Predicting red blood cell transfusion in hospitalized patients: role of hemoglobin level, comorbidities, and illness severity

Leading digit bias in hemoglobin thresholds for red cell transfusion

BACKGROUND

Leading digit bias is a heuristic whereby humans overemphasize the left-most digit when evaluating numbers (e.g., 9.99 vs. 10.00). The bias might affect the interpretation of hemoglobin results and influence red cell transfusion in hospitalized patients.

STUDY DESIGN AND METHODS

Adults who received a red cell transfusion while registered at the University Health Network (Toronto, Canada) between January 1, 2016 and January 1, 2022 (n = 6 years) were included. The primary analysis excluded apheresis, red cell disorders, radiology suites, and operating rooms. The primary comparison was a regression discontinuity analysis of transfusion occurrence above and below the hemoglobin threshold of 79 g/L (local units). Additional analyses tested other leading digit and control thresholds (71, 81, and 91 g/L). Secondary analyses explored temporal covariates and clinical subgroups.

RESULTS

A total of 211,872 red cell transfusions were identified over the study period (median pre-transfusion hemoglobin 76 g/L; interquartile range = 69-92 g/L), with 107,790 inpatient transfusions in the primary analysis. The 79 g/L threshold showed 815 fewer red cell units above the threshold (95% confidence interval [CI]: -1215 to -415). The 69 g/L threshold showed 2813 fewer transfused units (95% CI: -4407 to -1220), and 89 g/L showed 40 fewer units (95% CI: -408 to 328). The effect was accentuated during daytime, weekday, and May-June months, persisted in analyses including all transfusions, and was absent at control thresholds.

CONCLUSION

Leading digit bias might have a modest influence on the decision to transfuse red cells. The findings may inform practice guidelines and quasi-experimental study design in transfusion research.

Development and validation of a nomogram for blood transfusion during intracranial aneurysm clamping surgery: a retrospective analysis

PURPOSE

Intraoperative blood transfusion is associated with adverse events. We aimed to establish a machine learning model to predict the probability of intraoperative blood transfusion during intracranial aneurysm surgery.

METHODS

Patients, who underwent intracranial aneurysm surgery in our hospital between January 2019 and December 2021 were enrolled. Four machine learning models were benchmarked and the best learning model was used to establish the nomogram, before conducting a discriminative assessment.

RESULTS

A total of 375 patients were included for analysis in this model, among whom 108 received an intraoperative blood transfusion during the intracranial aneurysm surgery. The least absolute shrinkage selection operator identified six preoperative relative factors: hemoglobin, platelet, D-dimer, sex, white blood cell, and aneurysm rupture before surgery. Performance evaluation of the classification error demonstrated the following: K-nearest neighbor, 0.2903; logistic regression, 0.2290; ranger, 0.2518; and extremely gradient boosting model, 0.2632. A nomogram based on a logistic regression algorithm was established using the above six parameters. The AUC values of the nomogram were 0.828 (0.775, 0.881) and 0.796 (0.710, 0.882) in the development and validation groups, respectively.

CONCLUSIONS

Machine learning algorithms present a good performance evaluation of intraoperative blood transfusion. The nomogram established using a logistic regression algorithm showed a good discriminative ability to predict intraoperative blood transfusion during aneurysm surgery.

Multivariable statistical models to predict red cell transfusion in elective surgery

BACKGROUND

Predicting red cell transfusion may assist in identifying those most likely to benefit from patient blood management strategies. Our objective was to identify a simple statistical model to predict transfusion in elective surgery from routinely available data.

MATERIALS AND METHODS

Our final multicentre cohort consisted of 42,546 patients and contained the following potential predictors of red cell transfusion known prior to admission: patient age, sex, pre-admission hemoglobin, surgical procedure, and comorbidities. Missing data were handled by multiple imputation methods. The outcome measure of interest was administration of a red cell transfusion. We used multivariable logistic regression models to predict transfusion, and evaluated the performance by applying a 10-fold cross-validation. Model accuracy was assessed by comparing the area under the receiver operating characteristics curve. After applying an optimal probability cut-off we measured model accuracy, sensitivity, specificity, positive predictive value, and negative predictive value.

RESULTS

7.0% (n=2,993) of the study population received a red cell transfusion. Our most simple model predicted red cell transfusion based on admission hemoglobin and surgical procedure with a multiply imputed estimated area under the curve of 0.862 (0.856, 0.864). The estimated accuracy, sensitivity, specificity, positive predictive, and negative predictive values at the probability cut-off of 0.4 were 0.934, 0.257, 0.986, 0.573, and 0.946 respectively.

DISCUSSION

A small number of variables available prior to admission can predict red cell transfusion with very good accuracy. Our model can be used to flag high-risk patients most likely to benefit from pre-operative patient blood management measures.

