Importance of Estimated Blood Loss in Resource Utilization and Complications of Benign Hysterectomy

Abstract

OBJECTIVE:

To identify the variation in estimated blood loss (EBL) at the time of benign hysterectomy and to analyze how blood loss is associated with measures of resource utilization and complications.

METHODS:

We conducted a retrospective cohort study and analyzed hysterectomies for benign indications at hospitals in the Michigan Surgical Quality Collaborative between January 1, 2013, and May 30, 2015. A sensitivity analysis was performed to identify how EBL was associated with measures of utilization (transfusion, readmission, reoperation, and length of stay) and postoperative complications. A hierarchical logistic regression model was used to identify patient level factors independently associated with EBL >400 mL and to calculate a risk and reliability adjusted rate for each hospital.

RESULTS:

There were 18,033 benign hysterectomies from 61 hospitals included for analysis. The median EBL was 100 mL, and the 90^th percentile EBL was 400 mL. A sensitivity analysis demonstrated increased risks of transfusion, readmission, reoperation, length of stay, and major postoperative complications with EBL >400 mL. The proportion of hysterectomies at hospitals in the collaborative with an EBL >400 mL ranged from 3.5% to 16.9% after risk and reliability adjustment. The risk factors with the highest adjusted odds for an EBL >400 mL included abdominal surgery compared to laparoscopic hysterectomy (aOR 2.8, CI 2.3–3.5), surgical time >3 hours (aOR 3.9, CI 3.3–4.5), and specimen weight >250 grams compared to <100 grams (aOR 4.8, CI 3.9–5.8). Adhesive disease, low surgeon volume, age <40 years old, body mass index (BMI) >35, and need for a preoperative transfusion were also statistically significantly associated with EBL >400 mL.

CONCLUSION:

There is fivefold variation in the hospital rate of hysterectomies with an EBL >400 mL (90^th percentile)—a threshold associated with significantly higher rates of healthcare utilization and complications. Avoidance of abdominal hysterectomy when possible may reduce intraoperative blood loss and associated sequelae.

Précis

There is significant hospital variation in the rate of hysterectomies with estimated blood loss >400 mL, above which there is increased healthcare utilization and complications.

Introduction

Surgery complicated by high estimated blood loss (EBL) is associated with adverse outcomes, including overall postoperative morbidity, readmission, and reoperation.^1–4 We became interested in examining whether higher rates of EBL at the time of hysterectomy might be a marker for morbidity and high rates of healthcare utilization such as length of stay, readmission, and reoperation.⁵

Maintaining adequate hemostasis is a basic surgical principle. Though EBL is subjective, it is the most commonly used method of communicating surgical blood loss with other members of the care team.^6–8 For hysterectomy, increasing uterine weight, surgical route, surgeon expertise, and operative time^9–12 can all affect EBL. However, the threshold value of EBL above which complications significantly increase is unknown. It is also unknown how estimates of blood loss during hysterectomy vary across hospitals and to what degree EBL is associated with various perioperative factors, postoperative complications, and measures of healthcare utilization. It would be unwise to make system-wide changes when a single case is complicated by high EBL (due to numerous clinical factors), but there could be value in identifying a threshold value when comparing hospital systems or larger groups of surgeons. For instance, when there is a higher than expected rate of cases complicated by high EBL, this could identify opportunities for quality improvement.

Using data regarding benign hysterectomies performed in a statewide hospital collaborative, we sought to: 1) determine at what threshold value EBL is associated with increased risks of adverse outcomes, 2) analyze the variation of reported EBL across hospitals, and 3) examine risk factors associated with increased EBL.

Methods

We performed a retrospective cohort study including patients undergoing hysterectomy for benign disease from January 1, 2013 to May 30, 2015 within the Michigan Surgical Quality Collaborative (MSQC) database. Seventy-three hospitals contribute to the database, which is funded by Blue Cross and Blue Shield of Michigan Blue Care Network and includes patients from all insurance payers. Each contributing hospital receives support for a trained, dedicated nurse data abstractor who collects patient characteristics, intraoperative information, and 30-day postoperative outcomes for hysterectomy cases. The nurse abstractors also contact patients by phone to verify whether they were admitted postoperatively to a hospital other than where the hysterectomy was performed. This data collection method is standardized and validated through routine site visits, conference calls, and internal audits.^13,14

All studies using the Michigan Surgery Quality Collaborative dataset meet criteria for “exempt” status according to the University of Michigan Institutional Review Board-Medical (HUM00073978).

