Summary
Background
Delirium is a frequent and often detrimental postoperative complication that may be precipitated by perioperative inflammation. Packed red blood cell (PRBC) transfusions are important to maintain adequate oxygen delivery to ensure cerebral oxygenation, but can trigger inflammatory responses. This study aimed to evaluate the association between intraoperative PRBC transfusions, dependent on transfusion threshold and postoperative delirium.
Methods
This was a retrospective cohort study analyzing registry data from an academic tertiary-care center in Massachusetts, United States of America. Adult hospitalized patients who underwent general anesthesia for surgery between January 1, 2008 and January 15, 2024 with intraoperative hemoglobin measurements were included. Exclusion criteria included preoperative delirium or cognitive impairment, postoperative ventilation >72 h, an American Society of Anesthesiologists physical status >IV, and missing confounder data. The primary exposure was intraoperative allogeneic PRBC transfusions. The primary outcome was delirium diagnosed within seven days after surgery, identified through diagnostic codes, Confusion Assessment Method, and manual chart review.
Findings
Among 42,313 included patients, 6970 (16.5%) received intraoperative PRBC transfusions, and 2871 (6.8%) developed postoperative delirium. Patients receiving intraoperative PRBC transfusions had a higher risk of postoperative delirium (adjusted odds ratio 1.15, 95% confidence interval 1.03–1.29, P = 0.016) in a dose-dependent manner. The PRBC transfusion-associated risk of delirium was higher when intraoperative hemoglobin was at higher nadirs (P-for-interaction = 0.002). At nadirs below 7.3 g/dL, PRBC transfusions were not associated with delirium.
Interpretation
Intraoperative PRBC transfusions administered at higher hemoglobin nadirs, around 7.3 g/dL, were associated with an increased risk of postoperative delirium, suggesting that greater attention to delirium risk may be warranted among transfused patients and that transfusion triggers during surgery merit careful consideration.
Funding
Department of Anesthesia, Critical Care, and Pain Medicine, Beth Israel Deaconess Medical Center.
Keywords: Red blood cell transfusion, Postoperative delirium, Hemoglobin, Threshold, Restrictive
Research in context.
Evidence before this study
A PubMed search conducted on October 27, 2025, using the terms “red blood cell transfusion,” “delirium,” and “surgery” yielded 43 articles, including three in non-English. Five studies primarily examined the association between packed red blood cell (PRBC) transfusion or hemoglobin thresholds and delirium, but all were limited to patients undergoing hip fracture surgery. The present study addresses this gap by investigating how hemoglobin levels modify the relationship between PRBC transfusion and postoperative delirium across diverse surgical populations.
Added value of this study
This retrospective study included 42,313 patients from an academic hospital in the United States. Overall, 16.5% received intraoperative PRBC transfusions, and 6.8% developed postoperative delirium. After adjusting for patient- and surgery-related confounders, intraoperative PRBC transfusion was associated with a 15% higher risk of postoperative delirium in a dose-dependent manner. The transfusion-related risk was greater at higher intraoperative hemoglobin nadirs. Transfusion was not associated with delirium when nadirs were below 7.3 g/dL.
Implications of all the available evidence
We found that intraoperative PRBC transfusion, particularly at higher thresholds, was associated with an increased risk of postoperative delirium within seven days. These findings suggest that increased attention to delirium risk may be warranted among transfused patients and may inform consideration of transfusion thresholds in perioperative care. Despite extensive confounder adjustment, validation through multicenter interventional studies is warranted.
Introduction
Allogeneic packed red blood cell (PRBC) transfusions are essential for rapidly correcting anemia and restoring blood volume during acute bleeding.1 More than 1.3 million units of PRBCs are transfused intraoperatively each year in the United States.2 However, previous research identified an association between PRBC transfusion and an increased risk of postoperative short- and long-term mortality and morbidity.3,4 Therefore, recent guidelines have recommended more restrictive transfusion thresholds—typically defined as hemoglobin levels of 7–8 g/dL—over liberal transfusion strategies, commonly defined as thresholds of 9–10 g/dL, to minimize PRBC-associated complications.5
Delirium is one of the most common postoperative complications, occurring in about 5–10% of patients after low-risk surgeries and in up to 36–40% after cardiac or trauma surgeries.6 Previous research has linked intraoperative events including hypoxemia, hypocarbia and hypotension to an increased risk of postoperative delirium,7,8 likely due to effects on cerebral blood flow and oxygenation. PRBC transfusions can effectively improve cerebral oxygenation and may help prevent postoperative delirium.7,9 However, allogeneic PRBC transfusions also exacerbate postoperative systemic inflammation, potentially increasing patients’ risk of delirium.10 Previous data on the association between PRBC transfusions and postoperative delirium have been equivocal.11, 12, 13, 14, 15 In light of accumulating evidence supporting cardiovascular safety of a restrictive transfusion strategy,1 it remains unclear whether varying transfusion thresholds may modify the impact of PRBC transfusions on delirium.
This study aimed to examine the association between intraoperative PRBC transfusions and postoperative delirium in a general adult patient population undergoing a broad spectrum of surgical procedures. We hypothesized that PRBC transfusions are associated with a higher risk of postoperative delirium, and that this depended on perioperative hemoglobin levels.
Methods
Study design and ethics approval
This retrospective cohort study was conducted at Beth Israel Deaconess Medical Center (BIDMC), a tertiary academic medical center in Boston, Massachusetts, United States of America. The research protocol was approved by the Institutional Review Board of BIDMC (protocol number: 2024P000991), with a waiver of written informed consent granted due to the study's retrospective design. The manuscript complies with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (eTable 1). Data were obtained from the Anesthesia Research Data Repository (version 5.0) at BIDMC. This institutional repository integrates de-identified data from multiple sources, including the Anesthesia Information Management System for anesthesia records, the Perioperative Information Management System for surgical data, the billing registry for International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9/10-CM) codes, and the Admission-Discharge-Transfer database for patient demographics and hospitalization information.
