Outcomes of Nontransfusable and Transfusable Patients After Major Noncardiac Surgery: A Retrospective Cohort Study

Abstract

BACKGROUND:

Patients admitted to HELIOS Klinikum in Gotha and Erfurt, Germany, received one of 3 models of care. Nontransfusable patients received transfusion-free blood management, whereas transfusable patients received either patient blood management (PBM) or no PBM. Few studies have compared outcomes in patients undergoing these models of care within 1 hospital network. Our primary aim was to compare adult nontransfusable patients undergoing major surgery to transfusable patients. Our secondary aim was to compare transfusable patients receiving PBM strategies to those receiving no PBM.

METHODS:

A retrospective cohort study of 25,979 major adult noncardiac surgical admissions to 2 German hospitals between 2008 and 2020. We applied propensity score methods to multivariable regression models adjusting for age, sex, admission hemoglobin, comorbidities, surgical procedure group, surgical complexity, and admission year. Outcomes included in-hospital mortality, surgical site complications, renal complications, respiratory complications, acute myocardial infarction (AMI), readmissions within 30 days, length of stay, estimated blood loss, transfusion rate, and transfusion reactions.

RESULTS:

Patients receiving transfusion-free blood management had lower mortality (odds ratio [OR] 0.33, 95% confidence interval [CI], 0.26–0.42; P < .001), renal complications (OR 0.40, 95% CI, 0.34–0.48; P < .001), respiratory complications (OR 0.43, 95% CI, 0.37–0.49; P < .001), readmissions (OR 0.54, 95% CI, 0.48–0.60; P < .001), and shorter hospital stay (risk ratio [RR] 0.91, 95% CI, 0.90–0.93; P < .001) compared to transfusable patients. There were no AMI complications in the transfusion-free group compared to 0.3% (n = 78) in the transfusable group. Surgical site complications were not significantly different between groups (OR 0.94, 95% CI, 0.86–1.02; P = .140). In our secondary analysis of transfusable patients, PBM was associated with lower mortality (OR 0.79, 95% CI, 0.66–0.95; P = .012), surgical site complications (OR 0.62, 95% CI, 0.57–0.69, P < .001), renal complications (OR 0.76, 95% CI, 0.65–0.88; P < .001), respiratory complications (OR 0.68, 95% CI, 0.60–0.78; P < .001), and shorter hospital stay when compared to no PBM (RR 0.86, 95% CI, 0.85–0.87; P < .001). Hospital readmissions were higher in the PBM group (OR 1.28, 95% CI, 1.18–1.40; P < .001). The proportion of patients receiving a red cell transfusion, units transfused per patient, and estimated blood loss were lower in the PBM group when compared to no PBM. There were no transfusion complications coded in the PBM or no PBM groups.

CONCLUSIONS:

Our primary and secondary analyses demonstrate addressing anemia and minimizing or avoiding transfusion is associated with improved outcomes. The results of the study highlight the important role transfusion-free care and PBM have in improving outcomes for patients undergoing major surgery.

See Article, page 650

KEY POINTS.

• Question: What are the outcomes associated with transfusion-free blood management, patient blood management (PBM), and no PBM?

• Findings: Patients receiving transfusion-free blood management had lower mortality, morbidity, readmissions, and shorter hospital stay compared to transfusable patients, and similar improved outcomes were observed for transfusable patients receiving PBM when compared to no PBM.

• Meaning: Addressing anemia, minimizing blood loss and optimizing/managing coagulation/coagulopathy, and avoiding transfusion are associated with improved outcomes.

Major surgery is frequently associated with anemia and bleeding, and sometimes coagulopathy, all independently associated with morbidity and mortality. Transfusions are commonly administered to remedy these. However, transfusion itself is independently associated with morbidity and mortality and as a result is usually only administered when anemia, bleeding, or coagulopathy are considered severe.^1,2

Patient blood management (PBM), the judicious management of the patient’s own blood, including management of anemia, bleeding, and coagulopathy, is a solution to these independent risk factors for poor patient outcomes.³ Studies of comprehensive PBM programs have found they are associated with significant reductions in mortality as well as hospital length of stay, complications, and transfusions.^4,5

In a different study type, research investigating nontransfusable patients can provide natural experiment-like results and give a unique window into the outcomes of transfusion avoidance with or without the meticulous management of anemia, bleeding, and coagulopathy.^6,7 The terms “transfusible” and “nontransfusible” have been described elsewhere.⁷ In short, nontransfusable patients cannot be transfused because the patient declines consent (eg, for personal or religious reasons), or is unable (eg, rare blood types, blood shortages, etc) to receive allogeneic blood transfusions of red cells, white cells, platelets and plasma. By contrast, transfusable patients are those who are willing and able to be transfused if a physician deemed a transfusion to be clinically indicated, whether or not they are transfused. There is a subtle difference between studies comparing transfused versus nontransfused patients and studies comparing transfusable versus nontransfusable. Studying the latter also reports on effects of the blood-related therapeutic bundle the patients receive and includes patients who were successfully spared reaching the individual transfusion trigger. One systematic review and meta-analysis comparing transfusable and nontransfusable patients reported no statistically significant differences between the 2 groups. However, nontransfusable patients tended to have lower rates of infection, reoperation, and acute myocardial infarction (AMI).⁷