All hysterectomies performed for benign indications at participating collaborative hospitals during the study period were eligible for inclusion. Hysterectomies were excluded if cancer was included as a preoperative diagnosis or noted on final pathology, if the hysterectomy was performed in the immediate postpartum setting, or if important perioperative or demographic data were missing, including surgical indication or approach, EBL, or documentation regarding readmission or reoperation. We eliminated hospitals with fewer than 20 hysterectomies over the 29-month study period as the low case volume would result in wide confidence intervals and fails to provide an accurate estimate of performance. Data available included patient age, race, parity, body mass index (BMI), American Society of Anesthesiologists (ASA) Class, smoking status, insurance status, need for preoperative blood transfusion, medical comorbidities, surgical indications, surgical approach (abdominal, laparoscopic with or without robotic assistance, laparoscopic-assisted vaginal, and vaginal hysterectomy), surgical time, EBL, concomitant procedures, presence of moderate or severe adhesions, final pathology, and specimen weight. Surgeon volume was determined by annual hysterectomy volume divided by the number of years a surgeon appeared in the database. Surgeons who performed more than 11 hysterectomies per year (above the 75^th percentile in the collaborative) were considered high volume, as this was similar to definitions used in the literature.^15,16 Those performing 11 or fewer hysterectomies per year were defined as low volume surgeons. Hospital characteristics including number of beds and teaching status were analyzed.

Utilization measures including transfusion, length of stay, reoperation, readmission, and complications were obtained. Major postoperative complications were events occurring within 30 days and analyzed as a composite outcome, including deep or organ space surgical site infection, venous thromboembolism, unplanned postoperative intubation, cerebral vascular accident, pneumonia, myocardial infarction, cardiac arrest, Clostridium difficile infection, acute kidney injury, and sepsis. Deep surgical site infection was defined using the Centers for Disease Control and Prevention criteria.¹⁷ Superficial surgical site infections and urinary tract infections (catheter-associated or spontaneous) were categorized as minor postoperative complications and excluded from the composite major complication outcome. Postoperative blood transfusion was analyzed separately and deliberately excluded from the composite outcome, as it was expected to correlate strongly with EBL.

Patients included in the cohort were compared to those excluded for missing data with bivariate analysis. A sensitivity analysis was performed to identify how EBL was associated with measures of utilization (postoperative transfusion, readmission, reoperation, and length of stay) and postoperative complications. The association between EBL at various cut-offs (EBL 0–100 mL, >100–200 mL, >200–400 mL, and >400 mL) and patient and hospital characteristics was analyzed. An upper limit of 400 mL was chosen as it represented the 90^th percentile EBL among our cohort. Chi-square tests were used for categorical variables, Wilcoxon-Mann-Whitney tests for continuous variables with a median reported, and t-tests for continuous variables where the mean is provided. Bivariate analysis was used to identify candidate variables for a hierarchical logistic regression model. In hierarchical modeling, clinically relevant patient-level factors (fixed effects) and clustering at the hospital level (random effects) are considered. Using this model, risk and reliability adjusted estimates of the proportion of cases with EBL >400 mL were calculated for each hospital with 95% confidence intervals. Multicollinearity was assessed using Spearman Rank correlation coefficients and examined in the model by including and excluding those covariates.

Results

There were 22,093 hysterectomies eligible for inclusion. After exclusion of 4,060 hysterectomies as described in Figure 1, there were 18,033 benign hysterectomies from 61 collaborative hospitals available for analysis. During the 29-month study period, the number of hysterectomies performed at each hospital ranged from 27 to 752, the annual volume of hysterectomies performed by surgeons ranged from 1 to 123, with a median of 4.6 (interquartile range 1.3, 10.9), and the percentage of low volume surgeons at each hospital ranged from 0–87%. Robotic-assisted laparoscopic hysterectomy was the most common surgical approach (41.2%), followed by abdominal (22.8%), vaginal (13.4%), laparoscopic (11.9%), and laparoscopic-assisted vaginal hysterectomy (10.7%). Overall, 3.6% of patients required readmission within 30 days of surgery, 2.0% required reoperation, 2.1% had a postoperative blood transfusion, and 4.9% had a hospital stay of four days or longer. Most patients (93.2%) who were readmitted did not have a postoperative blood transfusion. In total, there were five deaths, for a mortality rate of 0.03%. Median EBL for the cohort was 100 mL (interquartile range 50, 200 mL) and an EBL of 400 mL defined the 90^th percentile.

Figure 1.

Sample selection flow diagram. *Items not mutually exclusive, patients could be missing data in more than one category.