Study population
Adult hospitalized patients who underwent surgery under general anesthesia at BIDMC between January 2008 and January 2024 and with intraoperative hemoglobin measurements were considered for inclusion in this study. We excluded patients diagnosed with delirium, dementia, or mild cognitive impairment within one month before surgery based on ICD-9/10-CM codes (eTable 2). Additionally, we excluded those requiring postoperative ventilation for more than 72 h, patients with an American Society of Anesthesiologists (ASA) physical status classification >IV, and individuals with missing data for confounding variables.
Primary analysis
The primary exposure in this study was intraoperative allogeneic PRBC transfusion, defined as the administration of PRBCs from the start to the end of surgery. These data were collected from electronic hospital records and the resulting variable was categorized as a binary variable (transfusion vs. no transfusion). The primary outcome was postoperative delirium within seven days after surgery, a period of heightened risk for developing delirium.16 The identification of delirium was determined based on a previously published approach, combining ICD-9/10-CM codes, results of the Confusion Assessment Method in the Intensive Care Unit (CAM-ICU), and a keyword-based search strategy paired with manual chart review (eFig. 1).16, 17, 18
Confounding model
Analyses were adjusted for an a priori defined confounding model. Potential confounders were variables associated with both PRBC transfusion and postoperative delirium, defined based on clinical plausibility and relevance as well as prior literature.11, 12, 13 Variables included demographic characteristics, including age, sex, body mass index (BMI), household income, and federal insurance. Comorbidity-related confounders included the Elixhauser comorbidity score, ASA physical status classification as well as a history of cerebral vascular disease, psychosis, depression, anxiety, pain-related conditions, hypertension, coronary artery disease (CAD), chronic obstructive pulmonary disease (COPD), anemia, tobacco use, drug or alcohol use, and preoperative benzodiazepine use. Procedure-related confounders included surgical service, work relative value units (RVU), year and duration of surgery, emergency status, night surgery (starting before 07:00 or after 17:00), intraoperative hypotension (≥5 min of mean arterial pressure <55 mmHg), intraoperative hypoxemia (≥5 min of pulse oximetry <90%), blood loss calculated using a validated formula based on body weight and hematocrit measured within 30 days preoperatively and 24 h postoperatively,19,20 total opioid dose converted to oral morphine equivalents, total propofol dose, total vasopressor dose converted to norepinephrine equivalents, median age-adjusted minimum alveolar concentration (MAC) of volatile anesthetics and nitrous oxide, use of regional anesthesia, autologous transfusion via cell saver, and transfusion of other allogeneic blood products (including thawed plasma, platelets, and cryoprecipitate). The detailed definitions of confounders are provided in eTable 3.
Secondary and exploratory analyses
In a key secondary analysis, we explored whether the association between intraoperative PRBC transfusions and postoperative delirium was dependent on nadir hemoglobin levels, defined as the lowest hemoglobin value measured from the start of surgery to 2 h postoperatively, accounting for a time lag between blood draw and result availability. An interaction term between the primary exposure PRBC transfusion and the possible modifier “hemoglobin nadir” was included in the confounder model. The hemoglobin nadir corresponding to the point where the 95% confidence interval (CI) of the adjusted odds ratio (ORadj) crossed was used as the cut-off value to stratify the nadir to create a binary variable. The association between hemoglobin nadirs and delirium was further analyzed separately for patients who received PRBC transfusion and those who did not.
To investigate a potential dose-response relationship, we examined whether the association between PRBC transfusion and delirium was magnified with increasing amount of transfused PRBC units, which was categorized as 0, 1, 2, and ≥3 units, and further analyzed as a continuous variable. The relationship between intraoperative PRBC transfusion and postoperative delirium was investigated in subgroups stratified by cardiac and non-cardiac surgeries.
In addition, to examine the impact of inflammation and oxidative stress induced by PRBC storage, we analyzed the association between the storage age of PRBCs and postoperative delirium among patients who received transfusions. Since a patient could receive multiple units of PRBCs with various storage ages, this analysis was conducted at both the unit and the patient level. At the patient level, the storage age of PRBCs was summarized as the maximum storage age among all units received by the patient. This analysis was adjusted for blood types, year of surgery, surgical service, total PRBC units transfused, and transfusion of other allogeneic blood, as blood bank staff were unaware of other confounders when issuing PRBCs.
With an exploratory intent, we described the percentage of patients receiving PRBC transfusion across hemoglobin nadirs. Variability in intraoperative PRBC transfusions across individual anesthesia providers was analyzed among anesthesiologists who managed more than 50 cases.
Sensitivity analyses
Multiple sensitivity analyses were conducted to confirm the robustness of our findings. (1) Preoperative anemia, which was defined initially using ICD-9/10-CM codes, was replaced with the most recent hemoglobin level measured within 30 days before surgery. (2) In patients with negative calculated blood loss values, we assigned them as zero in the primary analysis but in a sensitivity analysis, these values were imputed using random values between 0 and 100 mL. (3) The first intraoperative hematocrit measurement was used as the baseline for calculating blood loss in patients without a preoperative hematocrit measurement. These patients were excluded in a sensitivity analysis. (4) Data on estimated blood loss, jointly assessed by surgeons and anesthesiologists, became available for surgeries performed after 2020. A sensitivity analysis was conducted among these patients, replacing calculated blood loss with estimated blood loss. (5) Multiple imputation was performed to address missing data for all potential confounders. (6) Anesthesia provider was included in the model to account for provider-level variability in transfusion practices. (7) A validated PRBC transfusion risk prediction score was additionally adjusted for in the model in a post-hoc analysis.21
Statistical analysis
The distribution of confounding variables was evaluated using absolute standardized differences (ASD), with values greater than 0.1 considered imbalanced. Collinearity among the exposure and confounders was assessed using a Spearman correlation matrix, where coefficients below 0.7 indicated strong collinearity. In the primary analysis, confounding was adjusted using multivariable logistic regression and propensity score overlap weighting, which was chosen because it avoids the extreme weights that can occur with inverse probability of treatment weighting and can achieve covariate balance without requiring trimming or truncation.22 Propensity scores were estimated via logistic regression with PRBC transfusion as the dependent variable. Overlap weights were assigned as one minus the propensity score for the transfused group and the propensity score for the non-transfused group. After applying overlap weighting, ASDs were recalculated to confirm balance. When confounder balance was achieved, overlap-weighted logistic regression models were fitted with postoperative delirium as the dependent variable and PRBC transfusion as the independent variable. The results were reported as ORs and absolute risk differences (ARDs) with 95% CIs. Model performance was evaluated using the area under the receiver operating characteristic curve (AUROC) and the Hosmer–Lemeshow goodness-of-fit test. The robustness of the observed associations to potential unmeasured confounding was further assessed using the E-value.23
Restricted cubic splines were used to explore potential non-linear relationships, with degrees of freedom (3, 4, or 5) selected based on the minimum Akaike Information Criterion. Non-linearity was considered present if any spline term had a P < 0.05. Unit-level analysis of storage age was conducted using within-cluster resampling to prevent overrepresentation of large clusters and ensure balanced sampling across clusters. Pseudo-datasets were generated by randomly selecting one PRBC unit transfused per patient (without replacement) from the original dataset, and a logistic regression model was fitted to each pseudo-dataset. This process was repeated 1000 times to obtain stable parameter estimates. The OR was calculated as the average of estimates across iterations, and the pooled CI was derived by incorporating both within- and between-cluster variance. Variability in transfusion practices across anesthesia providers was assessed using mixed-effects logistic regression models with providers included as random intercepts.