A novel model forecasting perioperative red blood cell transfusion

We aimed to establish a predictive model assessing perioperative blood transfusion risk using a nomogram. Clinical data for 97,443 surgery patients were abstracted from the DATADRYAD website; approximately 75% of these patients were enrolled in the derivation cohort, while approximately 25% were enrolled in the validation cohort. Multivariate logical regression was used to identify predictive factors for transfusion. Receiver operating characteristic (ROC) curves, calibration plots, and decision curves were used to assess the model performance. In total, 5888 patients received > 1 unit of red blood cells; the total transfusion rate was 6.04%. Eight variables including age, race, American Society of Anesthesiologists' Physical Status Classification (ASA-PS), grade of kidney disease, type of anaesthesia, priority of surgery, surgery risk, and an 18-level variable were included. The nomogram achieved good concordance indices of 0.870 and 0.865 in the derivation and validation cohorts, respectively. The Youden index identified an optimal cut-off predicted probability of 0.163 with a sensitivity of 0.821 and a specificity of 0.744. Decision curve (DCA) showed patients had a standardized net benefit in the range of a 5–60% likelihood of transfusion risk. In conclusion, a nomogram model was established to be used for risk stratification of patients undergoing surgery at risk for blood transfusion. The URLs of web calculators for our model are as follows: http://www.empowerstats.net/pmodel/?m=11633_transfusionpreiction.

Hospital-Acquired Infection, Length of Stay, and Readmission in Elective Surgery Patients Transfused 1 Unit of Red Blood Cells: A Retrospective Cohort Study

BACKGROUND

Most patients transfused red blood cells in elective surgery receive small volumes of blood, which is likely to be discretionary and avoidable. We investigated the outcomes of patients who received a single unit of packed red blood cells during their hospital admission for an elective surgical procedure when compared to those not transfused.

METHODS

This retrospective cohort study included elective surgical admissions to 4 hospitals in Western Australia over a 6-year period. Participants were included if they were at least 18 years of age and were admitted for elective surgery between July 2014 and June 2020. We compared outcomes of patients who had received 1 unit of red blood cells to patients who had not been transfused. To balance differences in patient characteristics, we weighted our multivariable regression models using the inverse probability of treatment. In addition to propensity score weighting, our multivariable regression models adjusted for hemoglobin level, surgical procedure, patient age, gender, comorbidities, and the transfusion of fresh-frozen plasma or platelets. Outcomes studied were hospital-acquired infection, hospital length of stay, and all-cause emergency readmissions within 28 days.

RESULTS

Overall, 767 (3.2%) patients received a transfusion of 1 unit of red blood cells throughout their admission. In the propensity score weighted analysis, the transfusion of a single unit of red blood cells was associated with higher odds of hospital-acquired infection (odds ratio, 3.94; 95% confidence interval [CI], 2.99-5.20; P < .001). Patients who received 1 unit of red blood cells throughout their admission were more likely to have a longer hospital stay (rate ratio, 1.57; 95% CI, 1.51-1.63; P < .001) and had 1.42 (95% CI, 1.20-1.69; P < .001) times higher odds of 28-day readmission.

CONCLUSIONS

These results suggest that avoidance of even small volumes of packed red blood cells may prevent adverse clinical outcomes. This may encourage hospital administrators to implement strategies to avoid the transfusion of even small volumes of red blood cells by applying patient blood management practices.

Preoperative anemia is a risk factor for poor perioperative outcomes in ventral hernia repair

PURPOSE

Ventral hernia repairs (VHR) are among the most commonly performed operations by general surgeons. Despite advances in technology there remains high complication and readmission rates. Preoperative anemia has been linked to poor outcomes and readmission across several surgical procedures, however the link to ventral hernia repair outcomes is limited.

METHODS

Utilizing the American College of Surgeons National Safety and Quality Improvement Project (NSQIP) database for years 2016-2018, a total of 115,000 patients met inclusion criteria. Using propensity matching we matched two groups of patients who underwent VHR: (1) those with preoperative anemia and (2) those with normal hemoglobin levels. Anemia criteria was set forth by the World Health Organization (WHO).

RESULTS

Univariate analysis did demonstrate statistical significance in post-operative outcomes percentage of serious surgical site infection, poor renal outcomes, transfusion, and unplanned remission in those with preoperative anemia who underwent VHR. In a multivariate analysis, patients who underwent ventral hernia repair with pre-operative anemia had significantly greater odds of unplanned readmission (odds ratio 1.35, 95% confidence interval 1.16-1.57) and serious surgical site infection (odds ratio 1.35, 95% confidence interval 1.04-1.74) independent of known risk factors such as smoking, diabetes and obesity.

CONCLUSIONS

Preoperative anemia is a risk factor for poor postoperative outcomes in those undergoing ventral hernia repair and should be considered when evaluating a patient for repair.

Development and Validation of a Nomogram to Predict the Risk of Blood Transfusion in Orthognathic Patients

OBJECTIVE

This study aims to establish a nomogram to predict the probability of blood transfusion in patients with preoperative autologous blood donation before orthognathic surgery.

METHODS

The authors conducted a retrospective case-control study on consecutive orthognathic patients with preoperative autologous blood donation from January 2014 to December 2020. The outcome variable was the actual transfusion of autologous blood (ATAB). Predictors included patients' demographics, preoperative blood cell test, vital signs, American Society of Anesthesiologists classification, surgical procedure, operation duration, and blood loss. Univariable and multivariable logistic regressions were performed to identify independent risk factors associated with ATAB. A nomogram was constructed to predict the risk for ATAB. The performance of the nomogram was evaluated using the area under the receiver operating characteristic curve, calibration curve and the consistency index.

