A preoperative risk score for transfusion in infrarenal endovascular aneurysm repair to avoid type and cross

Machine-learning prediction models for any blood component transfusion in hospitalized dengue patients

BACKGROUND

Blood component transfusions are a common and often necessary medical practice during the epidemics of dengue. Transfusions are required for patients when they developed severe dengue fever or thrombocytopenia of 10×109/L or less. This study therefore investigated the risk factors, performance and effectiveness of eight different machine-learning algorithms to predict blood component transfusion requirements in confirmed dengue cases admitted to hospital. The objective was to study the risk factors that can help to predict blood component transfusion needs.

METHODS

Eight predictive models were developed based on retrospective data from a private group of hospitals in India. A python package SHAP (SHapley Additive exPlanations) was used to explain the output of the "XGBoost" model.

RESULTS

Sixteen vital variables were finally selected as having the most significant effects on blood component transfusion prediction. The XGBoost model presented significantly better predictive performance (area under the curve: 0.793; 95 % confidence interval: 0.699-0.795) than the other models.

CONCLUSION

Predictive modelling techniques can be utilized to streamline blood component preparation procedures and can help in the triage of high-risk patients and readiness of caregivers to provide blood component transfusions when required. This study demonstrates the potential of multilayer algorithms to reasonably predict any blood component transfusion needs which may help healthcare providers make more informed decisions regarding patient care.

A model based on electronic health records to predict transfusion events in on-pump cardiac surgery

Perioperative blood transfusion is costly and raises safety concerns. We developed and validated a model for predicting minor, moderate, or major transfusion given to patients during on-pump cardiac procedures based on two centers' database. Model performance incorporating 7 variables on the development set had an AUC of 0.803 [95% CI, 0.790-0.815] for minor transfusion; moderate transfusion, giving an AUC of 0.822 (95% CI, 0.803-0.841); and major transfusion, giving an AUC of 0.813 (95% CI, 0.759-0.866). Model performance on the validation set had an AUC of 0.739 (95% CI 0.714-0.765), 0.730 (95% CI 0.702-0.758), and 0.713 (95% CI 0.677-0.749), respectively. A model based entirely on readily available electronic health records can accurately predict intraoperative minor, moderate, or major transfusion and provide individualized transfusion risk profiles before surgery among those on-pump cardiac surgical patients, and may help guide patient management.

The Impact of Perioperative Morbidities, Lack of Discharge Aspirin, and Lack of Discharge Statin on Long Term Survival Following EVAR

OBJECTIVE

Adverse perioperative events and discharge medications both have the potential to impact survival following endovascular abdominal aortic aneurysm repair (EVAR). We hypothesize that variables such as blood loss, reoperation in the same hospital admission, and lack of discharge statin/aspirin have significant effect on long term survival following EVAR. Similarly, other perioperative morbidities, are hypothesized to affect long term mortality. Quantifying the mortality effect of perioperative events and treatment emphasizes to physicians the critical nature of preoperative optimization, case planning, operative execution and postoperative patient management.

METHODS

All EVAR in the Vascular Quality Initiative between 2003 and 2021 were queried. Exclusions were: ruptured/symptomatic aneurysm; concomitant renal artery or supra-renal intervention at the time of EVAR; conversion to open aneurysm repair at the time of initial operation; and undocumented mortality status at the 5 year mark postoperatively. 18,710 patients met inclusion criteria. Multivariable Cox regression time dependent analysis was performed to investigate the strength of mortality association of the exposure variables. Standard demographic variables and pre-existing major co-morbidities were included in the regression analysis to account for disproportionate, deleterious co-variables amongst those experiencing the various morbidities. Kaplan-Meier survival analysis was performed to provide survival curves for the key variables.

RESULTS

Mean follow up was 5.99 years and 5-year survival for included patients was 69.2%. Cox regression revealed increased long term mortality to be associated with the following perioperative events: reoperation during the index hospital admission (HR 1.21, P = .034), perioperative leg ischemia (HR 1.34, P = .014), perioperative acute renal insufficiency (HR 1.24, P = .013), perioperative myocardial infarction (HR 1.87, P < .001), perioperative intestinal ischemia (HR 2.13, P < .001), perioperative respiratory failure (HR 2.15, P < .001), lack of discharge aspirin (HR 1.26, P < .001), and lack of discharge statin (HR 1.26, P < .001). The following pre-existing co-morbidities correlated with increased long term mortality (P < .001 for all) : body mass index under 20 kg/m2, hypertension, diabetes, coronary artery disease, reported history congestive heart failure, chronic obstructive pulmonary disease, peripheral artery disease, advancing age, baseline renal insufficiency and left ventricular ejection fraction less than 50%. Females were more likely to have EBL >300 mL, reoperation, perioperative MI, limb ischemia and acute renal insufficiency than males (P < .01 for all). Female sex trended but was not associated with increased long term mortality risk (HR 1.06, 95% CI .995-1.14, P = .072).