Based on internal audits indicating that approximately 15% of major surgical patients receive PRBC transfusions and an estimated 5% incidence of postoperative delirium, we conducted an a priori power analysis. Assuming a sample size of 40,000 patients and a two-sided α of 0.05, the study had 93% power to detect an OR ≤ 0.80 and 82% power to detect an OR ≥ 1.20. Statistical analyses were performed using Stata (version MP 18.0; StataCorp LLC, USA) and R (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria) with the following packages: “tableone”, “survey”, “WeightIt”, “cobalt”, “mice”, “rms”, “lme4”, “pwr”, “pROC”, and “EValue”.
Role of the funding source
The funders had no involvement in study design, data collection, data analyses, data interpretation, or the writing of the report.
Results
A total of 50,521 patients were considered for inclusion in this study. After applying exclusion criteria (Fig. 1), the final cohort consisted of 42,313 patients. A total of 6970 patients (16.5%) received intraoperative PRBC transfusions, with a median of 2 units (interquartile range [IQR], 1–3). All transfused PRBC units underwent pre-storage leukoreduction. Postoperative delirium occurred in 2871 patients (6.8%). Among those who received PRBC transfusion, 734 (10.5%) patients developed postoperative delirium, compared to 2137 (6.1%) among patients not transfused. Most confounders were imbalanced between patients who received PRBC transfusion and those who did not (Table 1). The correlation matrix between all pairs of confounders indicated absence of strong collinearity (eFig. 2).
Fig. 1.
Study flow diagram. Abbreviation: ASA, American Society of Anesthesiologist Physical Status Classification; BMI, body mass index; RVU, relative value units; PRBC, packed red blood cell; n, number.
Table 1.
Patient characteristics.
| Variables | All patients (n = 42,313) | Transfusion (n = 6970) | No transfusion (n = 35,343) | ASD | ASD after PS-OW |
|---|---|---|---|---|---|
| Demographics | |||||
| Age (year) | 64 (15) | 66 (15) | 63 (15) | 0.20 | <0.01 |
| Sex | 0.19 | <0.01 | |||
| Male | 23,846 (56.4%) | 3390 (48.6%) | 20,456 (57.9%) | ||
| Female | 18,467 (43.6%) | 3580 (51.4%) | 14,887 (42.1%) | ||
| BMI (kg/m2) | 28.5 (6.4) | 27.8 (6.2) | 28.6 (6.4) | 0.13 | <0.01 |
| Household income ($/year) | 106,910 (38,332) | 104,915 (36,761) | 107,304 (38,623) | 0.06 | <0.01 |
| Federal insurance | 22,101 (52.2%) | 4087 (58.6%) | 18,014 (51.0%) | 0.16 | <0.01 |
| Comorbidities | |||||
| Elixhauser score | 10 [3, 19] | 15 [6, 24] | 9 [2, 18] | 0.40 | <0.01 |
| ASA Status | 0.45 | <0.01 | |||
| I and II | 7894 (18.7%) | 691 (9.9%) | 7203 (20.4%) | ||
| III | 20,561 (48.6%) | 2844 (40.8%) | 17,717 (50.1%) | ||
| IV | 13,858 (32.8%) | 3435 (49.3%) | 10,423 (29.5%) | ||
| Cerebral vascular disease | 7105 (16.8%) | 1312 (18.8%) | 5793 (16.4%) | 0.06 | <0.01 |
| Psychosis | 353 (0.8%) | 75 (1.1%) | 278 (0.8%) | 0.03 | <0.01 |
| Depression | 5653 (13.4%) | 995 (14.3%) | 4658 (13.2%) | 0.03 | <0.01 |
| Anxiety | 4797 (11.3%) | 817 (11.7%) | 3980 (11.3%) | 0.01 | <0.01 |
| Pain-related condition | 21,180 (50.1%) | 3819 (54.8%) | 17,361 (49.1%) | 0.11 | <0.01 |
| Hypertension | 27,171 (64.2%) | 4551 (65.3%) | 22,620 (64.0%) | 0.03 | <0.01 |
| CAD | 16,174 (38.2%) | 3169 (45.5%) | 13,005 (36.8%) | 0.18 | <0.01 |
| COPD | 3137 (7.4%) | 528 (7.6%) | 2609 (7.4%) | 0.01 | <0.01 |
| Anemia | 2468 (5.8%) | 708 (10.2%) | 1760 (5.0%) | 0.20 | <0.01 |
| Tobacco use | 16,034 (37.9%) | 2469 (35.4%) | 13,565 (38.4%) | 0.06 | <0.01 |
| Drug or alcohol use | 3046 (7.2%) | 669 (9.6%) | 2377 (6.7%) | 0.11 | <0.01 |
| Preoperative benzodiazepine use | 4423 (10.5%) | 722 (10.4%) | 3701 (10.5%) | <0.01 | <0.01 |
| Procedural characteristics | |||||
| Year of surgery (≥2012) | 19,168 (45.3%) | 2921 (41.9%) | 16,247 (46.0%) | 0.08 | |
| Surgical Service | 0.36 | <0.01 | |||
| General | 13,997 (33.1%) | 1719 (24.7%) | 12,278 (34.7%) | ||
| Orthopedic | 7134 (16.9%) | 1241 (17.8%) | 5893 (16.7%) | ||
| Transplant | 1971 (4.7%) | 507 (7.3%) | 1464 (4.1%) | ||
| Vascular | 4555 (10.8%) | 943 (13.5%) | 3612 (10.2%) | ||
| Neurology | 3253 (7.7%) | 266 (3.8%) | 2987 (8.5%) | ||
| Trauma | 1071 (2.5%) | 121 (1.7%) | 950 (2.7%) | ||
