Association between intraoperative blood transfusion and postoperative delirium in elderly patients undergoing total hip arthroplasty with combined spinal-epidural anesthesia: an IPTW-adjusted retrospective cohort study Authors: Yi Liu, Kun Wang, Juan Peng, Qing Li, Xianting Ke Published: 2025 ABSTRACT AbstractBackgroundPerioperative blood transfusion has been considered a potential risk factor for postoperative delirium (POD) in elderly patients. However, the association remains controversial, and prior studies have lacked a quantitative evaluation of the dose–response relationship.MethodsThis single-center retrospective cohort study included 610 patients aged 65 years or older who underwent their first total hip arthroplasty between January 2020 and April 2025. Patients were stratified into three groups according to intraoperative red blood cell transfusion volume: no transfusion, low-volume transfusion (< 4 units), and high-volume transfusion (≥ 4 units). Propensity score–based inverse probability of treatment weighting (IPTW) was utilized to control for baseline confounding and to evaluate the association between intraoperative transfusion and POD.ResultsIPTW-adjusted multivariable regression analysis revealed that overall transfusion was not associated with POD, but a significant dose–response was found when transfusion volume was stratified. The incidence of POD was significantly higher in the high-volume transfusion group compared with the no transfusion group (15.4% vs. 4.3%; adjusted odds ratio [OR] = 5.09, 95% confidence interval [CI]: 1.24–20.81; P = 0.024). Age ​was identified as​ a significant effect modifier (interaction P = 0.049), with patients aged 75 years or older exhibiting a heightened transfusion-related risk of POD.ConclusionsIn this retrospective study of older patients undergoing THA with CSEA, intraoperative high-volume transfusion was associated with an increased risk of POD compared with no transfusion.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12871-025-03554-5.Keywords: Blood transfusion, Postoperative delirium, Total hip arthroplasty Introduction With the accelerating demographic transition of population aging, the demand for surgical procedures among elderly patients has increased considerably. Total hip arthroplasty (THA) has become the standard surgical approach for treating severe hip joint disorders [1]. For elderly patients, hip pathology not only significantly impair quality of life but may also lead to increased functional dependence and mortality. THA provides an effective treatment approach for improving mobility and quality of life in elderly patients through reconstruction of joint function and alleviating pain [2]. However, elderly patients are at increased risk of perioperative complications due to reduced physiological capacity and increased comorbidity burden [3, 4]. Postoperative delirium (POD), a common syndrome of acute neurocognitive dysfunction in elderly patients following surgery, occurs at rates ranging from 4% to 53% in orthopedic procedures. It not only prolongs hospital stays and increases healthcare costs but is also closely associated with increased short-term and long-term mortality [5, 6]. Research has shown that POD may contribute to permanent cognitive impairment, significantly affecting patients’ long-term prognosis [7]. In patients undergoing THA, the combined effects of surgical trauma, pain, blood loss, and anesthetics result in a higher incidence of POD compared with other surgical procedures [8]. Combined spinal-epidural anesthesia (CSEA) is commonly used for hip arthroplasty. This technique provides stable hemodynamics, superior postoperative analgesia, and facilitates early mobilization. Some studies suggest that regional anesthesia reduces POD rates in elderly hip arthroplasty patients. This protective effect may result from avoiding direct central nervous system effects of general anesthetics [9]. However, recent studies report no significant increase in POD risk with general anesthesia. The impact of different anesthetic techniques on delirium remains controversial [10]. These findings suggest that elderly patients remain at POD risk even with regional anesthesia. Other perioperative factors may play important roles in delirium development. The association between perioperative blood transfusion and POD has long been an area of active investigation in clinical research. Existing studies show heterogeneous findings. Some indicate that intraoperative transfusion increases POD risk [11, 12], while others show no association or even protective effects