There are no studies that compare outcomes of nontransfusable patients receiving comprehensive transfusion-free blood management (TFBM) to transfusable patients treated either with or without PBM within a hospital network. Helios Klinikum Gotha and Helios Klinikum Erfurt are neighboring hospitals in Germany and part of the same hospital network. Between these 2 hospitals 3 models of care are available: (1) TFBM for nontransfusable patients, (2) PBM for transfusable patients, (3) and care without PBM for transfusable patients (no-PBM). The hospital in Gotha provides 2 models of care (TFBM and PBM), whereas the hospital in Erfurt has no formal PBM or TFBM program (no-PBM).

The primary objective of our study was to compare mortality and morbidity in adult nontransfusable patients undergoing major surgery and receiving TFBM to transfusable patients. Our secondary objective was to compare outcomes between transfusable patients receiving 2 different management strategies, PBM and no-PBM.

METHODS

To address our primary and secondary objectives, we designed a retrospective cohort study. Approval for this study was granted by the Ethics committee of the Chamber of Physicians Thuringia (approval number 42674/2020/78). The study was registered with, controlled, and approved by the HELIOS Kliniken GmbH data protection authority, the HELIOS Center for Research and Innovation (HCRI), and the institutional data protection authorities. The need for participant consent was waived based on national legislation as only anonymized data would be used for analysis, and data handling of identifiable data would only be performed within the hospital information system. The study was preregistered on the clinicaltrials.gov database (ID NCT06553833, principal investigator: P. Seeber, date of registration: May 5, 2024). We reported our results according to the “Strengthening The Reporting of Observational studies in Epidemiology” (STROBE) statement.⁸

Setting

Patients were included in our study if they were 18 years or over and underwent major surgery at HELIOS Klinikum Gotha (HKG) or HELIOS Klinikum Erfurt (HKE) between June 2008 and December 2020. HKG and HKE are 2 neighboring hospitals with overlapping personnel, purchasing, IT departments, and standard operating procedures. Both hospitals offer basic, advanced, and specialist care to their communities, and provide among them 3 different models of care, TFBM, PBM, and no-PBM. Patients were allocated to their treatment group at first hospital contact before the major surgical procedure.

All 3 models of care provided preoperative anemia screening, tranexamic acid for polytrauma and major orthopedic patients, factor-based coagulation management (fibrinogen, 4-factor prothrombin complex concentrate, etc) rather than plasma, electrocautery (or similar surgical devices) and local hemostatic agents, immediate hemostasis for patients with critical bleeding, and, in transfusable patients, leukocyte-depleted red cell concentrates. Anemia of chronic kidney disease was treated with low-dose erythropoiesis-stimulating agents. However, the 3 models of care we compared differed substantially in the management of the patient’s own blood and the use of allogeneic transfusions.

The TFBM service provided the most comprehensive application of PBM strategies with the goal of optimizing outcomes in nontransfusable patients. This service, which started in 2008, was established to provide care to patients declining allogeneic blood transfusions for personal or religious reasons. More details are available elsewhere,⁹ but in brief the service was staffed by a senior consultant surgeon, an anesthesiolgist, and a nurse, and received referrals from across Germany, Europe, and overseas. Where the service required was available, no referred patients were declined treatment. The individualized, structured approach to the detection and management of anemia, bleeding, and coagulopathy provided by this service has produced promising results.¹⁰ Key features of this service were particularly meticulous attention to minimizing blood loss, including the provision of immediate, even after-hours hemostasis in all bleeding patients with extensive use of cell salvage where possible, including during major cancer surgery. In addition to preoperative anemia screening, patients received an interdisciplinary root-caused-based individualized management of anemia, including total-dose intravenous iron, vitamins, erythropoiesis-stimulating agents, and cause-specific antianemic drugs (such as antirheumatic medications). The service also screened patients preoperatively using a questionnaire-based coagulation screening tool and the optimization of coagulation. Tranexamic acid was administered to patients with individual risk factors for bleeding (eg, blood thinners, intraoperative oozing, prolonged surgical time). Whenever possible, major surgery was postponed until anemia and coagulopathy were managed to the maximum extent possible. Every effort was made to minimize iatrogenic blood loss. Nontransfusable patients not treated under the TFBM program were excluded from this analysis.

