorrhage.10,11

Blood loss at the time of delivery is contained by both conical under-buttocks drapes and surgical sponges. Gravimetry is recommended by NPMS.9 However, the use of gravimetry is a multistep process requiring calculation and cumulative recording, which is challenging to perform during an acute bleeding situation. In addition, the use of gravimetry does not incorporate blood loss contained in the conical under-buttocks drape. In the present study, we have utilized a device for gravimetric and volumetric estimation of blood loss (tBL) that tracks cumulative blood loss over time based on gravimetry plus conical drape blood volume, subtracting sponge dry weight and amniotic fluid volume (Triton Labor and delivery [L&D], Gauss Surgical, Inc, Los Altos, CA).

Previous studies have compared visual estimation of blood loss (vBL) and gravimetry or calibrated drapes and report up to 30% underestimation by vBL.10,12,13 In addition, a study comparing a calibrated drape to gravimetry after VD reported no difference in PPH-related management or interventions.14 However, the impact of tBL after VD using a system that combines volumetric and gravimetric measurements is unknown. We hypothesized that the transition from vBL to tBL would be associated with increased odds of immediate PPH (≥500 mL blood loss) detection after VD. The primary aim of this study was to compare the change in odds of immediate PPH detection in VDs before and after implementation of the tBL device, using cesarean deliveries (CDs) as a concurrent control group. Secondary outcomes included blood loss ≥1000 mL, total blood loss, secondary uterotonic use, and a composite outcome related to PPH management (transfusion, vasopressor administration, and/or surgical procedures) before and after tBL implementation.

METHODS

Study Population

This was a single-center, retrospective, interrupted time series analysis with a concurrent control group. After institutional review board (IRB) approval and waiver of informed consent, all patients who had a singleton VD or CD at our institution between October 4 and December 31, 2017 (before tBL device implementation), and between February 1 and April 30, 2018 (after tBL device transition period), were identified. The tBL device was implemented for VDs only, so CDs were included as a control series that would likely also be affected by unofficial concurrent practice changes that could confound any observed association between tBL and immediate PPH detection in VDs.15–17 Patients were excluded if delivery blood loss data or any measured potential confounder (mother characteristics, baby characteristics, delivery details, or delivering physician or midwife) were missing. This article adheres to the applicable STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines.

tBL Device Implementation

On our L&D unit before tBL implementation, noncalibrated under-buttock drapes were used for all VDs, and obstetricians visually estimated blood loss. For tBL quantitation via the Triton L&D System, a calibrated under-buttock drape was introduced, and L&D nurses measured the tBL after VD. For CDs that occurred throughout the entire study period, anesthesiologists and obstetricians visually estimated blood loss. The Triton L&D System consists of an Apple iPad Pro 32GB Wi-Fi system linked by Bluetooth to a smart wireless scale. The iPad tracks cumulative blood loss over time by gravimetry, calculating sequential sponge weights on the scale and subtracting pre-entered sponge dry weights. In addition, the iPad has an input feature to enter initial amniotic fluid volume (ie, the initial fluid measured up to delivery of the neonate), which is subtracted from the final volume of fluid recorded from a calibrated conical under-buttocks drape. Of note, before the implementation of the device, the amniotic fluid volume was not recorded. Gravimetric and volumetric data are collected in real time for comprehensive and dynamic blood loss quantitation. An in-service training period from January 1 to 31, 2018, ensured that all L&D nurses, obstetricians, and anesthesiologists were oriented to the use of tBL as a standard of care.

Exposure of Interest and Potential Confounders

Our main exposure of interest was time period (ie, after versus before completion of transition from vBL to tBL). VDs with no evidence of tBL device use after transition were excluded from the primary analysis (31.9% [302/947]) but were included in a sensitivity analysis, as described in the Statistical Analysis section. Additional covariates of interest included date of delivery, delivery method, mother characteristics (age, body mass index [BMI], predelivery hematocrit [HCT], parity, and race), baby characteristics (gestational age and birth weight), and delivering clinician.

Patient Care

All patients admitted to L&D for VD or CD received standard of care during labor, delivery, and postpartum, and there were no system changes to obstetric care at our institution during the study periods. Our pre-established protocol for PPH after VD includes the activation of a “Stage 1 Variance” for blood loss exceeding 500 mL. A Stage 1 Variance directs the obstetrician/midwife and anesthesia teams to the bedside. The following interventions are considered: intravenous fluid resuscitation to maintain hemodynamic stability, additional intravenous access, secondary uterotonic administration, blood product transfusion, and discussion of whether to move the patient to the operating room for surgical intervention. All interventions related to PPH were initiated by the providing obstetric and anesthesia care teams in both groups, based on standard clinical criteria of hemodynamics (noninvasive blood pressure or arterial line, heart rate), coagulation tests (prothrombin time [PT], activated partial thromboplastin time [aPTT], fibrinogen, complete blood count [CBC]), and clinical etiology of bleeding in conjunction with vBL or tBL information. In the case of VD, providers were not blinded to tBL values. Management of all patients was based on the clinical judgment and decision-making of the medical team. While our L&D unit does not have a formal policy regarding active management of the third stage of labor, all providers are encouraged to use a prophylactic uterotonic, with oxytocin preferred.