in specific patient subgroups [13, 14]. Despite extensive research investigating the relationship between perioperative blood transfusion and POD, significant gaps in understanding and methodological limitations remain in this field. First, the dose-response relationship between transfusion volume and POD remains unclear. Most studies treat transfusion as a binary variable, failing to explore the dose-response effects of varying transfusion volumes on delirium risk. Additionally, adjustment for important confounding variables remains insufficient. Few studies have examined how different anesthetic approach affect the relationship between transfusion and delirium. This lack of stratified analysis may introduce bias in effect estimates. Finally, traditional statistical methods have limited ability in addressing selection bias, which may compromise the reliability of causal inference. This study aims to investigate the association between intraoperative transfusion volume and POD occurrence in elderly patients undergoing THA with CSEA. This study utilizes IPTW to balance baseline characteristics between groups and mitigate selection bias. We aim to provide evidence for developing precise perioperative blood management strategies. Methods Study design and participants This single-center retrospective cohort study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Data were retrospectively extracted from the electronic medical record system, surgical records, and postoperative follow-up files for patients aged ≥ 65 years who underwent THA at Taihe Hospital, Shiyan City, between January 2020 and April 2025. The study protocol was reviewed and approved by the Ethics Committee of Shiyan Taihe Hospital (Approval No. WZ2025KS101). Patients were excluded according to predefined exclusion criteria: (1) nine patients with a history of hip replacement/repair; (2) 30 patients with pre-existing cognitive impairment or psychiatric disorders, (3) nine patients with severe malignancies or end-stage renal disease, (4) 16 patients who received general anesthesia (i.e., those who were intubated or managed with a laryngeal mask airway). An additional 14 patients were excluded due to incomplete primary study data. Finally, 610 patients were included in the final analysis. The participant selection process is illustrated in Fig. 1. Flowchart of patient selection Exposure Patients were categorized into two groups according to intraoperative allogeneic red blood cell transfusion status: no transfused versus transfused cohorts. High-volume transfusion was defined as four or more units of packed red blood cells, consistent with the volume-dependent risk stratification for adverse outcomes employed by Shorr et al. in a large cohort of critically ill patients [15]. Consequently, the transfused cohort was further subdivided into low-volume (< 4 units) and high‐volume (≥ 4 units) subgroups. Intraoperative administration of allogeneic red blood cell (RBC) transfusions in elderly patients undergoing total hip arthroplasty (THA) was guided by established hemoglobin (Hb) thresholds combined with individualized clinical judgment. The general transfusion principles were as follows: when Hb was greater than 100 g/L, transfusion was generally not indicated; when Hb was less than 70 g/L, transfusion was routinely considered to prevent tissue hypoxia. For Hb levels between 70 and 100 g/L, transfusion decisions were individualized based on patient-specific factors, including age, estimated intraoperative or postoperative blood loss, cardiopulmonary reserve, and clinical manifestations suggestive of anemia or inadequate oxygen delivery (such as dizziness, fatigue, or hemodynamic instability). Outcome variables The primary outcome, postoperative delirium (POD), was determined through standardized assessments supplemented with a comprehensive review of clinical records. Trained evaluators confirmed POD occurrence within seven postoperative days through systematic chart review. This included progress notes, nursing evaluations, and consultation reports. The review process identified delirium-associated terminology including “delirium”, “confusion”, “lethargy”, “inattention”, “disorientation”, “hallucinations”, and “agitation”. Secondary outcomes included deep vein thrombosis (DVT), diagnosed by imaging examinations, and surgical site infection (SSI), defined according to Centers for Disease Control and Prevention (CDC) criteria. SSI diagnosis required documentation of at least one criterion: purulent drainage from the incision, pathogen isolation from cultures, or clinical