Patients treated under the PBM model of care received convenience measures of PBM and were transfused when a restrictive transfusion threshold was met. Convenience measures of PBM, that is, those that are convenient for the patient or health care providers as they do not interfere with the scheduled care, included cell salvage in major orthopedic and visceral surgery, anemia management without root-cause assessment based on intravenous iron, vitamins and occasional erythropoiesis-stimulating agents, tranexamic acid in major visceral surgery, and minimization of iatrogenic blood loss in monitored care settings only. Other measures of PBM that may have meant inconveniences for the patient or health care providers, such as postponing surgery in case of anemia, scheduling coagulation specialist care, or staging a planned surgery, were not routinely part of this PBM model of care.

The no-PBM group included transfusable patients at HKE where there was no PBM care model. These patients were transfused according to German transfusion guidelines,¹¹ an approach considered “usual care” in Germany. The guidelines recommend transfusions when the hemoglobin falls <6 g/dL, unless the patient has risk factors (chronic ischemic heart disease, heart failure, or stroke); then the recommended trigger is between 6 and 8 g/dL. When the patient presents with physiological transfusion triggers (such as tachycardia, hypotension, lactacidosis or ischemic signs in the ECG), transfusion is recommended when the hemoglobin falls <8 to 10 g/dL. For those patients, anemia of hematinic deficiency was managed based on oral iron, vitamins and standard transfusion. Cell salvage was not commonly used nor were efforts made to reduce diagnostic iatrogenic blood loss. Surgery was not postponed for the sole reason of anemia or coagulation management. More details on the interventions provided by the 3 services are detailed in Supplemental Digital Content, Supplemental Table 1, https://links.lww.com/AA/F435.

Participants

All consecutive adult participants hospitalized in HKG and HKE between June 2008 and December 2020 were potentially eligible to participate in the study if they underwent major surgical procedures in specialties offered by both hospitals (Supplemental Digital Content, Supplemental Table 2, https://links.lww.com/AA/F435, and Table 3, https://links.lww.com/AA/F435). Major surgery was defined as any surgical procedure from the German procedure classification system “Operationen- und Prozedurenschlüssel” (OPS) starting with the number 5, with the potential to cause anemia, substantial bleeding, or coagulopathy (Supplemental Digital Content, Supplemental Table 4, https://links.lww.com/AA/F435).

For this analysis, surgical specialties staffed by physicians who rented hospital facilities to treat their own patients were excluded as these clinicians were not bound by the hospital’s standard operating procedures. Also excluded were patients transferred to a specialized care facility within 6 hours of admission as these patients did not receive their primary care from our participating hospitals.

Initially, we planned to compare outcomes between the 3 study groups (TFBM, PBM, and no-PBM) over the entire study period (June 2008 to December 2020); however, data were not available for analysis before 2017 for one of our participating hospitals (HKE) because of a change in the hospital information system. As a result of the missing years, we updated our study design and analysis plan, with deviations from protocol available in Supplemental Digital Content, Supplemental Table 5, https://links.lww.com/AA/F435. For our primary analysis, we compared nontransfusable patients to transfusable patients for the entire study period. For our secondary analysis, we compared outcomes across 2 different transfusable treatment strategies (PBM and no-PBM) and limited our analysis to the most recent 4 years (2017–2020) as full data for both hospitals was available.

Exposures

To address our primary objective, our primary comparison (exposure) was nontransfusable patients receiving comprehensive TFBM to transfusable patients receiving PBM or usual care from HKG and HKE (PBM and no-PBM, respectively). Transfusable patients (the comparator group) were included whether they received an allogeneic blood transfusion or not. To address our secondary aim, a further comparison was made between transfusable patients undergoing major surgery at HKG (PBM) to transfusable patients at HKE, where there is no PBM care model (no-PBM).

Outcomes

For the purpose of this analysis, we did not dichotomize outcomes into primary or secondary measures. For our primary comparison the outcomes we analyzed included in-hospital mortality, surgical site complications, renal complications, respiratory complications, AMI, readmissions within 30 days, and length of stay. Surgical site complications were defined according to prespecified International Classification of Diseases (ICD) code lists (Supplemental Digital Content, Supplemental Table 6, https://links.lww.com/AA/F435) and included bleeding, hematoma, wound infection, wound necrosis, wound dehiscence, and anastomotic leak. Renal complications were defined according to RIFLE criteria.¹² Respiratory complications were defined as patients requiring ventilation support. AMI was defined as a documented increase in troponin in conjunction with an AMI diagnosis that was not identified as admission diagnosis. Readmission was defined as a readmission to the same hospital within 30 days regardless of the cause.

Although we had also considered reporting other outcome parameters such as deep vein thrombosis, pulmonary embolism, and sepsis, the available data set did not allow us to tell whether these were present at admission or developed thereafter. This is why we did not study these parameters.