RESULTS

A total of 142 patients (75 males and 67 females) with an average age of 22.72 ± 5.34 years donated autologous blood before their orthognathic surgery. Patients in the transfusion group (n = 56) had significantly lower preoperative red blood cell counts (4.74 ± 0.55 × 109/L versus 4.98 ± 0.45 × 109/L, P = 0.0063), hemoglobin (141.48 ± 15.18 g/dL versus 150.33 ± 14.73 g/dL, P = 0.0008), and hematocrit (41.05% ± 4.03% versus 43.32% ± 3.42%, P = 0.0006), more bimaxillary osteotomies (92.86% versus 56.98%, P < 0.001), longer operation duration (348.4 ± 111.10 minutes versus 261.6 ± 115.44 minutes, P < 0.001), and more intraoperative blood loss (629.23 ± 273.06 ml versus 359.53 ± 222.84 ml, P < 0.001) than their counterparts (n = 86) in the non-transfusion group. Univariable and multivariable logistic regression demonstrated that only hemoglobin (adjusted odds ratio [OR] 0.864, 95% confidence interval [CI]:0.76-0.98, P = 0.026), operation procedures (adjusted OR 8.14, 95% CI:1.69-39.16, P = 0.009), and blood loss (adjusted OR 1.006, 95% CI:1.002-1.009, P < 0.001) were independent risk factors for ATAB. The area under the receiver operating characteristic curve of the nomogram was 0.823. The consistency index of the nomogram was 0.823. The calibration curve illustrated that the nomogram was highly consistent with the actual observation.

CONCLUSIONS

The nomogram is a simple and useful tool with good accuracy and performance in predicting the risk for blood transfusion.

Development and Validation of a Transfusion Risk Score for Patients Receiving Maintenance Hemodialysis

BACKGROUND

In patients on dialysis with anemia, avoiding red blood cell transfusions is preferable. We sought to develop and validate a novel transfusion prediction risk score for patients receiving maintenance hemodialysis.

METHODS

This retrospective cohort study used United States Renal Data System data to create a model development cohort (patients who were point prevalent and on hemodialysis on November 1, 2012) and a validation cohort (patients who were point prevalent and on hemodialysis on August 1, 2013). We characterized comorbidity, inflammatory conditions, hospitalizations, anemia and anemia management, iron parameters, intravenous iron use, and vitamin D use during a 6-month baseline period to predict subsequent 3-month transfusion risk. We used logistic least absolute shrinkage and selection operator regression. In an exploratory analysis, model results were used to calculate a score to predict 6- and 12-month hospitalization and mortality.

RESULTS

Variables most predictive of transfusion were prior transfusion, hemoglobin, ferritin, and number of hospital days in the baseline period. The resulting c-statistic in the validation cohort was 0.74, indicating relatively good predictive power. The score was associated with a significantly increased risk of subsequent mortality (hazard ratios 1.0, 1.22, 1.26, 1.54, 1.71, grouped from lowest to highest score), but not with hospitalization.

CONCLUSIONS

We developed a transfusion prediction risk score with good performance characteristics that was associated with mortality. This score could be further developed into a clinically useful application, allowing clinicians to identify patients on hemodialysis most likely to benefit from a timely, proactive anemia treatment approach, with the goal of avoiding red blood cell transfusions and attendant risks of adverse clinical outcomes.

Blood transfusion trends by disease category in the United States, 2000 to 2014

BACKGROUND

Better understanding of blood usage rates could identify trends in transfusion practices over time and inform more efficient management.

METHODS

Inpatient admissions from the Healthcare Cost and Utilization Project National Inpatient Sample and State Inpatient Databases were analyzed for packed red blood cell (PRBC), plasma, platelet, and whole blood (WB) transfusions. The transfusion rates per admission and per prevalent case were calculated. Prevalence estimates were from the Global Burden of Disease 2017 study (GBD).

RESULTS

From 2000 to 2014, blood usage rates for most causes peaked around 2010. Across all causes, PRBC were the most commonly transfused component, followed by plasma, platelets, and WB. However, the relative use of each type varied by cause. Nutritional deficiencies (1.75 blood product units across all components per admission; 95 % uncertainty interval (UI) 1.62-1.87), neoplasms (0.95; 0.87-1.04), and injuries (0.92; 0.86 - 0.98) had the greatest blood use per admission. Cardiovascular diseases (96.9 units per 1000 prevalent cases; 89.3-105.0) and neoplasms (92.7 units per 1000 prevalent cases; 84.3-101.5) had the greatest blood use per prevalent case. Across all admissions, over three million blood units were saved in 2014 compared to 2011 due to transfusing at a reduced rate.

CONCLUSIONS

Blood transfusion rates decreased from 2011 to 2014 in the United States. This decline occurred in most disease categories, which points towards broad strategies like patient blood management systems and disease specific improvements like changes in surgical techniques being effective.