| Cardiac | 10,332 (24.4%) | 2173 (31.2%) | 8159 (23.1%) | ||
| Work RVU | 25.4 [17.5, 42.6] | 30.3 [18.7, 44.4] | 25.2 [17.4, 41.5] | 0.29 | <0.01 |
| Surgery duration (min) | 249 [171, 325] | 294 [221, 377] | 241 [164, 314] | 0.42 | <0.01 |
| Emergency | 7580 (17.9%) | 1743 (25.0%) | 5837 (16.5%) | 0.21 | <0.01 |
| Night surgery | 3003 (7.1%) | 665 (9.5%) | 2338 (6.6%) | 0.11 | <0.01 |
| Intraoperative hypotension | 11,104 (26.2%) | 2564 (36.8%) | 8540 (24.2%) | 0.28 | <0.01 |
| Intraoperative hypoxemia | 13,118 (31.0%) | 2610 (37.4%) | 10,508 (29.7%) | 0.16 | <0.01 |
| Hemoglobin nadir (g/dL) | 10.3 (2.2) | 8.31 (1.8) | 10.7 (2.0) | 1.26 | <0.01 |
| Calculated blood loss (mL) | 586 [154, 1072] | 624 [226, 1096] | 272 [0, 902] | 0.28 | <0.01 |
| Opioid dose (mg OME) | 74.9 [40.0, 250.0] | 92.0 [47.8, 250.0] | 71.5 [37.8, 222.5] | 0.05 | <0.01 |
| Propofol dose (mg) | 180 [113, 240] | 150 [85, 202] | 186 [120, 249] | 0.07 | <0.01 |
| Vasopressor dose (mg norepinephrine equivalents) | 0.32 [0.05, 0.77] | 0.61 [0.21, 1.28] | 0.27 [0.04, 0.68] | 0.28 | <0.01 |
| Age-adjusted MAC | 0.9 [0.8, 1.1] | 0.9 [0.8, 1.1] | 0.9 [0.8, 1.1] | 0.06 | <0.01 |
| Regional anesthesia | 5556 (13.1%) | 741 (10.6%) | 4815 (13.6%) | 0.09 | <0.01 |
| Autologous transfusion | 9133 (21.6%) | 2152 (30.9%) | 6981 (19.8%) | 0.26 | <0.01 |
| Transfusion of other allogeneic blood products | 3525 (8.3%) | 2010 (28.8%) | 1515 (4.3%) | 0.70 | <0.01 |
Continuous variables were reported as mean (standard deviation) for normally distributed data and median [interquartile range] for non-normally distributed data. Categorical variables were presented as number (percentage). Abbreviations: ASD, absolute standardized difference; PS-OW, propensity score overlap weighing; BMI, body mass index; ASA, American Society of Anesthesiologist Physical Status Classification; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; RVU, relative value units; OME, oral morphine equivalent; MAC, minimum alveolar concentration.
Primary analysis
In both unadjusted and adjusted analyses transfusion of PRBCs was associated with a higher risk of postoperative delirium (unadjusted regression: OR 1.83 [95% CI 1.67–2.00] and ARD 4.5% [95% CI 3.7%–5.2%], P < 0.001; adjusted regression: ORadj 1.15 [95% CI 1.03–1.29] and ARDadj 0.9% [95% CI 0.2%–1.6%], P = 0.016; eTable 4). The logistic regression model showed an AUROC of 0.75 and a Hosmer–Lemeshow test P value of 0.065 (eFig. 3). The E-values for the ORadj and the lower bound of its 95% CI were 1.57 and 1.21, respectively. Using propensity score overlap weighting (eTable 5), the distribution of all confounders achieved balance (Table 1 and eFig. 4), and PRBC transfusions were associated with a higher risk of postoperative delirium (ORadj 1.18 [95% CI 1.06–1.33] and ARDadj 1.0% [95% CI 0.3%–1.8%], P = 0.004). The E-values were 1.64 for the ORadj and 1.31 for the lower CI bound.
Secondary analyses
In the key secondary analysis, restricted cubic splines did not indicate a non-linear relationship between hemoglobin nadirs and delirium (eTable 6 and eFig. 5). A significant interaction between intraoperative PRBC transfusion and hemoglobin nadirs was observed (P-for-interaction = 0.002; eTable 7). PRBC-associated risk increased with higher hemoglobin nadirs (Fig. 2). The lower margin of the 95% CI for ORadj exceeded 1.0 when hemoglobin nadirs were above 7.3 g/dL, indicating increased risk of delirium with PRBC transfusions at a hemoglobin nadir >7.3 g/dL. When stratifying patients based on a hemoglobin nadir cut-off of 7.3 g/dL, PRBC transfusion was not associated with postoperative delirium among patients with hemoglobin nadirs <7.3 g/dL (ORadj 1.15, 95% CI 0.88–1.51, P = 0.31; eTable 8), but among those with hemoglobin nadirs ≥7.3 g/dL (ORadj 1.19, 95% CI 1.05–1.35, P = 0.006; eTable 9). A lower hemoglobin nadir was associated with an increased risk of postoperative delirium among patients who did not receive RBC transfusion (ORadj 1.04 per 1 g/dL-decrease in hemoglobin, 95% CI 1.01–1.06, P = 0.012; eFig. 6 and eTable 10), but no association was observed among patients who received PRBC transfusion (ORadj 1.03 per 1 g/dL-decrease, 95% CI 0.97–1.10, P = 0.32; eFig. 6 and eTable 11).