In addition to in-hospital mortality, surgical site complications, renal complications, respiratory complications, AMI, readmissions within 30 days, and length of stay, our secondary comparison of transfusable patients included the following outcomes: proportion of patients receiving a red cell transfusion, mean red cell units transfused per patient, mean estimated blood loss, and transfusion reactions.

Potential Confounders

Data were collected on the following potential confounders: patient age, sex, admission hemoglobin, comorbidities, surgical procedure group, surgical complexity score, and year of admission. Patient comorbidities were defined using the Charlson comorbidity index (Supplemental Digital Content, Supplemental Table 7, https://links.lww.com/AA/F435), and Supplemental Digital Content, Supplemental Table 4, https://links.lww.com/AA/F435, describes the surgical complexity scoring and provides more details on the surgical procedure groups.

Data on patient age were provided by age decile. Given the large number of categories, we decided before analysis to group the age deciles into 4 clinically meaningful categories: 18 to 39; 40 to 59; 60 to 79; 80+. We did not group year of admission into categories, rather, we analyzed it as mean years (standard deviation [SD]) since baseline admission (January 1, 2008), consistent with a previous large study.⁶

Statistical Methods

For the primary analysis, we applied propensity score methods and used the inverse probability of treatment weighting (IPTW) to control for baseline factors that might differ between treatment groups and estimated the average treatment effect (ATE). We chose propensity score weighting over propensity score matching to avoid discarding data through the matching process. The propensity score was defined as the probability of receiving treatment based on the baseline characteristics. This was estimated using a multivariable logistic regression model with patient age, sex, admission hemoglobin, comorbidities, surgical procedure group, surgical complexity score, and years since baseline as covariables. We chose these covariates as they were baseline characteristics measured before the treatment assignment that we judged related to the probability of receiving the treatment. We assigned each patient a weight based on the IPTW and checked balance diagnostics by comparing unweighted and weighted patient characteristic differences with standardized mean differences. We truncated weights at the 1st and 99th percentiles to reduce the impact of extreme weights and improve the robustness of our results.^13,14

After this, multivariable regression models assessing study outcomes were weighted using the IPTW and additionally included covariate adjustments for patient age, sex, admission hemoglobin, comorbidities, surgical procedure group, surgical complexity score, and years since baseline. We purposely did not choose potential confounders based on their p-value. Rather we chose to include these confounders based on our review of similar studies with the aim of controlling the possible effects of any differences in patient type or complexity between groups.^15–17

We planned to apply logistic regression models to assess the difference in our binary outcomes, in-hospital mortality, surgical site complications, renal injury, respiratory complications, AMI, readmissions within 30 days, transfusion reactions, receipt of a red cell transfusion. Poisson regression models were applied to investigate the mean units of red blood cells (RBCs) transfused per admission. In Poisson regression, it is assumed that the mean and variance of the outcome are equal. Overdispersion occurs when the variance of the data is larger than the mean, which violates this assumption and means a Poisson model may not be appropriate. This was the case for our length of stay outcome and after testing for overdispersion we chose negative binomial regression models to analyze length of stay. We fitted continuous variables into our multivariable models as restricted cubic splines to account for nonlinearity.

To manage missing data, we initially set a threshold of <5% for conducting a complete case analysis. Before analysis we identified data missing for admission hemoglobin (5.96%) and surgical procedure group (0.13%). Although the proportion of missing hemoglobin data was slightly above our predefined threshold, we considered this difference minimal and unlikely to introduce significant bias. Given the small deviation and our judgment that missingness was not systematically related to the outcome, we proceeded with a complete case analysis as the most appropriate and pragmatic approach. These statistical methods were applied to our primary and secondary analyses.

All statistical analyses were performed using R version 4.2.2 (the R Foundation for Statistical Computing), with P values 2-sided and the statistical significance level set at 5%.

RESULTS

Our data extract consisted of 36,165 admissions undergoing major surgical procedures between 2008 and 2020 at the 2 study hospitals (Figure 1). After exclusions were applied our final dataset consisted of 25,979 patients for analysis. Patient characteristics are available in Table 1. Overall, 52.4% (n = 13,607) were female, and most patients were aged 60–79 years (49.6%, n = 12,879). Of the patients undergoing major surgery, 425 (1.6%) requested treatment without allogeneic blood transfusions and were managed by the TFBM care model. Of the remaining 25,554 patients making up the transfusable groups, 14,546 received care from HKG and 11,008 from HKE.

Figure 1.

Flowchart of patient identification for the study cohort.

Table 1.