infection indicators (pain, tenderness, swelling, erythema, or warmth) warranting treatment. Length of stay (LOS) was calculated as the total duration of hospitalization from admission to discharge. Covariates Collected data included patient demographics, comorbidities, laboratory parameters, and intraoperative variables. Baseline characteristics included age, sex, Body ​Mass ​Index (BMI), smoking status, alcohol consumption, and American ​Society of ​Anesthesiologists (ASA) physical status classification. Owing to limited cases in ASA classes I and IV, patients were categorized into ASA I–II and ASA III–IV groups. Comorbidities included hypertension, diabetes mellitus, cardiovascular disease, and cerebrovascular disease. Preoperative laboratory parameters included platelet count, Hct, and white blood cell count. Intraoperative variables included surgical duration. Statistical analysis All data management and statistical analyses were conducted using R software (version 4.5.1). We applied propensity score-based IPTW to adjust for confounding bias. This method balanced baseline differences between transfusion groups and the control group. Propensity scores were estimated using logistic regression models, with transfusion status as the dependent variable and all baseline covariates as independent variables. IPTW weights targeting the average treatment effect were generated using the WeightIt package. As IPTW assigns differential weights to individuals based on their estimated propensity scores, the resulting weighted pseudo-population may have an effective sample size that differs from the original sample, representing the amount of information retained after weighting. Baseline characteristics were compared before and after weighting with the tableone and survey packages, respectively. Categorical variables are presented as frequencies and percentages; median [interquartile range, IQR] are reported for non-normally distributed continuous variable. Group comparisons were conducted using chi-square tests for categorical variables and the Kruskal-Wallis test for continuous variables. Covariate balance was assessed by standardized mean differences (SMD), with an SMD less than 0.1 indicating adequate balance. Outcome analyses were conducted using the weighted sample through implementation of the survey package to estimate weighted generalized linear models (svyglm). For binary outcomes (POD, DVT, SSI), weighted logistic regression analyses were employed, with results expressed as odds ratios (ORs) and their corresponding 95% confidence intervals (CIs). Length of stay (LOS) was evaluated using weighted linear regression models, with findings presented as regression coefficients (β) and their respective 95% CIs. Statistical significance was established at a two-sided alpha level of 0.05. Results Baseline characteristics and balance A total of 610 patients were enrolled in the study including 190 patients in the transfusion group and 420 in the control group. Demographic and clinical characteristics of the participants are summarized in Table 1. Prior to propensity score weighting, patients in the transfusion group were older and had a higher proportion of ASA physical status classification III–IV. The prevalence of hypertension and diabetes mellitus was similar between groups, whereas cerebrovascular disease was more prevalent in the transfusion group. Additionally, the transfusion group exhibited greater intraoperative blood loss and longer operative duration. After adjustment using propensity score–based inverse probability of treatment weighting, baseline characteristics were well balanced between groups (SMD<0.1). Table 1Patient demographics before and after IPTWVariablesEntire cohortIPTW cohortNo transfusion (n = 420)Transfusion (n = 190)SMDNo transfusion (n = 623)Transfusion (n = 592)SMDPatient characteristics Age, years71(68, 76)73(69, 80)0.28671(68, 78)72(68, 78)0.020 Sex, female, %178(42.4)79(41.6)0.016265(42.6)232(39.3)0.068 BMI, kg/m222.9(20.8, 25.4)22.7(20.2, 24.8)0.14122.7(20.7, 25.0)22.9(20.3, 24.8)0.040 ASA, Ⅲ-Ⅳ, %94(22.4)57(30.0)0.174166(26.6)139(23.5)0.072 Smoking, %99(23.6)42(22.1)0.035137(22.0)132(22.3)0.006 Alcohol, %117(27.9)41(21.6)0.146157(25.3)143(24.1)0.026Comorbidities (%) Hypertension170(40.5)78(41.1)0.012255(40.9)242(40.8)0.002 Diabetes58(13.8)27(14.2)0.01287(14.0)82(13.9)0.002 COPD34(8.1)21(11.1)0.10151(8.2)50(8.4)0.006 Coronary artery disease88(21.0)44(23.2)0.053126(20.2)125(21.1)0.022 Cerebrovascular diseases57(13.6)41(21.6)0.21298(15.7)92(15.5)0.008Laboratory data WBC, 10⁹/L6.17(5.09, 7.84)6.26(5.10, 