A sensitivity analysis was conducted to identify an EBL cut-off associated with increased resource utilization and perioperative complications. As shown in Figure 2, there was a consistent relationship between increasing blood loss and increasing rates of complications and utilization. Compared to patients whose hysterectomies had an EBL of ≤400 mL, those with EBL >400 mL had a 2.5-fold increase in prevalence of major postoperative complications (4.2 vs 1.7%, p<.001) and an 11-fold increase in prevalence of blood transfusion (12.5 vs 1.1%, p<.001). Reoperation rates were also twice as high (3.8 vs 1.9%, p<.001), with a 5.7-fold increased risk of length of stay of four days or more (20.0 vs 3.5%, p<.001). These patients were also more likely to be readmitted to the hospital (6.2 vs 3.4%, p<.001). Based on this analysis, we decided to use EBL >400 mL as a clinically relevant threshold.

Figure 2.

Percentage and number of benign hysterectomies by outcome and estimated blood loss.

Overall, 8.5% of hysterectomies in the cohort had an EBL >400 mL; the percentage varied significantly among hospitals, ranging from 3.5% to 16.9% after risk and reliability adjustment (Figure 3). Four hospitals had a rate that was significantly lower than the overall collaborative rate, while six hospitals had a rate that was significantly higher.

Figure 3.

Risk and reliability adjusted hospital rates of estimate blood loss >400 mL and 95% CIs for benign hysterectomies. Red line indicates Michigan Surgical Quality Collaborative rate.

We analyzed the bivariate associations between an EBL >400 mL and demographic, clinical, and surgical factors (Table 1). Women with an EBL >400 mL were younger, more frequently nulliparous, and non-white, with body mass index >35 kg per m² and American Society of Anesthesiologists class ≥3. These patients also had a higher prevalence of Medicaid with or without Medicare insurance or no insurance at all. The surgical indication for women with an EBL >400 mL was more frequently abnormal uterine bleeding, fibroids, or presence of a pelvic mass. The proportion of hysterectomies performed abdominally was three times higher among those with EBL >400 mL compared to those with EBL ≤400 mL. In total, an EBL >400 mL was reported for 22.7% of abdominal compared to 8.1% of laparoscopic, 3.9% of vaginal, and 2.0% of robotic hysterectomies (p<.001). EBL >400 mL was also more common with moderate or severe adhesions, longer operative time, specimen weight >250 grams, and lower surgeon volume. High volume hospitals (≥500 beds) had a higher proportion of hysterectomies with EBL >400 mL, but there was no significant effect of teaching hospital status on EBL.

Table 1.

Demographics and perioperative variables analyzed by estimated blood loss

Variable	EBL≤400 (n=16,508)	EBL>400 (n=1,525)	p-value
Age, years*	47.4 (10.8)	46.4 (8.5)	<.001
Race
White	12,724 (77.1)	801 (52.5)	<.001
Non-white	2,934 (17.8)	644 (42.2)
Unknown	850 (5.1)	80 (5.2)
Parity
Nulliparous	2,290 (13.9)	275 (18.0)	<.001
Parous	12,638 (76.6)	1,133 (74.3)
Unknown	1,580 (9.6)	117 (7.7)
Body Mass Index, kg/m2
<25	4,087 (24.8)	187 (12.3)	<.001
25–35	8,480 (51.4)	717 (47.0)
>35	3,876 (23.5)	614 (40.3)
Unknown	65 (0.4)	7 (0.5)
American Society of Anesthesiologists Class
<3	13,399 (81.2)	1,111 (72.9)	<.001
≥3	3,105 (18.8)	414 (27.1)
Unknown	4 (0.02)	0 (0.0)
Smoker	4,297 (26.0)	389 (25.5)	0.657
Surgical indication
Abnormal uterine bleeding	10,675 (64.7)	1,060 (69.5)	<.001
Fibroids	6,344 (38.4)	926 (60.7)	<.001
Pain (Endometriosis/chronic pain)	7,023 (42.5)	583 (38.2)	0.001
Pelvic organ prolapse	3,110 (18.8)	110 (7.2)	<.001
Cervical dysplasia	395 (2.4)	15 (1.0)	<.001
Endometrial hyperplasia	725 (4.4)	40 (2.6)	0.001
Pelvic mass	1,179 (7.1)	186 (12.2)	<.001
Pelvic inflammatory disease	57 (0.3)	12 (0.8)	0.014
Preoperative transfusion	83 (0.5)	55 (3.6)	<.001
Hysterectomy route
Abdominal	3,175 (19.2)	930 (61.0)	<.001
Laparoscopic	1,977 (12.0)	175 (11.5)
Robotic	7,285 (44.1)	151 (9.9)
LAVH	1,750 (10.6)	176 (11.5)
Vaginal	2,321 (14.1)	93 (6.1)
Concomitant procedures
Oophorectomy	124 (0.8)	44 (2.9)	<.001
Lysis of adhesions	2,222 (13.5)	380 (24.9)	<.001
Excision of endometriosis	426 (2.6)	9 (0.6)	<.001
Moderate/severe adhesions	3,719 (22.5)	574 (37.6)	<.001
Surgical time, hours*	2.0 (0.9)	2.8 (1.2)	<.001
Specimen weight, grams
<100	5,846 (35.4)	177 (11.6)	<.001
100–250	7,331 (44.4)	496 (32.5)
>250	2,993 (18.1)	820 (53.8)
Unknown	338 (2.0)	32 (2.1)
Insurance provider
Medicaid and/or Medicare	4,207 (25.5)	411 (27.0)	<.001
Private insurance†	11,372 (68.9)	971 (63.7)
All other/uninsured/unknown	929 (5.6)	143 (9.4)
Hospital volume, number of beds
<500	12,543 (76.0)	971 (63.7)	<.001
≥500	3,965 (24.0)	554 (36.3)
Teaching hospital	10,537 (63.8)	968 (63.5)	0.783
Surgeon volume, hysterectomies per year
Low (≤11)	5,175 (31.3)	781 (51.2)	<.001
High (>11)	11,333 (68.7)	744 (48.8)