Baseline Characteristics of the Study Population Stratified by Treatment Strategy

	Nontransfusable (n = 425)	Transfusable (n = 25,554)	P Value
Female sex	264 (62.1%)	13,343 (52.2%)	<.001
Admission hemoglobin	13.36 (1.74)	13.31 (2.01)	.599
Age category			<.001
18–39	22 (5.2%)	3878 (15.2%)
40–59	118 (27.8%)	5782 (22.6%)
60–79	246 (57.9%)	12,633 (49.4%)
80+	39 (9.2%)	3261 (12.8%)
Surgical complexity	1.31 (0.51)	1.18 (0.39)	<.001
Years since January 2008)	5.96 (3.31)	7.68 (3.35)	<.001
Charlson comorbidity index	2.08 (2.18)	1.82 (2.18)	.015
Surgical procedure group			<.001
Abdominal major surgery involving abdominal wall, appendix, ovary	24 (5.6%)	1147 (4.5%)
Amputation major	0 (0.0%)	322 (1.3%)
Biliodigestive anastomotic surgery	2 (0.5%)	172 (0.7%)
Bladder urinary	12 (2.8%)	1612 (6.3%)
Bone of chest and chest wall	8 (1.9%)	504 (2.0%)
Bone (major), joint, shoulder, knee	34 (8.0%)	2196 (8.6%)
Colorectal surgery	73 (17.2%)	2147 (8.4%)
Kidney, pararenal, ureter related major procedures	14 (3.3%)	1006 (3.9%)
Liver	15 (3.5%)	206 (0.8%)
Lung	2 (0.5%)	229 (0.9%)
Miscellaneous bone surgerya	3 (0.7%)	65 (0.3%)
Multivisceral surgery (involving 2 or more organs)	39 (9.2%)	757 (3.0%)
Muscle or skin related major surgery	9 (2.1%)	175 (0.7%)
Pancreas (nonmultivisceral)	9 (2.1%)	231 (0.9%)
Prostate	28 (6.6%)	1978 (7.7%)
Small bowel	11 (2.6%)	350 (1.4%)
Spine	17 (4.0%)	4382 (17.1%)
Stomach	14 (3.3%)	278 (1.1%)
Major thyroid, lymph nodes (major resection, neck dissection)	29 (6.8%)	628 (2.5%)
Total hip	63 (14.8%)	4130 (16.2%)
Uterus, C-section (major)	19 (4.5%)	3039 (11.9%)

Values are n (%) or mean (SD).

Abbreviation: SD, standard deviation.

^aCombination of 2 major bones or 1 major bone plus 1 major organ.

When compared to transfusable patients, nontransfusable patients were more likely to be female (62.1% vs 52.2%; P < .001), more likely to have a higher mean surgical complexity score (1.31 vs 1.18; P < .001), and higher mean Charlson comorbidity score (2.08 vs 1.82; P = .015). Mean hemoglobin level on admission was comparable between the 2 groups (13.4 vs 13.3 g/dL; P = .599).

These differences in baseline characteristics between groups were well balanced after propensity score weighting methods were applied. Covariate balance statistics comparing the original data and the propensity adjusted weighted observations in the nontransfusable and transfusable cohorts are available in Table 2.

Table 2.

Covariate Balance Statistics

	Unweighted			Propensity score weighted
Baseline characteristic	Nontransfusable	Transfusable	SMD	Nontransfusable	Transfusable	SMD
Female sex	62%	52%	0.201	54%	52%	0.031
Admission hemoglobin (g/dL)	13.36	13.31	0.027	13.45	13.31	0.073
Age category
40–59	28%	23%	0.118	26%	23%	0.068
60–79	58%	49%	0.170	48%	50%	0.025
80+	9%	13%	0.115	15%	13%	0.057
Surgical complexity	1.31	1.18	0.280	1.17	1.18	0.038
Years since January 2008	5.96	7.68	0.514	7.47	7.65	0.054
Charlson comorbidity index	2.08	1.82	0.119	1.62	1.82	0.096
Surgical procedure group
Amputation major	0%	1%	0.160	0%	1%	0.158
Biliodigestive anastomotic surgery	1%	1%	0.027	1%	1%	0.017
Bladder urinary	3%	6%	0.167	6%	6%	0.020
Bone of chest and chest wall	2%	2%	0.007	1%	2%	0.075
Bone, joint, shoulder, knee	8%	9%	0.022	11%	9%	0.065
Colorectal	17%	8%	0.265	9%	9%	0.030
Kidney, pararenal, ureter-related major surgery	3%	4%	0.034	4%	4%	0.006
Liver	4%	1%	0.188	1%	1%	0.014
Lung	1%	1%	0.052	1%	1%	0.048
Miscellaneous bone surgery	1%	0%	0.065	1%	0%	0.037
Multivisceral surgery (involving 2 or more organs)	9%	3%	0.262	3%	3%	0.007
Muscle or skin related major surgery	2%	1%	0.122	1%	1%	0.022
Pancreas (nonmultivisceral)	2%	1%	0.100	1%	1%	0.014
Prostate	7%	8%	0.045	7%	8%	0.017
Small bowel	3%	1%	0.088	1%	1%	0.003
Spine	4%	17%	0.438	19%	17%	0.063
Stomach	3%	1%	0.151	1%	1%	0.012
Major thyroid, lymph nodes (major resection, neck dissection)	7%	3%	0.208	2%	3%	0.005
Total hip	15%	16%	0.037	19%	16%	0.064
Uterus, C-section (major)	5%	12%	0.273	7%	12%	0.170

Abbreviation: SMD, standardized mean difference.