Outcomes

The primary study outcome was detection of immediate PPH, defined as ≥500 mL vBL or tBL immediately after VD, or ≥1000 mL vBL immediately after CD. While the American College of Obstetricians and Gynecologists (ACOG) defines PPH for VD as a cumulative blood loss ≥1000 mL, we defined PPH as an immediate blood loss ≥500 mL for VD because this volume retains utility to alert providers about early variance, and remains a valuable clinical trigger for initiating interventions for PPH.18 Secondary outcomes were a blood loss ≥1000 mL (regardless of delivery method), total blood loss, secondary uterotonic administration, and a composite outcome related to PPH management interventions (transfusion requirement, vasopressor administration, and/or surgical procedures [uterine tamponade with Bakri balloon placement, dilation and curettage, perineal or vaginal hematoma evacuation, and/or hysterectomy]).19 Additional post hoc secondary outcomes assessed in the subgroup of patients who had HCT measured both before and after VD during their hospitalization were nadir HCT, postpartum HCT reduction ≥10%, and the difference between vBL or tBL and calculated blood loss (cBL). To assess accuracy of the tBL device, laboratory-based methodologies for blood loss measurement, such as in vitro rinsing or spectrophotometry, have low error rates of 0%–12% but are impractical to perform.20,21 Therefore, cBL was calculated in patients who had a VD utilizing a pregnancy-specific formula described by Stafford et al.11 The algorithm derives a cBL in obstetric patients by multiplying calculated pregnancy blood volume (defined as 0.75 × {[maternal height (inches) × 50] + [maternal weight in pounds × 25]}) by percentage of blood volume lost (defined as [predelivery HCT − postdelivery HCT]/predelivery HCT).

Predelivery CBC was obtained from all patients on admission to the L&D unit; postdelivery HCTs were defined as HCTs collected within the first 24 hours after delivery. Of note, we do not have an established protocol to determine which patients require a postpartum CBC; this was left to each provider’s discretion and clinical judgment. Nadir HCT (collected within 24 hours of delivery) was adjusted if the patient received packed red blood cell (PRBC) transfusion by subtracting 3% for every PRBC unit transfused.22

Statistical Analysis

Mother, baby, and delivery characteristics were compared within and across delivery methods (VD and CD) and time periods (after versus before transition from vBL to tBL device for VDs) using unweighted and weighted23,24 analysis of variance (ANOVA) and 2-sample t tests for continuous variables, and χ2 or Fisher exact tests for categorical variables. Weighting methodology is described in Supplemental Digital Content, Appendix, http://links.lww.com/AA/C969.

Weighted logistic and linear segmented regression were used to assess the change in levels of binary and continuous outcomes, respectively, right after the device transition period while accounting for previous temporal trends.15,16,25 The difference between VD and CD controls in the change in outcomes right after transition period completion (ie, the difference-in-difference) was estimated for immediate postdelivery blood loss measures (PPH, blood loss ≥1000 mL, and total blood loss) to obtain stronger inferences for these outcomes in which large effect sizes were hypothesized. The changes in postdelivery interventions and HCT-based outcomes right after transition period completion were only estimated for VDs due to the exploratory nature of these analyses. Parameters estimated in each segmented regression model were defined. The impact of device implementation on study outcomes was expected to occur immediately after the transition period, so the estimate of primary interest for immediate postdelivery blood loss was the difference between VDs versus CDs in change in outcome level right after transition end (β6). For postdelivery interventions and HCT-based outcomes, the estimate of primary interest was the change in level right after the transition for VDs (β2). Although not of primary interest, all segmented regression models included terms to estimate the change in outcome slopes before and after the end of the transition period to obtain more precise estimates of changes in outcome levels.16 Projected trends in PPH detection and blood loss after the transition period based on preimplementation observations were estimated using weighted logistic and linear regression models, respectively, with parameters for level at the beginning of the study for CD, slope pretransition for CD, the difference in level at the beginning of the study between VD versus CD, and the difference in the slope pretransition for VD versus CD. To account for the increase in variance in estimated blood loss amount for VDs after implementation of the device, natural log-transformed blood loss amount was modeled using a multilevel segmented linear regression model with separate unstructured covariance matrices fit for each group (before VD, after VD, before CD, after CD). First-order autocorrelation between the model-based residuals of adjacent observations was not detected based on the Durbin-Watson statistics for blood loss and HCT measurements (2.04 and 2.00, respectively), so models were not adjusted for autocorrelation.

The primary analysis excluded patients with vBL despite VD in the post device implementation period. To account for the possibility that the excluded patients did not represent a random sample of deliveries, a sensitivity analysis repeated all analyses with these patients included in the post device implementation VD group. While we adjusted postdelivery HCT measurements for blood transfusions in our primary analysis of HCT-based outcomes, a second sensitivity analysis reexamined these outcomes solely among patients included in the primary analysis who did not receive a blood transfusion.

All statistical hypothesis tests were 2-sided, with no correction for multiple testing. Statistical analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC).

An a priori power calculation determined that a minimum of 220 patients per blood loss measurement method would be required to detect a clinically meaningful 6.3 percentage point increase in the percentage of patients with immediate PPH detected (beyond 1.9% detection reported previously)26 with 80% power at a 2-sided α level of .05 using a χ2 test.