The Swedish Scandinavian donations and transfusions database (SCANDAT3-S) - 50 years of donor and recipient follow-up

Background

The two previous versions of the Scandinavian donations and transfusions (SCANDAT) databases, encompassing data on blood donors, blood components, transfusions, and transfused patients linked to national health registers in Sweden and Denmark up until 2012, have been used to study donor health, disease transmission, the role of donor characteristics, and more.

Study Design and Methods

Here we describe the creation of the Swedish portion of the third iteration of SCANDAT – SCANDAT3‐S – with follow‐up from 1968 to the end of 2017, resulting in up to 50 years of uninterrupted follow‐up for donors and recipients. The database now also includes non‐transfused non‐donors with a blood typing result, increased temporal resolution for transfusions, and linkages to laboratory and drug prescription data.

Results

After data cleaning, the database contained 23 579 863 donation records, 21 383 317 transfusion records, and 8 071 066 unique persons with valid identification. In total, the database offers 28 638 436 person‐years of follow‐up for donors, 13 582 350 person‐years of follow‐up for transfusion recipients, and 65 613 639 person‐years of follow‐up for non‐recipient non‐donors, with possibility for future extension. Additionally, the database includes 167 820 412 dispense records for prescribed drugs and 316,338,442 laboratory test results. Since the latest update in 2012, >99.9% of all donations were traceable to a donor with valid identification, and >97% of all transfusions to a recipient with valid identification.

Conclusion

With extended follow‐up and more clinical detail, the Swedish portion of the third and latest iteration of the SCANDAT database should allow for more comprehensive analysis of donation and transfusion‐related research questions.

Individual- and hospital-level correlates of red blood cell, platelet, and plasma transfusions among hospitalized children and neonates: a nationally representative study in the United States

BACKGROUND

Factors associated with red blood cell (RBC), plasma, and platelet transfusions in hospitalized neonates and children across the United States have not been well characterized.

METHODS

Data from the Kids' Inpatient Database (KID) 2016 were analyzed. KID is a random sample of 10% of all uncomplicated in-hospital births and 80% of remaining pediatric discharges from approximately 4200 US hospitals. Sampling weights were applied to generate nationally representative estimates. Primary outcome was one or more RBC transfusion procedures; plasma and platelet transfusions were assessed as secondary outcomes. Analysis was stratified by age: neonates (NEO; ≤28 d), and nonneonates (PED; >28 d and <18 y). Multivariable logistic regression was used to estimate adjusted odds ratios (aORs) and 95% confidence intervals (95% CIs).

RESULTS

Among 5,604,984 total hospitalizations, overall prevalence of transfusions was 1.07% (95% CI, 0.94%-1.22%) for RBCs, 0.17% (95% CIs, 0.15%-0.21%) for plasma and 0.35% (95% CI, 0.30%-0.40%) for platelet transfusions. RBC transfusions occurred among 0.43% NEO admissions and 2.63% PED admissions. For NEO admissions, RBC transfusion was positively associated with nonwhite race, longer length of hospitalization, highest risk of mortality (aOR, 86.58; 95% CI, 64.77-115.73) and urban teaching hospital location. In addition to the above factors, among PED admissions, RBC transfusion was positively associated with older age, female sex (aOR, 1.10; 95% CI, 1.07-1.13), and elective admission status (aOR, 1.62; 95% CI, 1.46-1.80). Factors associated with plasma and platelet transfusions were largely similar to those associated with RBC transfusion, except older age groups had lower odds of plasma transfusion among PED admissions.

CONCLUSIONS

While there is substantial variability in the proportion of neonates and nonneonatal children transfused nationally, there are several similar, yet unique, nonlaboratory predictors of transfusion identified in these age groups.

Relationship between comorbidities and treatment decision-making in elderly hip fracture patients

Background

Elderly patients are at a higher risk for hip fracture. Moreover, hospitalized elderly patients with hip fracture are vulnerable to adverse outcomes including higher mortality rate and long-term disability. Treatment decision-making with respect to surgical procedure and perioperative management of these patients is typically challenging owing to the presence of multiple comorbid conditions.

Aims

The purpose of this study was to investigate the relationship between comorbidities in elderly patients with hip fracture and the treatment decision-making.

Methods

884 geriatric patients (age ≥ 60 years) with hip fracture were included. Comorbidities related to age were measured using the Charlson Co-morbidity Index (CCI) and age-adjusted CCI. The CCI of each geriatric hip fracture patient was calculated based on data retrieved from the medical records. The relationship of CCI and age-adjusted CCI with surgical procedure, time-to-surgery, length of hospital stay, and perioperative management (transfusion, anti-coagulation, and analgesia) was assessed.

Results

Mean age of patients was 78.01 ± 8.62 years. The mean CCI was 0.79 ± 0.036; the mean age-adjusted CCI was 4.15 ± 0.047. The CCI was significantly associated with time-to-surgery (P = 0.004), surgical treatment (P < 0.001), and transfusion (P = 0.023). The age-adjusted CCI was significantly associated with surgical treatment (P < 0.001), analgesia (P = 0.003) and transfusion (P < 0.001). The length of hospital stay was associated with both CCI (P = 0.041), age-adjusted CCI (P = 0.002), and hypertension (P = 0.012). Hospital expenses showed a significant association with CCI (P = 0.000), age-adjusted CCI (P = 0.029), osteoprosis (P = 0.007), and hypertension (P = 0.001).