Primary Comparison (Nontransfusable vs Transfusable)

Unadjusted and propensity weight adjusted patient outcomes are presented in Figure 2. Comparing nontransfusable to transfusable patients, the unadjusted odds of in-hospital death, surgical site complications, renal complications, respiratory complications, and readmissions were 0.54 (95% CI, 0.22–1.30; P = .166), 1.19 (95% CI, 0.87–1.63; P = .277), 0.62 (95% CI, 0.31–1.25; P = .179), 1.12 (95% CI, 0.73–1.73; P = .609), and 0.69 (95% CI, 0.46–1.03; P = .069), respectively. Unadjusted mean hospital length of stay was 11.28 days vs 11.39 days (rate ratio [RR] 0.99, 95% CI, 0.93–1.05; P = .752), between the nontransfusable and transfusable groups, respectively. There were no AMI complications in the nontransfusable group compared to 0.3% (n = 78) in the transfusable group.

Figure 2.

Adjusted mean estimated blood loss (mL) comparing nontransfusable patients to transfusable patients in major noncardiac surgery.

In the propensity score weighted analysis, nontransfusable patients treated by the TFBM care model had significantly fewer in-hospital deaths (OR 0.33, 95% CI, 0.26–0.42; P < .001), renal complications (OR 0.40, 95% CI, 0.34–0.48; P < .001), respiratory complications (OR 0.43, 95% CI, 0.37–0.49; P < .001), and hospital readmissions (OR 0.54, 95% CI, 0.48–0.60; P < .001) when compared to transfusable patients. Hospital length of stay was also significantly lower (RR 0.91, 95% CI, 0.90–0.93; P < .001) in the nontransfusable group when compared to transfusable patients. Surgical site complications were not statistically different between the 2 groups (OR 0.94, 95% CI, 0.86–1.02; P = .140).

Complete data on transfusions and estimated blood loss were not available before 2017. Between 2017 and 2020 there were 0 patients transfused RBCs in the TFBM group compared with 1227 (9.57%) patients in the transfusable group. Transfused patients received a median (Q1, Q3) of 2 (1, 4) RBC units per patient. The adjusted mean estimated blood loss was 305 mL (95% CI, 142–468 mL; P < .001) lower in the TFBM patients when compared to transfusable patients (Figure 3). Cell saver usage was several times higher in the TFBM cohort when compared to transfusable patients (28.6% vs 7.3%; P < .001).

Figure 3.

Adjusted relative risk of the different outcomes comparing nontransfusable patients to transfusable patients in major noncardiac surgery. Unadjusted numbers are n (%) unless otherwise stated. Relative risks presented are odds ratios for all outcomes except length of stay, for which rate ratios are presented. Propensity score adjustment involved multivariable regression models weighted using the inverse propensity for treatment and additionally included covariate adjustments for patient age, sex, admission hemoglobin, comorbidities, surgical procedure group, surgical complexity score, and years since baseline. There were no AMI complications in the nontransfusable group compared to 0.3% (n = 95) in the transfusable group. AMI indicates acute myocardial infarction; SD, standard deviation.

Secondary Comparison (Transfusable With and Without PBM)

Our secondary analysis compared 2 models of care: transfusable patients treated under a PBM care model and transfusable patients receiving standard care (no-PBM). Between 2017 and 2020 this subset of data contained 13,767 patients undergoing major surgery, made up of 5385 treated under a PBM care model and 8382 under the no-PBM care model. Patient baseline characteristics are available in Supplemental Digital Content, Supplemental Table 8, https://links.lww.com/AA/F435.

These differences in baseline characteristics between groups were well balanced after propensity score weighting methods were applied. Covariate balance statistics comparing the original data and the propensity adjusted weighted observations in the nontransfusable and transfusable cohorts are available in Supplemental Digital Content, Supplemental Table 9, https://links.lww.com/AA/F435.

The PBM cohort had fewer in-hospital deaths (OR 0.79, 95% CI, 0.66–0.95; P = .012), surgical site complications (OR 0.62, 95% CI, 0.57–0.69, P < .001), renal complications (OR 0.76, 95% CI, 0.65–0.88; P < .001), respiratory complications (OR 0.68, 95% CI, 0.60–0.78; P < .001), and shorter length of stay (RR 0.86, 95% CI, 0.85–0.87; P < .001) in the adjusted analysis. Hospital readmissions were higher in the PBM group compared to no-PBM (OR 1.28, 95% CI, 1.18–1.40; P < .001) (Figure 4). As there were small rates of AMI, we did not conduct an adjusted analysis. In the unadjusted analysis, the rate was 0.2% (n = 13) in the PBM group and 0.4% (n = 30) in the no-PBM group (OR 0.67, 95% CI, 0.35–1.29; P = .298).