8.01)0.0556.22(5.09, 8.16)6.22(4.84, 7.78)0.045 PLT, 10⁹/L203(161, 245)209(157, 256)0.028201(159, 242)207(149, 251)0.036 Hct, %0.37(0.34, 0.40)0.37(0.33, 0.40)0.2040.37(0.33, 0.40)0.37(0.33, 0.39)0.061 Blood loss, ml200(120, 250)300(150, 400)0.533200(130, 300)200(129, 300)0.044Operative factors Operative time1.8(1.5, 2.2)2.1(1.8, 2.5)0.6441.9(1.6, 2.3)1.9(1.6, 2.3)0.055 Patient demographics before and after IPTW Clinical outcomes associated with intraoperative transfusion The overall incidence of postoperative delirium (POD) within seven days was 5.2% (32/610), with rates of 7.4% in the transfusion group and 4.3% in the non-transfusion group. After adjustment using inverse probability of treatment weighting (IPTW), overall intraoperative transfusion was not significantly associated with POD (adjusted odds ratio [OR] = 0.94, 95% confidence interval [CI]: 0.41–2.16; P = 0.886). Secondary outcomes, including deep vein thrombosis (DVT), surgical site infection (SSI), and length of stay (LOS), showed no statistically significant differences between groups (Table 2). Table 2Risk of outcomes by intraoperative blood transfusion volume: before and after IPTW adjustmentNo Transfusion(n=420)Transfusion(n=190)Unadjusted Estimate/OR [95% CI]P-valueIPTW adjusted Estimate/OR [95% CI]P-valuePrimary outcome POD18(4.3)14(7.4)1.77(0.86–3.65.86.65)0.1180.94(0.41–2.16.41.16)0.886Secondary outcome DVT11(2.6)8(4.2)1.63(0.65–4.13.65.13)0.2990.45(0.12–1.68.12.68)0.232 SSI10(2.4)3(1.6)0.66(0.18–2.41.18.41)0.5280.50(0.13–2.00.13.00)0.325 LOS15(10–18)16(12–19)1.03(0.08–1.98.08.98)0.0350.32(−1.25-1.90)0.690*Estimate denotes the mean difference (for continuous outcomes) or odds ratio (for categorical outcomes), with 95% confidence intervals shown in parenthesesP-values are calculated from unadjusted/adjusted models as specified Risk of outcomes by intraoperative blood transfusion volume: before and after IPTW adjustment *Estimate denotes the mean difference (for continuous outcomes) or odds ratio (for categorical outcomes), with 95% confidence intervals shown in parentheses P-values are calculated from unadjusted/adjusted models as specified Dose-Response relationship between transfusion volume and postoperative delirium To investigate potential dose-dependent effects, patients were further stratified according to transfusion volume: 138 patients (22.6%) received low-volume transfusion, and 52 patients (8.5%) received high-volume transfusion. ​The estimated blood loss was comparable between the low- and high-volume transfusion groups (median [IQR]: 200 mL [150, 300] vs. 400 mL [200, 570], respectively; P < 0.001). The incidence of POD was 4.3% (6/138) in the low-volume transfusion group and 15.4% (8/52) in the high-volume transfusion group. Following IPTW adjustment and direct comparison with the non-transfusion group, low-volume transfusion was not associated with an increased risk of POD (adjusted OR = 0.64, 95% CI: 0.23–1.80; P = 0.396). In contrast, high-volume transfusion was associated with an increased risk of POD (adjusted OR = 5.09, 95% CI: 1.24–20.81; P = 0.024) (Table 3). Secondary outcomes, including DVT, SSI, and LOS, showed no statistically significant differences between either transfusion subgroup and the non-transfusion group. Table 3Risk of postoperative delirium and secondary outcomes by intraoperative blood transfusion volume: before and after inverse probability of treatment weighting (IPTW) adjustmentNo Transfusion (n = 420)Transfusion (n = 190)Unadjusted Estimate/OR [95% CI]P-valueIPTW adjusted Estimate/OR [95% CI]P-valuePrimary outcome POD18(4.3)14(7.4)1.77(0.86–3.65)0.1180.94(0.41–2.16)0.886Secondary outcome DVT11(2.6)8(4.2)1.63(0.65–4.13)0.2990.45(0.12–1.68)0.232 SSI10(2.4)3(1.6)0.66(0.18–2.41)0.5280.50(0.13–2.00.13.00)0.325 LOS15(10–18)16(12–19)1.03(0.08–1.98)0.0350.32(−1.25−1.90)0.690*Estimate denotes the mean difference (for continuous outcomes) or odds ratio (for categorical outcomes), with 95% confidence intervals shown in parenthesesP-values are calculated from unadjusted/adjusted models as specified Risk of postoperative delirium and secondary outcomes by intraoperative blood transfusion volume: before and after inverse probability of treatment weighting (IPTW) adjustment *Estimate denotes the mean difference (for continuous outcomes) or odds ratio (for categorical outcomes), with 95% confidence intervals shown in parentheses P-values are calculated from unadjusted/adjusted models as specified Subgroup analysis Subgroup analyses stratified by anemia status, sex, and age were conducted using IPTW-adjusted multivariable