Conclusion

In this study, comorbidities were positively associated with surgical procedure and perioperative management of elderly patients with hip fracture.

Electronic supplementary material

The online version of this article (10.1007/s40520-019-01134-5) contains supplementary material, which is available to authorized users.

Determinants of transfusion decisions in the ICU: haemoglobin concentration, what else? - a retrospective cohort study

BACKGROUND AND OBJECTIVES

The aim of this study was to assess potentially relevant clinical characteristics which influence the decision to transfuse red cells in critically ill patients with low haemoglobin concentrations (6.0-10.0 g/dl).

MATERIALS AND METHODS

This was a retrospective observational cohort study of patients admitted between November 2004 and May 2016 at the intensive care unit (ICU) of the Leiden University Medical Center, Netherlands. Haemoglobin measurements, clinical variables and the subsequent transfusion decision were extracted from the electronic health records. Clinical variables were grouped by organ system. We first examined the association of each of the clinical variables with the decision to transfuse during the following 6 h after a haemoglobin measurement using generalized estimating equations. We then compared the predictive abilities of single variables within an organ system and the predictive ability of an organ system's combined variables using the change in Akaike information criterion (AIC).

RESULTS

A total of 83 394 haemoglobin measurements of 10 947 ICU admissions were included. Haemoglobin concentration was the most predictive for red cell transfusion. After the haemoglobin concentration, the combined variables for General Health, followed by the organ systems Cardiovascular and Pulmonary, were most predictive for red cell transfusion. Within these organ systems, the APACHE II score, referring department, APACHE admission diagnosis subgroup, troponin concentration, lactate concentration, respiratory rate, PaO₂ /FiO₂ and ventilation mode had the highest predictive ability.

CONCLUSION

Haemoglobin concentration is the dominant predictor for red cell transfusion. Other clinical characteristics are also predictive, though to a lesser extent.

Effect of donor, component, and recipient characteristics on hemoglobin increments following red blood cell transfusion

Significant research has focused individually on blood donors, product preparation and storage, and optimal transfusion practice. To better understand the interplay between these factors on measures of red blood cell (RBC) transfusion efficacy, we conducted a linked analysis of blood donor and component data with patients who received single-unit RBC transfusions between 2008 and 2016. Hemoglobin levels before and after RBC transfusions and at 24- and 48-hour intervals after transfusion were analyzed. Generalized estimating equation linear regression models were fit to examine hemoglobin increments after RBC transfusion adjusting for donor and recipient demographic characteristics, collection method, additive solution, gamma irradiation, and storage duration. We linked data on 23 194 transfusion recipients who received one or more single-unit RBC transfusions (n = 38 019 units) to donor demographic and component characteristics. Donor and recipient sex, Rh-D status, collection method, gamma irradiation, recipient age and body mass index, and pretransfusion hemoglobin levels were significant predictors of hemoglobin increments in univariate and multivariable analyses (P < .01). For hemoglobin increments 24 hours after transfusion, the coefficient of determination for the generalized estimating equation models was 0.25, with an estimated correlation between actual and predicted values of 0.5. Collectively, blood donor demographic characteristics, collection and processing methods, and recipient characteristics accounted for significant variation in hemoglobin increments related to RBC transfusion. Multivariable modeling allows the prediction of changes in hemoglobin using donor-, component-, and patient-level characteristics. Accounting for these factors will be critical for future analyses of donor and component factors, including genetic polymorphisms, on posttransfusion increments and other patient outcomes.

Lezin EM, Erikstrup C, Ullum H, Glynn SA, Kleinman S, Hjalgrim H, Roubinian NH. Association of Blood Donor Sex and Prior Pregnancy With Mortality Among Red Blood Cell Transfusion Recipients

IMPORTANCE

Evidence regarding associations of blood donor sex with mortality among red blood cell transfusion recipients is conflicting.

OBJECTIVE

To study associations of donor sex and prior pregnancy with mortality of transfusion recipients.

DESIGN, SETTING, AND PARTICIPANTS

Data from 3 retrospective cohorts of transfusion recipients (the Kaiser Permanente Northern California [KPNC] and Recipient Epidemiology and Donor Evaluation Study-III [REDS-III] databases of data from January 2013 to December 2016 and the Scandinavian Donations and Transfusions [SCANDAT] database with data from January 2003 to December 2012) were analyzed. Final dates of follow-up were December 31, 2016, for the KPNC and REDS-III cohorts and December 31, 2012, for the SCANDAT cohort. Stratified Cox regression models were used to estimate associations between donor exposure groups with risk of mortality, adjusting for the number of red blood cell unit transfusions.

EXPOSURES

The number of transfused red blood cell units from female donors, previously pregnant donors, and sex-discordant donors (male donor and female recipient or female donor and male recipient).

MAIN OUTCOMES AND MEASURES

In-hospital mortality.