RESULTS

A total of 2486 singleton deliveries were included in the primary segmented regression analyses: 967 VDs pre-tBL device implementation (vBL), 645 VDs postimplementation (tBL), 456 CD preimplementation (vBL), and 418 CD postimplementation (vBL) (Figure 1). The tBL measurement device was used in 68.1% of singleton VDs in the postimplementation period (Figure 1). Before application of inverse probability of treatment weights to estimate the average treatment effect on the treated (IPTW-ATT),23,24 there were no differences detected in mother, baby, or delivery characteristics between time periods within each delivery method, but differences were detected in most mother and baby characteristics across all 4 groups (Table 1). After IPTW-ATT weighting, the only baseline difference between groups was a slightly higher proportion of patients who had a spontaneous VD (compared to vacuum and forceps) in the post- versus preimplementation VD groups (Supplemental Digital Content, Table 1, http://links.lww.com/AA/C969). A total of 105 clinicians performed VDs and/or CD during one or more of the study time periods, with 55 clinicians (52.4%) performing both VDs and CD during both study periods (Table 1).

Table 1. - Mother, Baby, and Delivery Characteristics for VD (Intervention Series) and CD (Control Series) Before and After Transition From Visual to Quantitative Blood Loss Estimation for Vaginal Births
VD CD All Groups
P a
VD tBL After Versus VD vBL Before
P a
CD vBL After Versus CD vBL Before
P a
vBL Before tBL After vBL Before vBL After
n = 967 n = 645 n = 456 n = 418
Age (years), mean (SD) 31.6 (4.9) 32.0 (4.9) 33.1 (5.0) 32.9 (5.0) <.001 .104 .617
Race, n (%) .018 .106 .999
 Asian 108 (11.2) 88 (13.6) 40 (8.8) 36 (8.6)
 African American 117 (12.1) 69 (10.7) 67 (14.7) 65 (15.6)
 Hispanic 62 (6.4) 26 (4.0) 22 (4.8) 20 (4.8)
 Caucasian 573 (59.3) 374 (58.0) 256 (56.1) 234 (56.0)
 Other 89 (9.2) 71 (11.0) 62 (13.6) 55 (13.2)
 Unknown 18 (1.9) 17 (2.6) 9 (2.0) 8 (1.9)
Delivery BMI (kg/m2), mean (SD) 26.6 (5.9) 26.3 (5.7) 28.1 (6.7) 28.3 (6.4) <.001 .361 .677
Spontaneous VD, n (%) 883 (91.3) 602 (93.3) .140
Manual removal of placenta, n (%) 16 (1.7) 9 (1.4) .680
Gestational age (weeks), mean (SD) 38.9 (1.6) 38.9 (1.8) 38.4 (1.9) 38.1 (2.2) <.001 .507 .057
Birth weight (g), mean (SD) 3292.8 (481.6) 3287.3 (513.5) 3239 (584.3) 3197.4 (649.4) .010 .826 .319
Predelivery HCT (%), mean (SD) 35.9 (3.1) 36 (3.1) 35.4 (3.4) 35.8 (3.2) .025 .594 .153
Predelivery HCT <30%, n (%) 29 (3.0) 18 (2.8) 21 (4.6) 19 (4.5) .197 .808 .966
Multiparous, n (%) 495 (51.2) 334 (51.8) 254 (55.7) 221 (52.9) .441 .815 .401
Anesthesia type, n (%) <.001 .084 .125
 CSE 16 (1.7) 2 (0.3) 32 (7.0) 29 (6.9)
 Spinal 5 (0.5) 3 (0.5) 263 (57.7) 234 (56.0)
 Epidural 809 (83.7) 563 (87.3) 158 (34.6) 142 (34.0)
 General 2 (0.4) 9 (2.2)
 Local 53 (5.5) 28 (4.3)
 Other 3 (0.3) 3 (0.5) 1 (0.2) 4 (1.0)
 None 81 (8.4) 46 (7.1)
Oxytocin administration, n (%) 964 (99.7) 644 (99.8) .654
Number of clinicians; and median (IQR) number of deliveries per clinician 89; 10 (4–17) 88; 5 (2–11) 60; 7 (4–10) 61; 5 (3–8)
A
bbreviations: BMI, body mass index; CD, cesarean deliveries (the control series); CSE, combined spinal and epidural anesthesia; HCT, hematocrit; IQR, interquartile range; SD, standard deviation; tBL, gravimetric and volumetric estimation of blood loss; vBL, visual estimation of blood loss; VD, vaginal deliveries (the intervention series).
a
P values correspond to analysis of variance and 2-sample t tests for continuous variables and χ2 or Fisher exact tests for categorical variables.

Figure 1.
Figure 1.:
Flow chart of patient inclusion. CD indicates cesarean delivery; singleton CD, singleton cesarean delivery (the control series); singleton VD, singleton VD (the intervention series); tBL, gravimetric and volumetric estimation of blood loss; vBL, visual estimation of blood loss; VD, vaginal delivery.