models. Interaction analysis identified age as a significant effect modifier (interaction P = 0.049). Among patients aged < 75 years, low-volume transfusion demonstrated a potential protective effect compared to no transfusion (OR = 0.16). Conversely, among patients aged ≥ 75 years, both low-volume and high-volume transfusion were associated with an increased risk of POD (OR = 1.87 and 2.38, respectively). No significant interactions were detected for other subgroup variables as shown in Fig. 2. Association between the transfusion volume and POD in subgroups Discussion Postoperative delirium (POD) represents a serious complication following total hip arthroplasty (THA) in elderly patients. This study utilized inverse probability of treatment weighting (IPTW) to evaluate the dose-response relationship between intraoperative transfusion volume and POD risk in this patient population. Findings indicate that high-volume transfusion (≥ 4 units) may be associated with an increased risk of POD, whereas low-volume transfusion did not demonstrate a significant increase in risk. These results provide preliminary data to inform intraoperative transfusion practices and early POD surveillance among elderly THA patients, although validation in large-scale, multicenter studies is required. Previous investigations have yielded conflicting results regarding the association between intraoperative transfusion and POD. Chou et al. [12] reported that elderly orthopedic patients with preoperative anemia who received intraoperative transfusion had the highest POD risk. In contrast, Fan et al. found no significant differences in POD incidence between restrictive and liberal transfusion strategies, while Kim et al. observed no clear association between transfusion and POD following knee or hip arthroplasty [13, 16]. These inconsistencies likely reflect variability in study design, patient characteristics, anesthetic approaches, and transfusion protocols. To address this variability, we categorized transfusion by volume and used IPTW to reduce selection bias. Our findings suggest that, compared with the non-transfusion group, high-volume transfusion is associated with an increased risk of postoperative delirium (POD) in elderly patients, whereas low-volume transfusion shows no such association. The pathophysiological mechanisms underlying the association between high-volume transfusion and POD are likely multifactorial. Allogeneic transfusion has been shown to activate immune pathways, with proinflammatory mediators potentially compromising central nervous system function through blood–brain barrier disruption [17]. Transfusion-related immunomodulation (TRIM) may further contribute to neuroinflammatory responses [18]. Additionally, circulatory volume fluctuations and hemorheological alterations associated with massive transfusion may compromise cerebral microcirculatory perfusion and enhance oxidative stress [19]. Biochemical alterations in stored erythrocytes may release free hemoglobin and reactive oxygen species, potentially exacerbating endothelial and neuronal injury. These mechanisms may demonstrate interactive effects at higher transfusion volumes, thus contributing to increased delirium susceptibility [20]. Notably, perioperative blood loss was significantly higher in the high-volume transfusion group in our cohort. Massive bleeding and prolonged operative time, which typically accompany high transfusion volumes, exacerbate systemic inflammatory cascade reactions and stress responses, thereby increasing the risk of POD. Currently, there is no universally accepted, standardized definition for “massive transfusion” or high-volume transfusion in non-trauma elective surgery. The trauma literature predominantly defines massive transfusion as ≥ 10 units of red blood cells (RBCs) per 24 h [21]. However, this threshold reflects a substantially different clinical context and risk profile than that encountered in elective surgical populations. To address this discrepancy, we adapted the stratification framework established by Shorr and colleagues, which was derived from a large cohort of critically ill patients [15]. This framework risk-stratifies adverse outcomes based on transfusion volume. We therefore set our operational threshold at ≥ 4 units of RBCs. This lower threshold enables more sensitive detection of clinically significant transfusion-related risks specific to the elective surgery setting. Future studies should explore and validate transfusion volume thresholds associated with postoperative delirium across diverse surgical populations. Current research