Figure 4.

Adjusted relative risk of the different outcomes comparing transfusable patients treated in a patient blood management care model to transfusable patients with no patient blood management care model. Unadjusted numbers are n (%) unless otherwise stated. Relative risks presented are odds ratios for all outcomes except length of stay, for which rate ratios are presented. Propensity score adjustment involved multivariable regression models weighted using the inverse propensity for treatment and additionally included covariate adjustments for patient age, sex, admission hemoglobin, comorbidities, surgical procedure group, surgical complexity score, and years since baseline. CI indicates confidence interval; PBM, patient blood management; SD, standard deviation.

There were 460 (8.54%) patients transfused RBCs in the PBM group compared with 870 (10.38%) patients in the no-PBM group. Overall, transfused patients in the PBM group and no-PBM groups received a median (Q1, Q3) of 2 (1, 4) RBC units per patient.

After adjusting for confounders, the odds of receiving a red cell transfusion were 0.84 times lower (95% CI, 0.76–0.93; P < .001) for patients treated under the PBM care model when compared to the no-PBM care model, and the mean units of red cells transfused was 0.83 times lower (95% CI, 0.79–0.87; P < .001). The adjusted mean estimated blood loss was 41 mL (95% CI, 9–74 mL; P = .013) lower in the PBM group when compared to no-PBM. Cell saver usage was several times higher in the PBM group when compared to no-PBM group (19.9% vs 0.3%; P < .001). There were no transfusion complications coded in the PBM or no-PBM groups.

DISCUSSION

In our propensity weighted analysis, nontransfusable patients receiving comprehensive TFBM had significantly lower adjusted in-hospital mortality, renal complications, respiratory complications, hospital readmissions, and hospital length of stay when compared to transfusable patients. Surgical site complications were not statistically different between the 2 groups.

Few studies compare nontransfusible and transfusible patients undergoing major noncardiac surgery, most of which include very few nontransfusible participants. In fact, only Harwin et al¹⁸ published a study with a considerable number of nontransfusible participants, namely 164, who underwent total knee replacement and reported acceptable orthopedic results without any mortality.

However, the results of our primary analysis are consistent with a large “natural experiment” propensity matched study from the Cleveland Clinic comparing the outcomes of patients undergoing cardiac surgery who were declining transfusions and received “severe blood conservation.”⁶ These transfusion-free patients had better 1-year survival, fewer acute complications, and shorter length of stay than matched patients receiving transfusions. A more recent review of cardiac surgery in patients declining transfusions compared to those not declining transfusions showed comparable mortality with a trend to favor patients declining transfusions.¹⁹ Our research adds another important study to the literature as it compares outcomes in major noncardiac surgery patients.

A recent systematic review and meta-analysis reported no statistically significant differences in mortality or morbidity in nontransfusable patients when compared to transfusable patients.⁷ However, those considered nontransfusable were more likely to have a shorter ICU stay, and although there were no statistically significant differences in morbidity outcomes, the nontransfusable patients tended to have fewer infections, AMI, and reoperations.

This potential difference in results may be due to the variability in the care provided. Not all nontransfusable patients in that meta-analysis were treated in a comprehensive TFBM model of care. The comprehensive TFBM service described in this article made full use of all available measures to minimize blood loss, optimize the patient’s own red cell mass, and manage coagulation, without resorting to allogeneic transfusions.

Incorporating PBM strategies into the care of patients undergoing major surgery was associated with significant reductions in blood loss. This was especially pronounced in patients receiving TFBM where intense efforts were made to avoid blood loss. In addition to anemia management and transfusion avoidance, significant reductions in blood loss are likely to be one of the main contributing factors for reduced mortality. Other studies have confirmed an association between blood loss and mortality.^20–22

Blood loss is common after noncardiac surgery. In their study of over 16,000 patients, Roshanov et al²³ reported that the postoperative bleeding incidence was 36%, with most bleeding events occurring within 3 days of surgery. These results strongly suggest there is substantial room for improvement for any surgeon performing major surgery. The study further reported that a significant proportion of deaths could potentially be avoided by addressing bleeding, demonstrating that bleeding is no innocent or unchangeable bystander of major surgery.²³

Another difference between the patients in our study and those in other studies is our use of a factor-based coagulation management, where fibrinogen concentrates and 4-factor PCCs were used. Cryoprecipitate was not used, and even in transfusable patients, platelet concentrates and plasma were used only very rarely. Such a strategy ensures coagulation factors are available quicker, are given blood—group independent and do not carry the same risks as plasma or platelets (transfusion-transmittable infections, volume overload, storage lesions, alloimmunization, etc). This is not standard practice and needs to be kept in mind when comparing this study with other published studies reporting on outcomes of PBM.