Fig. 2.
Association between intraoperative packed red blood cell transfusion and postoperative delirium across various hemoglobin nadirs. ORs and 95% CIs were estimated using 1000 bootstrap iterations based on a multivariable logistic regression model that included an interaction term between packed red blood cell transfusion and hemoglobin nadir levels, as well as all the demographic, comorbidity-, and procedure-related confounders. Patients who did not receive PRBC transfusion served as the reference (OR = 1, dashed black line). The lower limit of the 95% confidence interval crossed 1 at a hemoglobin nadir level of 7.3 g/dL. Abbreviations: OR, odds ratio; CI, confidence interval.
A dose-response relationship was observed in the multivariable logistic regression model, when treating PRBC units as an ordinal variable (P-for-trend<0.001; Fig. 3A and eTable 12) or as a continuous variable (ORadj 1.03 per 1 unit, 95% CI 1.00–1.06; P = 0.024; Fig. 3B and eTable 13). In the subgroup analysis, patients receiving PRBC transfusion had a higher risk of postoperative delirium in non-cardiac surgeries (ORadj 1.19, 95% CI 1.04–1.36, P = 0.013). No association was observed in cardiac surgeries (ORadj 1.02, 95% CI 0.82–1.27, P = 0.86; eTable 14).
Fig. 3.
Dose-responses analysis of the units of PRBC transfused and the risk of postoperative delirium analyzed as an ordinal variable (A) and as a continuous variable (B). The analysis was adjusted for all the demographic, comorbidity-, and procedure-related confounders, including calculated blood loss, using a multivariable logistic regression model. A P-for-trend was calculated by including the median value of each category in the model as a continuous variable. The risk of postoperative delirium was estimated by calculating the average predicted probabilities and corresponding confidence intervals across a sequence of values for the PRBC transfusion amount ranging from 0.1% to 0.9%, while keeping all other confounders intact. Abbreviations: n, number; ARD, absolute risk difference; CI, confidence interval; PRBC, packed red blood cell.
Median PRBC storage time was 22 days (IQR, 15–29). A longer storage age of PRBCs was not associated with a higher risk of delirium at both the patient level (ORadj 0.94 per 7 days, 95% CI 0.88–1.01, P = 0.10, eTable 15) and the unit level (ORadj 0.95 per 7 days, 95% CI 0.88–1.02, P = 0.14). Similarly, no association was observed when categorizing the storage age by quartiles as thresholds (P-for-trend = 0.15, eFig. 7 and eTable 16) or using restricted cubic splines (eFig. 8 and eTable 17) at the patient level.
Exploratory analysis
The percentage of patients who received intraoperative PRBC transfusion decreased as hemoglobin nadirs increased (Fig. 4A). Mixed-effects models revealed substantial variability in the adjusted odds of PRBC transfusion among providers, with ORadj (95% CI) ranging from 0.45 (0.25–0.82) to 1.97 (1.21–3.21) compared to the providers with the median transfusion odds (Fig. 4B and eTable 18).
Fig. 4.
(A) Number and percentage of patients who received intraoperative red blood cell transfusion across various hemoglobin level nadirs. (B) Adjusted odds ratios and 95% confidence intervals of transfusion odds among anesthesia providers compared with the provider with the median transfusion odds. The analysis was adjusted for demographic confounders, Elixhauser score, American Society of Anesthesiologists Physical Status, cerebral vascular disease, coronary artery disease, chronic obstructive pulmonary disease, baseline hemoglobin level, work relative value units, year and duration of surgery, surgical service, emergency, and calculated blood loss, using a mixed-effects model.
Sensitivity analysis
Results of all sensitivity analyses were consistent with the primary analysis (eTable 19), including analyses replacing ICD code-defined preoperative anemia with baseline hemoglobin level (ORadj 1.27, 95% CI 1.14–1.42, P < 0.001, eTable 20), imputing negative calculated blood loss values with random values between 0 and 100 mL (ORadj 1.15, 95% CI 1.02–1.29, P = 0.017, eTable 21), excluding patients without preoperative hematocrit measurements (ORadj 1.16, 95% CI 1.03–1.31, P = 0.011, eTable 22), replacing calculated with estimated blood loss (ORadj 1.31, 95% CI 1.01–1.70, P = 0.041, eTable 23), addressing missing confounder data through multiple imputation (ORadj 1.16, 95% CI 1.04–1.30, P = 0.007, eTable 24), accounting for provider-level variability in transfusion practices (ORadj 1.15, 95% CI 1.02–1.30, P = 0.021, eTable 25), and additionally adjusting for the PRBC transfusion prediction score (ORadj 1.16, 95% CI 1.03–1.30, P = 0.017; P-for-interaction = 0.003 for hemoglobin nadir, eTable 26).
Discussion
This retrospective cohort study including 42,313 hospitalized patients undergoing surgery with intraoperative hemoglobin measurements indicated that intraoperative allogeneic PRBC transfusions were associated with a higher risk of postoperative delirium in a dose-dependent manner. This association was dependent on intraoperative hemoglobin nadir and magnified at higher hemoglobin nadir levels. Findings were robust across several sensitivity analyses and were adjusted for patient- and procedure-related confounder, including blood loss and the administration of other fluids.