Table 2 shows summary statistics within each time period and weighted effect sizes from comparisons of outcome levels right after versus before completion of transition from vBL to tBL. All effect sizes and corresponding confidence intervals (CI) reported in Table 2 were obtained from transformation of estimates obtained from the IPTW-ATT segmented regression models reported in Table 3. The weighted odds ratio (wOR) of immediate PPH detection post-VD right after versus before completion of transition from vBL to tBL was 2.74 (95% CI, 1.39–5.41) (P = .004; Figure 2A). The wOR of immediate PPH detection after CD (estimated blood loss ≥1000 mL) right after versus before completion of the transition period was 1.43 (95% CI, 0.72–2.85) (P = .304; Figure 2B). A greater increase in the odds of PPH detection right after versus before completion of the transition period was not detected for VD versus CD (interaction wOR 1.91 [95% CI, 0.73–5.03; P = .190]). A difference in the trends (slopes) in the odds of PPH detection after versus before the end of the transition period was not detected for VDs (wOR = 1.02 [95% CI, 0.95–1.09; P =.616]) but was for CD (wOR = 1.07 [95% CI, 1.00–1.15; P =.050] [interaction P =.291]). The wOR of estimated blood loss ≥1000 mL right after versus before completion of the transition period was 1.19 (95% CI, 0.30–4.76; P =.806) for VDs versus 1.43 (95% CI, 0.72–2.85; P = .304) for CD (interaction wOR = 0.83 [95% CI, 0.18–3.90; P = .812]). No differences were detected in geometric mean estimated blood loss for VDs or CD right after versus before completion of the transition period (Table 2; Figure 3A). For VDs in the predevice implementation period, 56.4% of the estimated blood loss was in the range of 300 to <500 mL, while for the postdevice implementation period, only 24.7% of the blood loss was in the same range (Figure 3B). No difference was detected in secondary uterotonic administration or PPH-related interventions right after versus before completion of the transition period amongst VDs (Tables 2−3).

Table 2. - Changes in Outcome Levels for VD (Intervention Series) and CD (Control Series) Immediately After Completion of Transition From Visual to Quantitative Blood Loss Estimation for Vaginal Births, Accounting for Pretransition Trends
Outcome VD CD VD Versus CD
vBL Before tBL After Effect Size (95% CI) P vBL Before vBL After Effect Size (95% CI) P Effect Size (95% CI) P
Immediate Postdelivery Blood Loss n = 967 n = 645 n = 456 n = 418
PPH, n (%) 111 (11.5) 173 (26.8) 2.74 (1.39–5.41)a .004 91 (20.0) 85 (20.3) 1.43 (0.72–2.85)a .304 1.91 (0.73–5.03)a .190
≥1000 mL blood loss, n (%) 20 (2.1) 43 (6.7) 1.19 (0.30–4.76)a .806 91 (20.0) 85 (20.3) 1.43 (0.72–2.85)a .304 0.83 (0.18–3.90)a .812
Blood loss (mL), geometric mean (SE) 307 (4) 290 (10) 0.98 (0.84–1.15)b .831 767 (10) 793 (11) 1.10 (0.99–1.21)b .081 0.90 (0.74–1.09)b .265
Postdelivery Interventions n = 967 n = 645
Secondary uterotonic use, n (%) 90 (9.3) 62 (9.6) 0.92 (0.42–2.04)c .838
PPH management interventions, n (%) 20 (2.1) 13 (2.0) 0.54 (0.11–2.68)c .450
 Blood transfusion, n (%) 13 (1.3) 8 (1.2)
 Vasopressor administration, n (%) 6 (0.6) 5 (0.8)
 Surgical management, n (%) 13 (1.3) 6 (0.9)
Hematocrit-Based Outcomes n = 160 n = 127
Hematocrit nadir (%), mean (SD) 28.5 (4.8) 29.9 (4.3) 1.1 (−1.6 to 3.8)d .421
Hematocrit drop, ≥10%, n (%) 50 (31.3) 27 (21.3) 0.63 (0.17–2.35) c .489
Difference between vBL or tBL and cBL (mL), mean (SD) −600 (596) −237 (522) 349 (13–684)d .042
P
PH management interventions: transfusion, vasopressor administration, and/or surgical procedures.
A
bbreviations: cBL, calculated blood loss; CD, cesarean deliveries (the control series); CI, confidence interval; IPTW-ATT, inverse probability of treatment weights to estimate the average treatment effect on the treated; PPH, postpartum hemorrhage defined as blood loss ≥500 mL for VDs and ≥1000 mL for CDs; SD, standard deviation; SE, standard error; tBL, gravimetric and volumetric estimation of blood loss; vBL, visual estimation of blood loss; VD, vaginal deliveries (the intervention series).
a
Weighted odds ratios and 95% CIs obtained using the combination of β2 and β6 for VD, β2 for CD, and β6 for VD versus CD from each of the corresponding IPTW-ATT-weighted segmented logistic regression models presented in Table 3.
b
Weighted ratios of geometric means and 95% CIs obtained using the combination of β2 and β6 for VD, β2 for CD, and β6 for VD versus CD from each of the corresponding IPTW-ATT–weighted segmented linear regression models presented in Table 3.
c
Weighted odds ratios and 95% CIs obtained using β2 from each of the corresponding IPTW-ATT–weighted segmented logistic regression models presented in Table 3.
d
Weighted differences in means and 95% CIs obtained using β2 from each of the corresponding IPTW-ATT–weighted segmented linear regression models presented in Table 3.