Data presented as n (%) except where noted.

^*Mean (SD)

^†Includes Health Maintenance Organization plans

EBL: Estimated blood loss; LAVH: laparoscopically assisted vaginal hysterectomy

A hierarchical logistic regression model was developed to adjust for risk factors independently associated with an EBL >400 mL (see Table 2); those with the highest adjusted odds included an abdominal surgical route, surgical time >3 hours, and specimen weight >250 grams. The presence of adhesive disease, age <40 years, BMI >35 kg per m², lower surgeon volume, insurance status of uninsured, other, or unknown, and need for a preoperative transfusion were also significantly associated with EBL above this threshold. After controlling for these factors, we found that hospital volume was not significantly associated with an elevated EBL. Using these risk factors, the model C-statistic for predicting EBL >400 mL was 0.86.

Table 2.

Crude and adjusted odds ratios associated with estimated blood loss >400 mL

Variable	Crude OR (95% CI)	P-value	Adjusted OR (95% CI)	P-value
Age, years
<40	1.59 (1.26–2.01)	<.001	1.33 (1.01–1.74)	0.040
40–49	2.39 (1.94–2.96)	<.001	1.24 (0.96–1.60)	0.101
50–59	2.05 (1.62–2.58)	<.001	1.13 (0.86–1.48)	0.383
>60	Reference		Reference
Body Mass Index, kg/m2
<25	Reference		Reference
25–35	1.85 (1.57–2.18)	<.001	1.50 (1.26–1.79)	<.001
>35	3.46 (2.92–4.10)	<.001	2.42 (2.0–2.91)	<.001
Unknown	2.35 (1.06–5.20)	0.034	1.41 (0.58–3.39)	0.447
Specimen weight, grams
<100	Reference		Reference
100–250	2.23 (1.88–2.66)	<.001	1.98 (1.63–2.40)	<.001
>250	9.05 (7.65–10.71)	<.001	4.76 (3.89–5.83)	<.001
Unknown	3.13 (2.11–4.63)	<.001	2.53 (1.65–3.88)	<.001
Preoperative transfusion	7.41 (5.25–10.46)	<.001	2.37 (1.58–3.55)	<.001
Hysterectomy route
Robotic	0.23 (0.19–0.29)	<.001	0.21 (0.17–0.27)	<.001
Abdominal	3.31 (2.79–3.93)	<.001	2.83 (2.31–3.46)	<.001
Laparoscopic	Reference		Reference
LAVH	1.14 (0.91–1.41)	0.253	1.42 (1.12–1.82)	0.004
Vaginal	0.45 (0.35–0.59)	<.001	1.07 (0.80–1.42)	0.666
Moderate/severe adhesions	2.08 (1.86–2.32)	<.001	1.72 (1.51–1.96)	<.001
Surgical time >3 hours	3.25 (2.90–3.65)	<.001	3.88 (3.34–4.50)	<.001
Insurance
Medicaid and/or Medicare	1.14 (1.01–1.29)	0.028	1.08 (0.94–1.25)	0.272
Private insurance	Reference		Reference
All other/uninsured/unknown	1.80 (1.49–2.18)	<.001	1.40 (1.10–1.78)	0.005
Hospital volume, number of beds
<500 beds	Reference		Reference
≥500 beds	1.80 (1.62–2.02)	<.001	1.09 (0.76–1.56)	0.653
Surgeon volume
Low (≤10.9)	2.30 (2.07–2.56)	<.001	1.41 (1.24–1.61)	<.001
High (>10.9)	Reference		Reference