We observed that the transfusion-associated risk of delirium was more pronounced in patients with higher hemoglobin nadirs. The hemoglobin nadirs likely reflected the lowest hemoglobin levels physicians were willing to tolerate when deciding whether to proceed with a transfusion, and suggest a transition in risk direction at hemoglobin levels around 7.3 g/dL. This value represents a data-driven estimate derived from the continuous exposure–response relationship rather than a definitive threshold, and is consistent with the transfusion threshold of 7–8 g/dL recommended in restrictive transfusion strategies by the American Association of Blood Banks guidelines.5 Notably, these recommendations are based on evidence regarding postoperative mortality, myocardial infarction, stroke, and infection,1 while the optimal hemoglobin threshold for minimizing risk of delirium remained unclear; our findings help address this gap. We acknowledge that clinicians may be inclined to initiate transfusion at higher hemoglobin thresholds in patients with greater comorbidity burden or bleeding risk, or in those undergoing more aggressive surgical procedures, which could introduce confounding by clinical indication. To mitigate this concern, we extensively adjusted for patient- and procedure-related confounders, including the Elixhauser Comorbidity Index, ASA physical status, work RVU, surgical duration, and estimated blood loss, achieving satisfactory covariate balance using propensity score overlap weighting. Provider-level variability in transfusion practices was accounted for in sensitivity analyses. To further address residual confounding, effect modification by hemoglobin nadir was re-evaluated after adjustment for a validated PRBC transfusion prediction score, with consistent results.
Previous studies investigated a potential relationship between PRBC transfusion and postoperative delirium. A secondary analysis of a multicenter randomized controlled trial found that anemia was a risk factor for delirium among hip fracture patients with a median age over 80 years, which is in line with our findings. The authors further observed that PRBC protected from delirium in patients with nadirs below 9.7 g/dL.24 However, due to the limited sample size, the study was unable to further segregate hemoglobin nadirs, and it is possible that the hemoglobin nadir associated with a protective effect of PRBC transfusion in this study was overestimated.24 Another randomized controlled trial conducted in a similar population showed that a liberal transfusion threshold of 10 g/dL did not increase the severity of postoperative delirium compared to a restrictive threshold of 8 g/dL; however, incidences of delirium were not compared.25 We found that a higher hemoglobin nadir was associated with a lower risk of delirium among patients who did not receive PRBC transfusion, highlighting the potential benefit of correcting anemia through preoperative nutrient supplementation. Previous studies have suggested a target hemoglobin level of 13 g/dL for both males and females.26
Our cohort encompassed patients undergoing a diverse range of procedures, which broadens the scope of investigation relative to previous studies that primarily focused on orthopedic surgeries.11, 12, 13, 14, 15 Of note, our subgroup analysis revealed PRBC transfusions were associated with higher odds of delirium in noncardiac surgeries, while this was not the case for cardiac surgeries. Cardiac patients are more likely to experience inadequate cerebral perfusion and oxygenation as a result of hemodynamic instability and the cardiopulmonary bypass.27 Therefore, optimization of oxygen delivery through PRBC transfusion may be particularly important in cardiac patients.
In this study, the risk of delirium did not increase with longer storage duration of transfused PRBCs, consistent with findings from a multicenter RCT among critically ill, pediatric patients.28 These findings do not support the practice of avoiding older PRBCs as a strategy to prevent delirium. We also observed a wide range of adjusted transfusion odds between anesthesiologists. Similarly, previous research has demonstrated significant variability in transfusion practice across institutions.29 This highlights the necessity of standardizing transfusion practices through audits and physician education.
A possible pathophysiological mechanism linking PRBC transfusion to postoperative delirium is neuroinflammation and oxidative stress triggered by the transfusion.30 For example, prior metabolomic and proteomic studies showed that PRBC transfusion, even when leukoreduced, leads to the accumulation of reactive oxygen species and the release of inflammatory cytokines.31 The transfusion of stored PRBCs led to the generation of free heme and non-transferrin-bound iron in healthy human volunteers.32 This process has been shown to further induce neuroinflammation and cognitive impairment in rats.33 However, it is unclear whether these changes contribute to delirium in humans. Previous studies have also found that regional cerebral desaturation, as measured by near-infrared spectroscopy, is associated with postoperative delirium.34,35 Effect modification by hemoglobin nadir is biologically plausible and may reflect a balance between transfusion-related inflammation and improved oxygen delivery. At lower hemoglobin levels, the benefit of improved oxygenation may outweigh these inflammatory harms, whereas at higher hemoglobin nadirs, the incremental oxygen benefit is likely limited and the inflammatory burden may predominate.
This study has limitations. First, the observational design of this study limited our ability to establish causal relationships due to the potential residual confounding effects, including confounding by clinical indication for PRBC transfusion. Moreover, the E-value corresponding to the lower 95% CI bound for the ORadj in the primary analysis suggests that unmeasured confounding with an effect size greater than 1.21 could potentially attenuate the association to non-significance. However, this study employed an a priori-defined confounding adjustment strategy that combined logistic regression with propensity score overlap weighting, effectively mitigating confounder imbalance. In addition, multiple sensitivity analyses were conducted to further minimize the impact of residual confounding. Second, the findings of this single-center study should be generalized to other institutions with caution, as perioperative management standards may vary across settings. Third, the blood loss was calculated from pre- and postoperative hematocrits, which may not closely align with gravimetric methods or visual estimation. However, this method provides the advantage of capturing occult blood loss and has been employed in several large trials.19,20 The results remained consistent across multiple adjustments performed during the sensitivity analyses. Finally, some patients were excluded from the cohort due to missing confounder data. However, they represented less than 10% of the population meeting the inclusion criteria, and the results remained robust after multiple imputation was applied to address the missing data.
In conclusion, intraoperative PRBC transfusions, especially at higher thresholds, were associated with an increased risk of seven-day postoperative delirium. These findings suggest that heightened attention to delirium risk may be warranted among patients receiving PRBC transfusions and support careful consideration of transfusion thresholds in clinical practice.
Contributors
XX contributed to conceptualisation, data curation, formal analysis, investigation, methodology, visualisation, and writing of the original draft of the manuscript. LJW contributed to data curation, investigation, methodology, and writing of the original draft. BSP and SR contributed to data curation and investigation. EA contributed to data curation and writing of the original draft. GC contributed to methodology. TT, MM, HM, and KO contributed to data curation. DT contributed to project administration, resources, and supervision. MSS contributed to conceptualisation, data curation, formal analysis, funding acquisition, investigation, methodology, software, project administration, resources, supervision, and writing of the original draft. All authors read and approved the final version of the manuscript. XX and MSS accessed and verified the underlying data.