Table 3. - Changes in Levels and Trends in Blood Loss–Related Metrics for VD (Intervention Series) Versus CD (Control Series) After Transition From Visual to Quantitative Blood Loss Estimation for Vaginal Births
Immediate Postdelivery Blood Loss
PPH a ≥1000 mL Blood Loss a Blood Loss b (Log [mL])
Parameter β (95% CI) P β (95% CI) P β (95% CI) P
Level at beginning of study for CD, β0 −1.29 (−1.62 to −0.96) <.001 −1.29 (−1.62 to −0.96) <.001 6.64 (6.59–6.70) <.001
Slope before transition for CD, β1 −0.04 (−0.09 to 0.01) .115 −0.04 (−0.09 to 0.01) .115 0 (−0.01 to 0) .247
Change in level right after transition for CD, β2 0.36 (−0.33 to 1.05) .304 0.36 (−0.33 to 1.05) .304 0.09 (−0.01 to 019) .081
Change in slope (after transition minus before transition) for CD, β3 0.07 (0–0.14) .050 0.07 (0–0.14) .050 0 (−0.01 to 0.01) .732
Difference between VD versus CD in level at beginning of study, β4 −0.71 (−1.23 to −0.20) .007 −3.17 (−4.26 to −2.09) <.001 −0.90 (−0.97 to −0.82) <.001
Difference between VD versus CD in slope after transition, β5 0.03 (−0.04 to 0.11) .367 0.13 (−0.01 to 0.26) .064 0 (−0.01 to 0.01) .845
Difference between VD versus CD in change in level right after transition, β6 c 0.65 (−0.32 to 1.62) .190 −0.19 (−1.73 to 1.36) .812 −0.11 (−0.30 to 0.08) .265
Difference between VD versus CD in change in slope (after transition minus before transition), β7 d −0.05 (−0.15 to 0.04) .291 −0.13 (−0.29 to 0.02) .096 0 (−0.02 to 0.02) .979
Postdelivery Interventions
Secondary Uterotonic Use a PPH Management Interventions a
Parameter β (95% CI) P β (95% CI) P
Level at beginning of study for VD, β0 −2.36 (−2.80 to −1.91) <.001 −4.24 (−5.22 to −3.26) <.001
Slope pretransition for VD, β1 0 (−0.01 to 0.01) .686 0.06 (−0.06 to 0.18) .363
Change in level right after transition for VD, β2 −0.08 (−0.88 to 0.71) .838 −0.62 (−2.23 to 0.99) .450
Change in slope (after transition minus before transition) for VD, β3 e 0 (−0.01 to 0.01) .707 −0.07 (−0.25 to 0.11) .441
Hematocrit-Based Outcomes
Hematocrit Nadir f (%) Hematocrit Drop a , ≥10% Difference of vBL or tBL and cBL f (mL)
Parameter β (95% CI) P β (95% CI) P β (95% CI) P
Level at beginning of study for VD, β0 28.9 (27.4–30.5) <.001 −0.75 (−1.46 to −0.04) .038 −586 (−777 to −394) <.001
Slope pretransition for VD, β1 −0.1 (−0.3 to 0.1) .549 −0.01 (−0.10 to .09) .914 −2 (−27 to 22) .864
Change in level right after transition for VD, β2 1.1 (−1.6 to 3.8) .421 −0.46 (−1.78 to 0.85) .489 349 (13−684) .042
Change in slope (after transition minus before transition) for VD, β3 e 0.2 (−0.1 to 0.5) .274 −0.02 (−0.16 to 0.13) .826 7 (−29 to 43) .690
P
PH management interventions = transfusion, vasopressor administration, and/or surgical procedures.
Abbreviations:
β, segmented regression coefficients; cBL, calculated blood loss; CD, cesarean deliveries (the control series); CI, confidence interval; IPTW-ATT, inverse probability of treatment weights to estimate the average treatment effect on the treated; PPH, postpartum hemorrhage defined as blood loss ≥500 mL for VDs and ≥1000 mL for CDs; tBL, gravimetric and volumetric estimation of blood loss; vBL, visual estimation of blood loss; VD, vaginal deliveries (the treatment series).
a
“Level” refers to log odds of outcome, and “slope” refers to change in log odds of outcome per week as estimated using IPTW-ATT–weighted segmented logistic regression.
b
“Level” refers to log of outcome and “slope” refers to change in log of outcome per week as estimated using IPTW-ATT–weighted segmented linear regression.
c
Parameter of primary interest for immediate postdelivery blood loss outcomes. The difference (95% CI) in the change in the log odds of PPH immediately after the tBL transition period between VDs and CD was 0.65 (−0.32 to 1.62). This parameter was not estimated using analogous models for the postdelivery intervention and hematocrit-based outcomes due to the exploratory nature of these secondary outcome analyses.
d
Despite its relevance in many interrupted time series analyses, this parameter was not of main interest in this study because the impact of the blood loss quantitation device was expected to remain constant after the transition period. This parameter was estimated with the main goal of obtaining a more precise estimate of the parameter of main interest, β6.
e
Parameter of primary interest for postdelivery interventions and hematocrit-based outcomes.
f
“Level” refers to mean of outcome, and “slope” refers to change in mean outcome per week as estimated using IPTW-ATT–weighted segmented linear regression.