In our secondary comparison, transfusable patients treated under a PBM model of care were less likely to have in-hospital deaths, surgical site complications, renal complications, respiratory complications, and more likely to have shorter length of stay, when compared to patients receiving no-PBM. By contrast, hospital readmissions were more likely to be higher in the PBM group compared to no-PBM. We are unsure why readmissions were higher in the PBM group; however, this was likely not due to lower discharge hemoglobin levels as these were similar between groups. Because we were not able to determine the cause for readmissions, we were unable to state whether the readmissions were related or unrelated to the patient’s previous admission.

Overall, these results are consistent with a systematic review and meta-analysis by Althoff et al⁴ that investigated the impact of comprehensive PBM on patient outcomes, pooling 17 studies and over 200,000 surgical patients. PBM was considered comprehensive if it included at least 1 strategy from each of the 3 pillars or PBM (anemia management, minimize blood loss and bleeding, optimize anemia tolerance). Their review concluded comprehensive PBM was associated with lower mortality, hospital complications, and length of stay.

When combined, the collective results of our primary and secondary analyses suggest addressing anemia, bleeding and coagulopathy, and minimizing or avoiding transfusions is associated with improved patient outcomes. Several systematic reviews and meta-analyses have reported increased mortality and morbidity associated with preoperative anemia^24–26 and red cell transfusion.^27–30 Our results, when interpreted with the results of systematic reviews and meta-analyses, collectively highlight the important role TFBM and PBM care models have for nontransfusable and transfusable patients in improving the outcomes of patients undergoing major surgery.

Strengths and Limitations

Our primary comparison was similar to the studies by Pattakos et al⁶ and Shander et al.³¹ However, 1 key difference was that our comparator group consisted of patients able to receive transfusions regardless of whether they were or were not administered a transfusion. This means our primary analysis compared different bundles of care, namely the treatment nontransfusable and transfusable patients received.

One of the strengths of our study was its multi-center design and large sample size as well as the use of well-defined criteria for outcome parameters. We also collected data on important blood-related management strategies (cell salvage, estimated blood loss, and transfusion). The majority of studies comparing nontransfusable to transfusable patients have used propensity score matching,⁷ however one of the key strengths of the propensity score weighting methods we applied is that, unlike matching, it makes use of all available data. This may result in more efficient estimates of treatment effects.^32,33

Like all observational studies our analysis has limitations. We applied propensity score methods to reduce bias. However, the propensity methods we applied can account only for those variables that were available in our dataset, and it is possible there is residual confounding we were unable to adjust for. Furthermore, data were not available on units of blood transfused before 2017 because before this year, the specific number of units of blood transfused were not available from the coded data.

Another limitation of this study is that, although the 2 hospitals included were similar in key characteristics (same region and hospital network, similar patient characteristics, overlapping personnel), it is possible that some of the observed differences in outcomes may be influenced by unmeasured factors or subtle differences between the hospitals, unrelated to the presence of a TFBM or PBM program.

CONCLUSIONS

For patients undergoing major surgery within the same hospital network, our study compared outcomes associated with 3 different strategies: (i) comprehensive TFBM for nontransfusable patients (ii) PBM for transfusable patients, and (iii) no-PBM for transfusable patients. Our collective results are consistent with what has already been published in the literature and suggest that each level of care (PBM and TFBM) is associated with improved patient outcomes. This may be a result of each level of care representing a more rigorous approach to addressing anemia, minimizing bleeding and blood loss, and avoiding transfusion. To maximize patient outcomes from PBM programs, our findings suggest that health care organizations should implement a broader and more comprehensive application of PBM strategies. These include early preoperative screening and treatment of anemia and iron deficiency, for example, with intravenous iron and erythropoiesis-stimulating agents, intraoperative factor-based hemostatic therapy, use of cell salvage and tranexamic acid, and strategies to reduce diagnostic blood loss perioperatively in high-risk surgical procedures. Embedding these practices within perioperative pathways offers tangible opportunities to improve surgical outcomes across a range of procedures.

ACKNOWLEDGMENTS

We acknowledge Mr Jakuschinski and Mrs Oehmer for their support with data extraction and data protection, and Mr Lendholt and Mrs Müller for their review of our article.

DISCLOSURES

Conflicts of Interest: K. M. Trentino reports honoraria from Event Troop, support for attending meetings and consulting from International Foundation for Patient Blood Management (IFPBM), outside the submitted work. M. Lucas and P. Seeber receive royalties for patient blood management (PBM)-related books. No other authors declared Conflicts of Interest. Funding: None. This manuscript was handled by: Shannon L. Farmer, DHSc.