RESULTS

The study population included 34 662 patients (mean age, 69 years; 18 652 [54%] women) from the KPNC cohort, 93 724 patients (mean age, 61 years; 48 348 [52%] women) from the REDS-III cohort, and 918 996 patients (mean age, 72 years; 522 239 [57%] women) from the SCANDAT cohort. The median number of red blood cell transfusions per patient was 3 in the KPNC cohort, 2 in the REDS-III cohort, and 3 in the SCANDAT cohort. The percentage of transfusions from previously pregnant or parous donors was 9% in the KPNC cohort, 18% in the REDS-III cohort, and 25% in the SCANDAT cohort. The percentage of transfusions in the 3 cohorts from female donors ranged from 39% to 43%, from previously pregnant or parous donors ranged from 9% to 25%, and from sex-discordant donors ranged from 44% to 50%. There were 3217 in-hospital deaths in the KPNC cohort, 8519 in the REDS-III cohort, and 198 537 in the SCANDAT cohort. There were no statistically significant associations between any of the 3 donor exposures and in-hospital mortality in the 3 cohorts. Hazard ratios for in-hospital mortality per transfused unit from female donors were 0.99 (95% CI, 0.96-1.03) for the KPNC cohort, 1.00 (95% CI, 0.99-1.01) for the REDS-III cohort, and 1.00 (95% CI, 0.99-1.00) for the SCANDAT cohort. For units from previously pregnant or parous female donors, hazard ratios were 1.00 (95% CI, 1.00-1.01) for the KPNC cohort, 1.01 (95% CI, 0.98-1.03) for the REDS-III cohort, and 1.00 (95% CI, 1.00-1.01) for the SCANDAT cohort. For units from sex-discordant transfusions, hazard ratios were 1.02 (95% CI, 0.99-1.05) for the KPNC cohort, 0.99 (95% CI, 0.98-1.00) for the REDS-III cohort, and 1.00 (95% CI, 0.99-1.00) for the SCANDAT cohort.

CONCLUSIONS AND RELEVANCE

Among red blood cell transfusion recipients, transfusions from female, previously pregnant, or sex-discordant donors were not significantly associated with increased mortality.

Red blood cell transfusions for emergency department patients with gastrointestinal bleeding within an integrated health system

STUDY OBJECTIVE

To assess trends over time in red blood cell (RBC) transfusion practice among emergency department (ED) patients with gastrointestinal (GI) bleeding within an integrated healthcare system, inclusive of 21 EDs.

METHODS

Retrospective cohort of ED patients diagnosed with GI bleeding between July 1st, 2012 and September 30th, 2016. The primary outcome was receipt of an RBC transfusion in the ED. Secondary outcomes included 90-day rates of RBC transfusion, repeat ED visits, rehospitalization, and all-cause mortality. Logistic regression was used to obtain confounder-adjusted outcome rates.

RESULTS

A total of 24,868 unique patient encounters were used for the primary analysis. The median hemoglobin level in the ED prior to RBC transfusion decreased from 7.5 g/dl to 6.9 g/dl in the first versus last twelve months of the study period (p < 0.0001). A small trend was observed in the overall adjusted rate of ED RBC transfusion (absolute quarterly change of -0.1%, R2 = 0.18, p = 0.0001) largely attributable to the subgroup of patients with hemoglobin nadirs between 7.0 and 9.9 g/dl (absolute quarterly change of -0.4%, R2 = 0.38, p < 0.0001). Rates of RBC transfusions through 90 days likewise decreased (absolute quarterly change of -0.4%, R2 = 0.85, p < 0.0001) with stable to decreased corresponding rates of repeat ED visits, rehospitalizations and mortality.

CONCLUSION

Rates of ED RBC transfusion decreased over time among patients with GI bleeding, particularly in those with hemoglobin nadirs between 7.0 and 9.9 g/dl. These findings suggest that ED providers are willing to adopt evidence-based restrictive RBC transfusion recommendations for patients with GI bleeding.

Factors associated with red blood cell, platelet, and plasma transfusions among inpatient hospitalizations: a nationally representative study in the United States

BACKGROUND

Demographic and hospital-level factors associated with red blood cell (RBC), plasma, and platelet transfusions in hospitalized patients across the U.S. are not well characterized.

METHODS

We conducted a retrospective analysis of the National Inpatient Sample (2014). The unit of analysis was a hospitalization; sampling weights were applied to generate nationally-representative estimates. The primary outcome was having ≥ 1 RBC transfusion procedure; plasma and platelet transfusions were similarly assessed as secondary outcomes. For each component, factors associated with transfusion were measured using adjusted prevalence ratios (adjPR) and 95% confidence intervals (95% CI) estimated by multivariable Poisson regression.

RESULTS

The prevalence of RBC, plasma, and platelet transfusion was 5.8%, 0.9%, and 0.7%, respectively. RBC transfusions were associated with older age (≥ 65 vs. < 18 years; adjPR = 1.80; 95% CI = 1.66-1.96), female sex (adjPR = 1.13; 95% CI = 1.12-1.14), minority race/ethnic status, and hospitalizations in rural hospitals compared to urban teaching hospitals. Prevalence of RBC transfusion was lower among hospitalizations in the Midwest compared to the Northeast (adjPR = 0.73; 95% CI = 0.67-0.80). All components were more likely to be transfused in patients with a primary hematologic diagnosis, patients with a higher number of total diagnoses, patients who experienced a higher number of other procedures, and patients who eventually died in the hospital. In contrast to RBC transfusions, prevalence of platelet transfusion was greater in urban teaching hospitals (vs. rural; adjPR = 1.71; 95% CI = 1.49-1.98) and lower in blacks (vs. whites; adjPR = 0.80; 95% CI = 0.76-0.85).