Figure 2.
Figure 2.:
Detection of immediate postpartum hemorrhage before and after the implementation of a device for gravimetric and volumetric estimation of blood loss after VD. The device was implemented on January 1, 2018 with a 1-mo transition period (weeks 14–17). The vertical axis is on a logit scale. Filled and hollow circles represent the percentage of VDs (A) and CDs (B), respectively, in which PPH was detected (blood loss ≥500 mL for VDs and ≥1000 mL for CDs) each week. Black and gray solid lines depict the preimplementation and postimplementation temporal trends in PPH detection for VDs and CDs, respectively, estimated from a weighted segmented logistic regression model. Black and gray dashed lines show the projected postimplementation temporal trend in PPH detection for VDs and CDs, respectively, predicted from a weighted logistic regression model developed using only preimplementation observations. For each delivery method, the difference in height between the solid and dashed lines right after the transition period is the change in the level (log odds) of PPH detection associated with device implementation. CD indicates cesarean delivery; PPH, postpartum hemorrhage; VD, vaginal delivery.
Figure 3.
Figure 3.:
Immediate postdelivery blood loss before and after the implementation of the device for gravimetric and volumetric estimation of blood loss after VD. The device was implemented on January 1, 2018 with a 1-mo transition period (weeks 14–17). Filled and hollow circles represent the geometric mean immediate postpartum blood loss for VDs and CDs, respectively, per week. Black and gray solid lines depict the temporal trends in geometric mean blood loss before and after implementation for VDs and CDs, respectively, estimated from a segmented linear regression model. Black and gray dashed lines show the projected postimplementation temporal trends in geometric mean blood loss for VDs and CDs, respectively, predicted from a linear regression model developed using only preimplementation observations. For each delivery method, the difference in height between the solid and dashed lines right after the transition period is the change in the level (geometric mean) of estimated blood loss associated with device implementation. B, Histograms of natural log-transformed estimated blood. CD after indicates cesarean deliveries, after device implementation; CD before, cesarean deliveries, before device implementation; VD after, vaginal deliveries, after device implementation; VD before, vaginal deliveries, before device implementation.

In the subgroup of VD patients for whom predelivery and postdelivery HCT values were available (127 patients in the tBL group [19.7%] and 160 patients in the vBL group [16.5%]), there was no difference detected in mean nadir HCT during the first 24 hours after delivery in the tBL group versus the vBL group (Tables 2−3). Moreover, there was no difference detected in the odds of HCT drop of 10% or more during the first 24 hours after delivery for the tBL group versus the vBL group (Tables 2−3). The mean difference between delivery blood loss and cBL was smaller in the tBL (mean ± standard deviation [SD]: −237 ± 522 mL) versus the vBL group ([−600 ± 596 mL] [weighted difference in means [wDM] [95% CI]: 349 mL [13–684]; P = .042]).

Sensitivity analyses including VDs both with and without evidence of tBL after device implementation (sensitivity analysis 1), as well as excluding patients who received a blood transfusion (sensitivity analysis 2), produced similar results to the primary analysis (Supplemental Digital Content, Table 2, http://links.lww.com/AA/C969).

DISCUSSION

Better quantitation of blood loss to determine early PPH is a priority for reducing maternal morbidity and mortality. Delayed or imprecise vBL occurs commonly27 and is a driver of preventable PPH morbidity.7,8 The NPMS recommends quantitative and cumulative measurement of blood loss.9 In our study, tBL implementation was associated with increased odds of identifying PPH after VD, a finding that is consistent with previous studies.10–13 However, this change was not greater than a concurrent increased odds of vBL ≥1000 mL in the CD control series. The failure to detect a difference between delivery methods was partly due to an unanticipated increase in the odds of PPH after CD despite lack of intervention. This may be explained in part by enhanced focus on vBL during CD by staff who also provide care during VDs (a Hawthorne effect).28

The wOR of detecting PPH by tBL versus vBL was consistently high for both the primary and sensitivity analyses but remained constant in the comparison using weighted geometric means. This suggests that a benefit of tBL is to overcome the clinician tendency to visually estimate any nonalarming blood loss as <500 mL. In our study, there was a preponderance of cases with vBL between 300 and 500 mL. While identification of blood loss ≥500 mL may not necessitate immediate intervention, this threshold may help identify patients who warrant closer monitoring for blood loss beyond the immediate postpartum period. Our post hoc subgroup comparison of the difference between vBL or tBL and 24-hour cBL showed a notably smaller difference between tBL and cBL, suggesting greater accuracy with tBL.

While we did not detect a change in the odds of postdelivery interventions between the vBL versus tBL periods, this study was not powered to test whether tBL impacts outcomes. A prior cluster-randomized trial of 25,381 VDs in 13 European countries reported no improvement or change in PPH outcomes using a calibrated collector bag after vaginal delivery compared to vBL.19 A trial of similar size and design would be required to test whether blood loss quantitated by volumetric plus gravimetric methods versus vBL impacts patient outcomes.

Our study demonstrated increased odds of PPH detection with a quantitative approach compared to vBL after VD. A study comparing blood loss measurements after CD by visual estimation, gravimetry, and quantitation utilizing a photographic analysis of blood-soaked sponges and suction canister contents revealed only weak correlation between any measuring modality and the postpartum hemoglobin values.29 However, the study was performed under conditions of low blood loss to investigate correlation between blood loss and postpartum hemoglobin, and not to assess a quantitative device for PPH detection or diagnosis. There are several conditions of our study which may explain why the odds of PPH detection was higher using the tBL method in our study. First, the under-buttocks VD drape utilized at our institution has historically been uncalibrated. Implementation of the tBL system required the introduction of drapes that are calibrated. Toledo et al13 demonstrated that vBL accuracy worsens with increased volume (41% underestimation at 2000 mL) and that the use of calibrated drapes reduces inaccuracy to <15% error at all volumes measured. Incorporating calibrated VD drapes in the tBL phase may have further enhanced blood loss quantitation compared to the vBL with an unmarked drape. Second, the clinical situation of elective, low-risk CD is relatively standardized compared to VD with varying PPH risk factors including duration of labor, oxytocin exposure, and lacerations.