Data sharing statement
Due to the sensitive nature of the data collected for this study, requests to access the dataset from qualified researchers trained in human subject research and confidentiality may be sent to Maximilian S. Schaefer at msschaef@bidmc.harvard.edu.
Declaration of interests
None of the authors reported disclosures pertaining to this manuscript. All other relationships: XX received funding from National Natural Science Foundation of China (project number: 72304281) and Peking Union Medical College Hospital Talent Cultivation Program (Category D: UHB11977). She is an editor of BioMed Central Anesthesiology. EA is an associate editor for BioMed Central Anesthesiology. TT received a grant from the Deutsche Forschungs-Gemeinschaft (DFG, German Research Foundation - Walter Benjamin Fellowship, project number: 522518834). LJW received funding for an investigator-initiated study from Merck & Co. He is an associate editor for BioMed Central Anesthesiology. MSS received funding for investigator-initiated studies from Merck & Co. He is an associate editor for BioMed Central Anesthesiology. He received honoraria for lectures from Mindray Medical International Limited and an unrestricted philanthropic grant from Jeffrey and Judith Buzen. All other authors declare no competing interests.
Acknowledgements
This study was supported by institutional and departmental funding from the Department of Anesthesia, Critical Care and Pain Medicine at Beth Israel Deaconess Medical Center, United States, without specific grant numbers. Preliminary data of this study was presented at the American Society of Anesthesiologists annual meeting in Oct 2025 in San Antonio, Texas (abstract number: A3023).
Footnotes
Supplementary data related to this article can be found at https://doi.org/10.1016/j.eclinm.2026.103885.
Appendix A. Supplementary data
References
- 1.Carson J.L., Stanworth S.J., Dennis J.A., et al. Transfusion thresholds for guiding red blood cell transfusion. Cochrane Database Syst Rev. 2021;12(12) doi: 10.1002/14651858.CD002042.pub5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Jones J.M., Sapiano M.R.P., Mowla S., Bota D., Berger J.J., Basavaraju S.V. Has the trend of declining blood transfusions in the United States ended? Findings of the 2019 National Blood Collection and Utilization Survey. Transfusion. 2021;61 Suppl 2(Suppl 2):S1–S10. doi: 10.1111/trf.16449. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Morris F.J.D., Fung Y.L., Craswell A., Chew M.S. Outcomes following perioperative red blood cell transfusion in patients undergoing elective major abdominal surgery: a systematic review and meta-analysis. Br J Anaesth. 2023;131(6):1002–1013. doi: 10.1016/j.bja.2023.08.032. [DOI] [PubMed] [Google Scholar]
- 4.Xu X., Zhang Y., Gan J., Ye X., Yu X., Huang Y. Association between perioperative allogeneic red blood cell transfusion and infection after clean-contaminated surgery: a retrospective cohort study. Br J Anaesth. 2021;127(3):405–414. doi: 10.1016/j.bja.2021.05.031. [DOI] [PubMed] [Google Scholar]
- 5.Carson J.L., Stanworth S.J., Guyatt G., et al. Red blood cell transfusion: 2023 AABB International Guidelines. JAMA. 2023;330(19):1892–1902. doi: 10.1001/jama.2023.12914. [DOI] [PubMed] [Google Scholar]
- 6.Swarbrick C.J., Partridge J.S.L. Evidence-based strategies to reduce the incidence of postoperative delirium: a narrative review. Anaesthesia. 2022;77(Suppl 1):92–101. doi: 10.1111/anae.15607. [DOI] [PubMed] [Google Scholar]
- 7.Ahrens E., Tartler T.M., Suleiman A., et al. Dose-dependent relationship between intra-procedural hypoxaemia or hypocapnia and postoperative delirium in older patients. Br J Anaesth. 2023;130(2):e298–e306. doi: 10.1016/j.bja.2022.08.032. [DOI] [PubMed] [Google Scholar]
- 8.Wachtendorf L.J., Azimaraghi O., Santer P., et al. Association between intraoperative arterial hypotension and postoperative delirium after noncardiac surgery: a retrospective multicenter cohort study. Anesth Analg. 2022;134(4):822–833. doi: 10.1213/ANE.0000000000005739. [DOI] [PubMed] [Google Scholar]
- 9.Gouvea Bogossian E., Rass V., Lindner A., et al. Factors associated with brain tissue oxygenation changes after RBC transfusion in acute brain injury patients. Crit Care Med. 2022;50(6):e539–e547. doi: 10.1097/CCM.0000000000005460. [DOI] [PubMed] [Google Scholar]
- 10.Brummel N.E., Hughes C.G., McNeil J.B., et al. Systemic inflammation and delirium during critical illness. Intensive Care Med. 2024;50(5):687–696. doi: 10.1007/s00134-024-07388-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Raphael J., Hensley N.B., Chow J., et al. Red blood cell transfusion and postoperative delirium in hip fracture surgery patients: a retrospective observational cohort study. Anesthesiol Res Pract. 2021;2021 doi: 10.1155/2021/8593257. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kim S., Oh T.K., Song I.A. Perioperative blood transfusion and delirium after total knee or hip arthroplasty: retrospective analysis. J Pers Med. 2024;14(6):576. doi: 10.3390/jpm14060576. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.OuYang C.L., Hao X.Y., Yu Y., et al. Intraoperative allogeneic transfusion is associated with postoperative delirium in older patients after total knee and hip arthroplasty. Front Surg. 2022;9 doi: 10.3389/fsurg.2022.1048197. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Elsamadicy A.A., Adil S.M., Charalambous L., et al. Independent association between type of intraoperative blood transfusion and postoperative delirium after complex spinal fusion for adult deformity correction. Spine (Phila Pa 1976) 2020;45(4):268–274. doi: 10.1097/BRS.0000000000003260. [DOI] [PubMed] [Google Scholar]