CONCLUSIONS

Nationally, there is heterogeneity in factors associated with transfusion between each blood component, including by hospital type and location. This variability presents patient blood management programs with potential opportunities to reduce transfusions.

Bioelectrical Impedance Phase Angle-Predictor of Blood Transfusion in Cardiac Surgery

OBJECTIVE

To determine whether bioelectrical impedance-derived phase angle (PA) can be a predictor of red blood cell (RBC) transfusion in patients undergoing cardiac surgery.

DESIGN

An observational retrospective study of prospectively collected data.

SETTING

Single center, tertiary referral university hospital.

PARTICIPANTS

The study sample comprised 642 adult patients undergoing elective cardiac surgery.

INTERVENTIONS

Patient demographic and clinical variables were collected. The body composition of the patients was evaluated by bioelectrical impedance analysis (BIA) the day prior to surgery. The rates of postoperative RBC transfusion were recorded.

MEASUREMENTS AND MAIN RESULTS

Among the 642 patients (67.8% men, median age of 66 [range 59-73]) included in the present study, 210 (32.7%) received at least 1 RBC unit postoperatively. Hypertension, preoperative stroke, renal failure, preoperative hemoglobin and hematocrit values, BIA-derived PA, aortic crossclamp time, and cardiopulmonary bypass (CPB) time were associated with the risk of RBC transfusion in the univariate analysis, and were included in the final multivariate regression model. Preoperative stroke (odds ratio [OR] 0.394; 95% confidence interval [CI]: 0.183-0.848; p = 0.017), preoperative hemoglobin values (OR 0.943; 95% CI: 0.928-0.960; p < 0.001), PA <15th percentile (OR 2.326; 95% CI: 1.351-4.000; p = 0.002), and CPB time (OR 1.013; 95% CI: 1.008-1.018; p < 0.001) were identified as independent predictors of RBC transfusion.

CONCLUSION

Several factors were identified to be associated significantly with postoperative RBC transfusion in patients undergoing cardiac surgery. Among the conventional predictors, the value of the BIA-derived PA was indicated as a potent prognostic tool.

Demographic and epidemiologic characterization of transfusion recipients from four US regions: evidence from the REDS-III recipient database

BACKGROUND

Blood transfusion is one of the most common medical procedures during hospitalization in the United States. To understand the benefits of transfusion while mitigating potential risks, a multicenter database containing detailed information on transfusion incidence and recipient outcomes would facilitate research.

STUDY DESIGN AND METHODS

The Recipient Epidemiology and Donor Evaluation Study-III (REDS-III) program has developed a comprehensive transfusion recipient database utilizing data from hospital electronic health records at 12 participating hospitals in four geographic regions. Inpatient and outpatient data on transfusion recipients from January 1, 2013 to December 31, 2014 included patient age, sex, ethnicity, primary diagnosis, type of blood product provided, issue location, pretransfusion and post-transfusion hemoglobin (Hgb), and hospital outcomes. Transfusion incidence per encounter was calculated by blood product and various patient characteristics.

RESULTS

During the 2-year study period, 80,362 (12.5%) inpatient encounters involved transfusion. Among inpatients, the most commonly transfused blood products were red blood cells (RBCs; 10.9% of encounters), followed by platelets (3.2%) and plasma (2.9%). Among patients who received transfusions, the median number of RBC units was one, the pretransfusion Hgb level was 7.6 g/dL, and the Hgb increment per unit was 1.4 g/dL. Encounter mortality increased with patient age, the number of units transfused, and the use of platelet or plasma products. The most commonly reported transfusion reaction was febrile nonhemolytic.

CONCLUSION

The database contains comprehensive data regarding transfusion use and patient outcomes. The current report describes an evaluation of the first 2 years of a planned, 4-year, linked blood donor-component-recipient database, which represents a critical new resource for transfusion medicine researchers.

Variation in Use of Blood Transfusion in Primary Total Hip and Knee Arthroplasties

BACKGROUND

There is growing clinical and policy emphasis on minimizing transfusion use in elective joint arthroplasty, but little is known about the degree to which transfusion rates vary across US hospitals. This study aimed to assess hospital-level variation in use of allogeneic blood transfusion in patients undergoing elective joint arthroplasty and to characterize the extent to which variability is attributable to differences in patient and hospital characteristics.

METHODS

The study population included 228,316 patients undergoing total knee arthroplasty (TKA) at 922 hospitals and 88,081 patients undergoing total hip arthroplasty (THA) at 606 hospitals from January 1, 2009 to December 31, 2011 in the Nationwide Inpatient Sample database, a 20% stratified sample of US community hospitals.