There is an ongoing need for clinicians to evaluate how interventions may impact the incidence of PPH after VD. The rate of blood loss, percentage of blood volume lost, and evaluation of clinical parameters such as the perfusion and shock indices are all outcomes of interest.30 The use of refined tBL devices as in our study may have utility to enhance the dynamic comparison of blood loss after any intervention proposed in a research setting.

We acknowledge the limitations of this study. First, the study design was retrospective and we cannot exclude bias or residual confounding.28 Second, vBL methodology was not standardized. Third, time to intervention for PPH management was a variable we could not collect retrospectively, and timing of administration for agents such as secondary uterotonics can play a major role in final blood loss and associated morbidity. Fourth, although postdelivery HCTs were collected within 24 hours of delivery, the time interval was not controlled and additional factors such as fluid administration were not considered. Such variables may have impacted our postdelivery HCT and cBL calculation. Fifth, the compliance rate of 68.1% for the tBL device may have introduced selection bias in our study. However, a sensitivity analysis including VDs without evidence of tBL after device implementation produced similar results to the primary analysis. A postimplementation survey on barriers to use revealed challenges locating a device, finding devices uncharged, and rapid delivery with no time to get the device. Sixth, we did not compare the tBL device with manual volumetric and gravimetric quantification of blood loss. However, this was not the aim of the present study, and more research is needed to investigate if the tBL device is more accurate compared to manual volumetric and gravimetric measurements. Finally, this study compared vBL with noncalibrated under-buttock drapes against the combination of the device with calibrated under-buttock drapes, making it impossible to disentangle the influence of calibrated drapes versus the device itself on any differences in results.

In conclusion, this is the first study to our knowledge that evaluates a comprehensive tBL device in comparison to visual estimation for measuring blood loss after VD. Enhanced and more accurate detection of PPH was achieved with the use of tBL. No differences were detected in outcomes related to PPH in the setting of a highly resourced and proactive tertiary center. Whether a higher probability of PPH detection using tBL can improve maternal outcomes warrants further study. E

DISCLOSURES

Name: Mario I. Lumbreras-Marquez, MBBS, MMSc.

Contribution: This author helped with conception of the work, data collection, analysis, and interpretation; and helped to draft the work and revise it critically and approve the final version of the manuscript.

Name: Sharon C. Reale, MD.

Contribution: This author helped with data collection and interpretation; and helped to draft the work and revise it critically and approve the final version of the manuscript.

Name: Daniela A. Carusi, MD, MS.

Contribution: This author helped with conception of the work and data interpretation; and helped to draft the work and revise it critically and approve the final version of the manuscript.

Name: Julian N. Robinson, MD.

Contribution: This author helped with conception of the work and data interpretation; and helped to draft the work and revise it critically and approve the final version of the manuscript.

Name: Nora Scharf, RN, MS.

Contribution: This author helped with conception of the work and data interpretation; and helped to draft the work and revise it critically and approve the final version of the manuscript.

Name: Kara G. Fields, MS.

Contribution: This author helped with conception of the work, data collection, analysis, and interpretation; and helped to draft the work and revise it critically and approve the final version of the manuscript.

Name: Michaela K. Farber, MD, MS.

Contribution: This author helped with conception of the work, data collection, analysis, and interpretation; and helped to draft the work and revise it critically and approve the final version of the manuscript.

This manuscript was handled by: Jill M. Mhyre, MD.

FOOTNOTES

GLOSSARY

ACOG = = American College of Obstetricians and Gynecologists

ANOVA = = analysis of variance

aPTT = = activated partial thromboplastin time

BMI = = body mass index

CBC = = complete blood count

cBL = = calculated blood loss

CD = = cesarean delivery

CI = = confidence interval

CSE = = combined spinal and epidural anesthesia

HCT = = hematocrit

IPTW-ATT = = inverse probability of treatment weighting-average treatment effect in the treated

IQR = = interquartile range

IRB = = institutional review board

L&D = = labor and delivery

NPMS = = National Partnership for Maternal Safety

PPH = = postpartum hemorrhage

PRBC = = packed red blood cell

PT = = prothrombin time;