- 15.Kwon Y.S., Kim J.H., Lee J.J., Seo E.M. The relationship between perioperative blood transfusion and postoperative delirium in patients undergoing spinal fusion surgery: clinical data warehouse analysis. Medicina (Kaunas) 2022;58(2):268. doi: 10.3390/medicina58020268. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Ma H., Ahrens E., Wachtendorf L.J., et al. Intraoperative use of phenylephrine versus ephedrine and postoperative delirium: a multicenter retrospective cohort study. Anesthesiology. 2024;140(4):657–667. doi: 10.1097/ALN.0000000000004774. [DOI] [PubMed] [Google Scholar]
- 17.Kaiser L., Paschold B.S., Wachtendorf L.J., et al. Dose-dependent relationship between intra-operative ketamine administration and postoperative delirium: a retrospective cohort study. Anaesthesia. 2025;80(11):1360–1369. doi: 10.1111/anae.16681. [DOI] [PubMed] [Google Scholar]
- 18.Ahrens E., Wachtendorf L.J., Shay D., et al. Association between neuromuscular blockade and its reversal with postoperative delirium in older patients: a Hospital Registry Study. Anesth Analg. 2025;141(2):363–372. doi: 10.1213/ANE.0000000000007489. [DOI] [PubMed] [Google Scholar]
- 19.Camarasa M.A., Olle G., Serra-Prat M., et al. Efficacy of aminocaproic, tranexamic acids in the control of bleeding during total knee replacement: a randomized clinical trial. Br J Anaesth. 2006;96(5):576–582. doi: 10.1093/bja/ael057. [DOI] [PubMed] [Google Scholar]
- 20.Sentilhes L., Senat M.V., Le Lous M., et al. Tranexamic acid for the prevention of blood loss after cesarean delivery. N Engl J Med. 2021;384(17):1623–1634. doi: 10.1056/NEJMoa2028788. [DOI] [PubMed] [Google Scholar]
- 21.Eyth A., Borngaesser F., Rudolph M.I., et al. Development and validation of a risk model to predict intraoperative blood transfusion. JAMA Netw Open. 2025;8(4) doi: 10.1001/jamanetworkopen.2025.5522. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Zajichek A.M., Grunkemeier G.L. Propensity scores used as overlap weights provide exact covariate balance. Eur J Cardiothorac Surg. 2024;66(3) doi: 10.1093/ejcts/ezae318. [DOI] [PubMed] [Google Scholar]
- 23.VanderWeele T.J., Ding P. Sensitivity analysis in observational research: introducing the E-Value. Ann Intern Med. 2017;167(4):268–274. doi: 10.7326/M16-2607. [DOI] [PubMed] [Google Scholar]
- 24.van der Zanden V., Beishuizen S.J., Scholtens R.M., de Jonghe A., de Rooij S.E., van Munster B.C. The effects of blood transfusion on delirium incidence. J Am Med Dir Assoc. 2016;17(8):748–753. doi: 10.1016/j.jamda.2016.04.008. [DOI] [PubMed] [Google Scholar]
- 25.Gruber-Baldini A.L., Marcantonio E., Orwig D., et al. Delirium outcomes in a randomized trial of blood transfusion thresholds in hospitalized older adults with hip fracture. J Am Geriatr Soc. 2013;61(8):1286–1295. doi: 10.1111/jgs.12396. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Shander A., Corwin H.L., Meier J., et al. Recommendations from the international consensus conference on anemia management in surgical patients (ICCAMS) Ann Surg. 2023;277(4):581–590. doi: 10.1097/SLA.0000000000005721. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Smulter N., Lingehall H.C., Gustafson Y., et al. Disturbances in oxygen balance during cardiopulmonary bypass: a risk factor for postoperative delirium. J Cardiothorac Vasc Anesth. 2018;32(2):684–690. doi: 10.1053/j.jvca.2017.08.035. [DOI] [PubMed] [Google Scholar]
- 28.Traube C., Tucci M., Nellis M.E., et al. Transfusion-associated delirium in children: no difference between short storage versus standard issue RBCs. Crit Care Med. 2022;50(2):173–182. doi: 10.1097/CCM.0000000000005393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Fitzgerald D.C., Simpson A.N., Baker R.A., et al. Determinants of hospital variability in perioperative red blood cell transfusions during coronary artery bypass graft surgery. J Thorac Cardiovasc Surg. 2022;163(3):1015–1024.e1. doi: 10.1016/j.jtcvs.2020.04.141. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Maldonado J.R. Neuropathogenesis of delirium: review of current etiologic theories and common pathways. Am J Geriatr Psychiatry. 2013;21(12):1190–1222. doi: 10.1016/j.jagp.2013.09.005. [DOI] [PubMed] [Google Scholar]
- 31.D'Alessandro A., Anastasiadi A.T., Tzounakas V.L., et al. Red blood cell metabolism in vivo and in vitro. Metabolites. 2023;13(7):793. doi: 10.3390/metabo13070793. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Rapido F., Brittenham G.M., Bandyopadhyay S., et al. Prolonged red cell storage before transfusion increases extravascular hemolysis. J Clin Invest. 2017;127(1):375–382. doi: 10.1172/JCI90837. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wei X., Zhang F., Cheng D., et al. Free heme induces neuroinflammation and cognitive impairment by microglial activation via the TLR4/MyD88/NF-kappaB signaling pathway. Cell Commun Signal. 2024;22(1):16. doi: 10.1186/s12964-023-01387-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Eertmans W., De Deyne C., Genbrugge C., et al. Association between postoperative delirium and postoperative cerebral oxygen desaturation in older patients after cardiac surgery. Br J Anaesth. 2020;124(2):146–153. doi: 10.1016/j.bja.2019.09.042. [DOI] [PubMed] [Google Scholar]
- 35.Cui F., Zhao W., Mu D.L., et al. Association between cerebral desaturation and postoperative delirium in thoracotomy with one-lung ventilation: a prospective cohort study. Anesth Analg. 2021;133(1):176–186. doi: 10.1213/ANE.0000000000005489. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.