RESULTS

The median hospital transfusion rates were 11.0% (interquartile range, 3.5%-18.5%) in TKA and 15.9% (interquartile range, 5.4%-26.2%) in THA. After fully adjusting for patient- and hospital-related factors using mixed-effects logistic regression models, the average predicted probability of blood transfusion use in TKA was 6.3%, with 95% of the hospitals having a predicted probability between 0.37% and 55%. For THA, the average predicted probability of blood transfusion use was 9.5%, with 95% of the hospitals having a predicted probability between 0.57% and 66%. Hospital transfusion rates were inversely associated with hospital procedure volume and directly associated with length of stay.

CONCLUSION

The use of blood transfusion in elective joint arthroplasty varied widely across US hospitals, largely independent of patient case-mix and hospital characteristics.

Protocol for a national blood transfusion data warehouse from donor to recipient

INTRODUCTION

Blood transfusion has health-related, economical and safety implications. In order to optimise the transfusion chain, comprehensive research data are needed. The Dutch Transfusion Data warehouse (DTD) project aims to establish a data warehouse where data from donors and transfusion recipients are linked. This paper describes the design of the data warehouse, challenges and illustrative applications.

STUDY DESIGN AND METHODS

Quantitative data on blood donors (eg, age, blood group, antibodies) and products (type of product, processing, storage time) are obtained from the national blood bank. These are linked to data on the transfusion recipients (eg, transfusions administered, patient diagnosis, surgical procedures, laboratory parameters), which are extracted from hospital electronic health records.

APPLICATIONS

Expected scientific contributions are illustrated for 4 applications: determine risk factors, predict blood use, benchmark blood use and optimise process efficiency. For each application, examples of research questions are given and analyses planned.

CONCLUSIONS

The DTD project aims to build a national, continuously updated transfusion data warehouse. These data have a wide range of applications, on the donor/production side, recipient studies on blood usage and benchmarking and donor-recipient studies, which ultimately can contribute to the efficiency and safety of blood transfusion.

Nomogram to predict perioperative blood transfusion for hepatopancreaticobiliary and colorectal surgery

Background

Predictive tools assessing risk of transfusion have not been evaluated extensively among patients undergoing complex gastrointestinal surgery. In this study preoperative variables associated with blood transfusion were incorporated into a nomogram to predict transfusion following hepatopancreaticobiliary (HPB) or colorectal surgery.

Methods

A nomogram to predict receipt of perioperative transfusion was developed using a cohort of patients who underwent HPB or colorectal surgery between January 2009 and December 2014. The discriminatory ability of the nomogram was tested using the area under the receiver operating characteristic (ROC) curve and internal validation performed via bootstrap resampling.

Results

Among 4961 patients undergoing either a HPB (56·3 per cent) or colorectal (43·7 per cent) resection, a total of 1549 received at least 1 unit of packed red blood cells, giving a perioperative transfusion rate of 31·2 per cent. On multivariable analysis, age 65 years and over (odds ratio (OR) 1·52), race (versus white: black, OR 1·58; Asian, OR 1·86), preoperative haemoglobin 8·0 g/dl or less (versus over 12·0 g/dl: OR 26·79), preoperative international normalized ratio more than 1·2 (OR 2·44), Charlson co-morbidity index score over 3 (OR 1·86) and procedure type (versus colonic surgery: major hepatectomy, OR 1·71; other pancreatectomy, OR 2·12; rectal surgery, OR 1·39; duodenopancreatectomy, OR 2·65) were associated with a significantly higher risk of transfusion and were included in the nomogram. A nomogram was constructed to predict transfusion using these seven variables. Discrimination and calibration of the nomogram revealed good predictive abilities (area under ROC curve 0·756).

Conclusion

The nomogram predicted blood transfusion in major HPB and colorectal surgery.

Opportunities and challenges in developing risk prediction models with electronic health records data: a systematic review

OBJECTIVE

Electronic health records (EHRs) are an increasingly common data source for clinical risk prediction, presenting both unique analytic opportunities and challenges. We sought to evaluate the current state of EHR based risk prediction modeling through a systematic review of clinical prediction studies using EHR data.

METHODS

We searched PubMed for articles that reported on the use of an EHR to develop a risk prediction model from 2009 to 2014. Articles were extracted by two reviewers, and we abstracted information on study design, use of EHR data, model building, and performance from each publication and supplementary documentation.

RESULTS

We identified 107 articles from 15 different countries. Studies were generally very large (median sample size = 26 100) and utilized a diverse array of predictors. Most used validation techniques (n = 94 of 107) and reported model coefficients for reproducibility (n = 83). However, studies did not fully leverage the breadth of EHR data, as they uncommonly used longitudinal information (n = 37) and employed relatively few predictor variables (median = 27 variables). Less than half of the studies were multicenter (n = 50) and only 26 performed validation across sites. Many studies did not fully address biases of EHR data such as missing data or loss to follow-up. Average c-statistics for different outcomes were: mortality (0.84), clinical prediction (0.83), hospitalization (0.71), and service utilization (0.71).

CONCLUSIONS

EHR data present both opportunities and challenges for clinical risk prediction. There is room for improvement in designing such studies.