SD = = standard deviation

SE = = standard error

STROBE = = STrengthening the Reporting of OBservational studies in Epidemiology

tBL = = gravimetric and volumetric estimation of blood loss

vBL = = visual estimation of blood loss

VD = = vaginal delivery

wDM = = weighted difference in means

wOR = = weighted odds ratio

REFERENCES

1. Berg CJ, Harper MA, Atkinson SM, et al. Preventability of pregnancy-related deaths: results of a state-wide review. Obstet Gynecol. 2005;106:1228–1234.
2. Main EK, McCain CL, Morton CH, Holtby S, Lawton ES. Pregnancy-related mortality in California: causes, characteristics, and improvement opportunities. Obstet Gynecol. 2015;125:938–947.
3. Mehrabadi A, Hutcheon JA, Lee L, Kramer MS, Liston RM, Joseph KS. Epidemiological investigation of a temporal increase in atonic postpartum haemorrhage: a population-based retrospective cohort study. BJOG. 2013;120:853–862.
4. Kramer MS, Berg C, Abenhaim H, et al. Incidence, risk factors, and temporal trends in severe postpartum hemorrhage. Am J Obstet Gynecol. 2013;209:449.e1–449.e7.
5. Lutomski JE, Byrne BM, Devane D, Greene RA. Increasing trends in atonic postpartum haemorrhage in Ireland: an 11-year population-based cohort study. BJOG. 2012;119:306–314.
6. Al-Zirqi I, Vangen S, Forsen L, Stray-Pedersen B. Prevalence and risk factors of severe obstetric haemorrhage. BJOG. 2008;115:1265–1272.
7. Della Torre M, Kilpatrick SJ, Hibbard JU, et al. Assessing preventability for obstetric hemorrhage. Am J Perinatol. 2011;28:753–760.
8. Khan KS, Wojdyla D, Say L, Gülmezoglu AM, Van Look PF. WHO analysis of causes of maternal death: a systematic review. Lancet. 2006;367:1066–1074.
9. Main EK, Goffman D, Scavone BM, et al.; National Partnership for Maternal Safety; Council on Patient Safety in Women’s Health Care. National partnership for maternal safety: consensus bundle on obstetric hemorrhage. Obstet Gynecol. 2015;126:155–162.
10. Patel A, Goudar SS, Geller SE, et al. Drape estimation vs visual assessment for estimating postpartum hemorrhage. Int J Gynaecol Obstet. 2006;93:220–224.
11. Stafford I, Dildy GA, Clark SL, Belfort MA. Visually estimated and calculated blood loss in vaginal and cesarean delivery. Am J Obstet Gynecol. 2008;199:519.e1–519.e7.
12. Al Kadri HM, Al Anazi BK, Tamim HM. Visual estimation versus gravimetric measurement of postpartum blood loss: a prospective cohort study. Arch Gynecol Obstet. 2011;283:1207–1213.
13. Toledo P, McCarthy RJ, Hewlett BJ, Fitzgerald PC, Wong CA. The accuracy of blood loss estimation after simulated vaginal delivery. Anesth Analg. 2007;105:1736–1740.
14. Ambardekar S, Shochet T, Bracken H, Coyaji K, Winikoff B. Calibrated delivery drape versus indirect gravimetric technique for the measurement of blood loss after delivery: a randomized trial. BMC Pregnancy Childbirth. 2014;14:276.
15. Bernal JL, Cummins S, Gasparrini A. Interrupted time series regression for the evaluation of public health interventions: a tutorial. Int J Epidemiol. 2017;46:348–355.
16. Mascha EJ, Sessler DI. Segmented regression and difference-in-difference methods: assessing the impact of systemic changes in health care. Anesth Analg. 2019;129:618–633.
17. Huitema BE, McKean JW. Design specification issues in time-series intervention models. Educ Psychol Meas. 2000;60:38–58.
18. Shields LE, Goffman D, Caughey AB. ACOG practice Bulletin No. 183: postpartum hemorrhage. Obstet Gynecol. 2017;130:168–186.
19. Zhang WH, Deneux-Tharaux C, Brocklehurst P, Juszczak E, Joslin M, Alexander S; EUPHRATES Group. Effect of a collector bag for measurement of postpartum blood loss after vaginal delivery: cluster randomised trial in 13 European countries. BMJ. 2010;340:c293.
20. Chua S, Ho LM, Vanaja K, Nordstrom L, Roy AC, Arulkumaran S. Validation of a laboratory method of measuring postpartum blood loss. Gynecol Obstet Invest. 1998;46:31–33.
21. Konig G, Holmes AA, Garcia R, et al. In vitro evaluation of a novel system for monitoring surgical hemoglobin loss. Anesth Analg. 2014;119:595–600.
22. Walker RH. Mathematical calculations in transfusion medicine. Clin Lab Med. 1996;16:895–906.
23. Stuart EA, Huskamp HA, Duckworth K, et al. Using propensity scores in difference-in-differences models to estimate the effects of a policy change. Health Serv Outcomes Res Methodol. 2014;14:166–182.
24. Schulte PJ, Mascha EJ. Propensity score methods: theory and practice for anesthesia research. Anesth Analg. 2018;127:1074–1084.
25. Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther. 2002;27:299–309.
26. Rubenstein AF, Zamudio S, Al-Khan A, et al. Clinical experience with the implementation of accurate measurement of blood loss during cesarean delivery: influences on hemorrhage recognition and allogeneic transfusion. Am J Perinatol. 2018;35:655–659.
27. Frank SM, Savage WJ, Rothschild JA, et al. Variability in blood and blood component utilization as assessed by an anesthesia information management system. Anesthesiology. 2012;117:99–106.
28. Ho AMH, Phelan R, Mizubuti GB, et al. Bias in before–after studies: narrative overview for anesthesiologists. Anesth Analg. 2018;126:1755–1762.
29. Fedoruk K, Seligman KM, Carvalho B, Butwick AJ. Assessing the association between blood loss and postoperative hemoglobin after cesarean delivery: a prospective study of 4 blood loss measurement modalities. Anesth Analg. 2019;128:926–932.
30. Borovac-Pinheiro A, Pacagnella RC, Cecatti JG, et al. Postpartum hemorrhage: new insights for definition and diagnosis. Am J Obstet Gynecol. 2018;219:162–168.

Supplemental Digital Content

Copyright © 2019 International Anesthesia Research Society