shows that regional anesthesia does not increase the incidence of postoperative delirium compared to general anesthesia [10, 22]. In our study center, CSEA was the primary anesthetic technique used for THA, with general anesthesia reserved exclusively for patients with contraindications to neuraxial techniques. Consequently, general anesthesia recipients were excluded from the analysis due to their limited representation. This methodological approach helped standardize the anesthetic technique and minimize potential confounding effects, allowing for a clearer assessment of the association between transfusion and delirium. To further attenuate residual confounding attributable to baseline imbalances, inverse probability of treatment weighting (IPTW) was subsequently implemented. However, our findings may have limited generalizability to patients who undergo general anesthesia. This study has several inherent limitations. First, its single-center, retrospective nature may limit the applicability of the findings to other settings and precludes definitive conclusions regarding causality. Second, residual confounding is likely, as our propensity score model did not include several potential confounders (e.g., preoperative hemoglobin levels, intraoperative hypotension, vasopressor use, fluid balance, and sedative/opioid doses). The omission of these variables represents a major limitation, as they could account for the observed relationship between transfusion and delirium. Thirdly, POD was identified based on clinical documentation rather than standardized tools such as the Confusion Assessment Method (CAM) or CAM-ICU. This reliance on unstructured clinical records may have led to underdetection or misclassification bias, ​which could also explain the relatively low overall incidence of POD observed in our cohort compared to studies using prospective screening, potentially attenuating the observed associations. The association between high-volume transfusion and POD was derived from a relatively small patient subgroup, limiting the precision of the effect estimate. The wide confidence interval and the low number of events highlight the potential influence of random error or sparse data bias. The observed dose–response pattern appears clinically plausible; however, these findings should be interpreted with caution and confirmed in larger, multicenter cohorts. Despite these limitations, the current findings provide preliminary evidence suggestive of a dose-response relationship between transfusion volume and POD risk. Conclusion In this retrospective study of older patients undergoing total hip arthroplasty (THA) with combined spinal-epidural anesthesia (CSEA), intraoperative high-volume transfusion was associated with an increased risk of postoperative delirium (POD) compared with patients who did not receive transfusion, whereas low-volume transfusion showed no significant association with POD risk. These findings indicate that high-volume transfusion (≥ 4 units) may be an independent predictor of postoperative delirium. Nevertheless, the underlying mechanisms remain to be clarified, and large-scale prospective studies are required to further validate these findings and provide evidence for the development of targeted preventive strategies. Supplementary Information Supplementary Material 1 Acknowledgements Not applicable. Abbreviations postoperative delirium inverse probability of treatment weighting odds ratio confidence interval total hip arthroplasty combined spinal-epidural anesthesia deep vein thrombosis surgical site infection Centers for Disease Control and Prevention length of stay body ​mass ​index American ​Society of ​Anesthesiologists standardized mean differences Authors’ contributions Yi Liu: Writing original draft; Kun Wang: Data curation and analysis; Kun Wang and Juan Pen: Data collection; Qing Li: Review and editing; Xianting Ke: Supervision. All authors have read and agreed to the published version of the manuscript. Funding Not applicable. Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Ethical approval of the study was obtained from Shiyan Taihe Hospital ethics committees (Ethics No.: [WZ2025KS101]), and this study adhered to the Declaration of Helsinki. The Institutional Review Board of the Shiyan Taihe Hospital granted an informed consent waiver. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yi Liu and Kun Wang contributed equally to this work. Contributor Information Qing Li, Email: liqing8801@163.com. Xianting Ke, Email: xiantingke@126.com. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.