C-statistic=0.86

^*Includes Health Maintenance Organization plans

LAVH: laparoscopically assisted vaginal hysterectomy; OR: odds ratio; CI: confidence interval

The 1,356 patients excluded for missing data or facility type were compared to those who were included; they were significantly more likely to be younger, white, and have had a robotically assisted laparoscopic approach at lower-volume hospitals with shorter surgical time and smaller specimen weights. There was no significant difference, however, in BMI, need for preoperative transfusion, or presence of significant adhesive disease.

Discussion

In this retrospective analysis of benign hysterectomy in a statewide database, EBL above the 90^th percentile (400 mL) was associated with postoperative complications and healthcare utilization measures including prolonged hospital stay, reoperation, and readmission. We observed five-fold variation among hospitals in the proportion of benign hysterectomies with EBL >400 mL, with rates ranging from 3.5% to 16.9%. Specimen weight >250 grams, surgical time >3 hours and abdominal route of surgery were factors most strongly associated with EBL >400 mL.

In the past 15 years, the Centers for Medicare and Medicaid Services has developed hospital value-based purchasing programs to penalize hospitals with high rates of costly surgical complications. The variation in EBL among hospitals within the Michigan Surgical Quality Collaborative is important for hospitals benchmarking their performance and considering quality improvement strategies. As expected, surgical blood loss is associated with patient-level risk factors such as obesity, a larger uterus, or more significant adhesive disease. These factors may also make it more likely to plan for or convert to an open abdominal approach. However, when controlling for these and other factors including the number of hospital beds and teaching status (which were not significant), there are also site-specific factors associated with higher surgical blood loss. We found that hysterectomies performed by surgeons who performed fewer than 11 hysterectomies per year had 40% increased odds of EBL >400 mL. This, along with evidence regarding increased morbidity with an abdominal approach, reinforces the notion that surgical experience and technique play a role in the observed variation. These findings are supported by a growing body of literature about the role of surgical experience and improved outcomes.¹⁵

There are several limitations to consider with this analysis. Our study focused on estimated blood loss—a subjective and imperfect measure.¹⁸ We chose this measure because less ambiguous events such as transfusion have a low enough prevalence that they may miss the opportunity for quality improvement. We were therefore interested in identifying how increasing EBL, not just transfusion, is clinically significant. Another limitation is that we are able to report outcome data only up to 30 days following surgery. Though we expect the most significant effect of surgical blood loss to be within the month following surgery, some studies have suggested longer-term sequelae^2,3 that would be interesting to investigate in future studies. Finally, while we were able to report variation in hospital rates of EBL >400 mL, the small number of hospitals with rates significantly above and below the collaborative rate precluded any statistical analysis to identify trends to explain this variation.

The strengths of this study include use of a large, multi-site statewide database with validated data abstraction. We were able to collect and report clinically relevant major postoperative complications up to 30 days after surgery, along with valuable data on measures of healthcare utilization. The sensitivity analysis in this paper adds to the literature by providing an understanding of how estimated blood loss is associated with clinically relevant outcomes. In our analysis, we were able to identify associations between an EBL >400 mL and postoperative complications and important measures of healthcare utilization using a large, multi-site database. This differs from previous single institution studies. With sample sizes of 200 to 3,100 patients, these studies found an increased risk of hospital readmission and a trend toward increased risk of febrile morbidity with increasing EBL, but data on the association of EBL with other complications and utilization measures has been lacking.^5,19 Furthermore, there has been no published report of an EBL threshold above which these risks increase for hysterectomy. With 18,033 hysterectomies collected across 61 sites, our analysis had the power to identify this pattern of associations between estimated blood loss and clinically relevant outcomes.

In conclusion, EBL above the 90^th percentile (400 mL) at the time of benign hysterectomy is associated with increased risks of major postoperative complications, blood transfusion, hospital stay >4 days, reoperation, and readmission within 30 days of surgery. The proportion of hysterectomies with EBL >400 mL varies significantly across hospitals and may be a potential marker for quality improvement initiatives. Abdominal route of hysterectomy and surgeon volume are potentially modifiable factors that we found to be independently associated with increased odds of EBL >400 mL. Continued efforts to increase utilization of minimally invasive routes of hysterectomy could reduce EBL as well as the associated postoperative complications and measures of healthcare utilization.