CONCLUSIONS

Survival after EVAR is improved with optimal operative planning to facilitate evading the need for reoperation and ensuring patients without contra-indication are discharged with aspirin and statin medications. Females and patients with pre-existing co-morbidity are at particularly higher risk for perioperative limb ischemia, renal insufficiency, intestinal ischemia and myocardial ischemia necessitating appropriate preparation and preventative measures.

Drugs to reduce bleeding and transfusion in major open vascular or endovascular surgery: a systematic review and network meta-analysis

BACKGROUND

Vascular surgery may be followed by internal bleeding due to inadequate surgical haemostasis, abnormal clotting, or surgical complications. Bleeding ranges from minor, with no transfusion requirement, to massive, requiring multiple blood product transfusions. There are a number of drugs, given systemically or applied locally, which may reduce the need for blood transfusion.

OBJECTIVES

To assess the effectiveness and safety of anti-fibrinolytic and haemostatic drugs and agents in reducing bleeding and the need for blood transfusion in people undergoing major vascular surgery or vascular procedures with a risk of moderate or severe (> 500 mL) blood loss.

SEARCH METHODS

We searched: Cochrane Central Register of Controlled Trials; MEDLINE; Embase; CINAHL, and Transfusion Evidence Library. We also searched the WHO ICTRP and ClinicalTrials.gov trial registries for ongoing and unpublished trials. Searches used a combination of MeSH and free text terms from database inception to 31 March 2022, without restriction on language or publication status.

SELECTION CRITERIA

We included randomised controlled trials (RCTs) in adults of drug treatments to reduce bleeding due to major vascular surgery or vascular procedures with a risk of moderate or severe blood loss, which used placebo, usual care or another drug regimen as control.

DATA COLLECTION AND ANALYSIS

We used standard Cochrane methods. Our primary outcomes were units of red cells transfused and all-cause mortality. Our secondary outcomes included risk of receiving an allogeneic blood product, risk of reoperation or repeat procedure due to bleeding, risk of a thromboembolic event, risk of a serious adverse event and length of hospital stay. We used GRADE to assess certainty of evidence.

MAIN RESULTS

We included 22 RCTs with 3393 participants analysed, of which one RCT with 69 participants was reported only in abstract form, with no usable data. Seven RCTs evaluated systemic drug treatments (three aprotinin, two desmopressin, two tranexamic acid) and 15 RCTs evaluated topical drug treatments (drug-containing bioabsorbable dressings or glues), including fibrin, thrombin, collagen, gelatin, synthetic sealants and one investigational new agent. Most trials were conducted in high-income countries and the majority of the trials only included participants undergoing elective surgery. We also identified two ongoing RCTs. We were unable to perform the planned network meta-analysis due to the sparse reporting of outcomes relevant to this review. Systemic drug treatments We identified seven trials of three systemic drugs: aprotinin, desmopressin and tranexamic acid, all with placebo controls. The trials of aprotinin and desmopressin were small with very low-certainty evidence for all of our outcomes. Tranexamic acid versus placebo was the systemic drug comparison with the largest number of participants (2 trials; 1460 participants), both at low risk of bias. The largest of these included a total of 9535 individuals undergoing a number of different higher risk surgeries and reported limited information on the vascular subgroup (1399 participants). Neither trial reported the number of units of red cells transfused per participant up to 30 days. Three outcomes were associated with very low-certainty evidence due to the very wide confidence intervals (CIs) resulting from small study sizes and low number of events. These were: all-cause mortality up to 30 days; number of participants requiring an allogeneic blood transfusion up to 30 days; and risk of requiring a repeat procedure or operation due to bleeding. Tranexamic acid may have no effect on the risk of thromboembolic events up to 30 days (risk ratio (RR) 1.10, 95% CI 0.88 to 1.36; 1 trial, 1360 participants; low-certainty evidence due to imprecision). There is one large ongoing trial (8320 participants) comparing tranexamic acid versus placebo in people undergoing non-cardiac surgery who are at high risk of requiring a red cell transfusion. This aims to complete recruitment in April 2023. This trial has primary outcomes of proportion of participants transfused with red blood cells and incidence of venous thromboembolism (DVT or PE). Topical drug treatments Most trials of topical drug treatments were at high risk of bias due to their open-label design (compared with usual care, or liquids were compared with sponges). All of the trials were small, most were very small, and few reported clinically relevant outcomes in the postoperative period. Fibrin sealant versus usual care was the topical drug comparison with the largest number of participants (5 trials, 784 participants). The five trials that compared fibrin sealant with usual care were all at high risk of bias, due to the open-label trial design with no measures put in place to minimise reporting bias. All of the trials were funded by pharmaceutical companies. None of the five trials reported the number of red cells transfused per participant up to 30 days or the number of participants requiring an allogeneic blood transfusion up to 30 days. The other three outcomes were associated with very low-certainty evidence with wide confidence intervals due to small sample sizes and the low number of events, these were: all-cause mortality up to 30 days; risk of requiring a repeat procedure due to bleeding; and risk of thromboembolic disease up to 30 days. We identified one large trial (500 participants) comparing fibrin sealant versus usual care in participants undergoing abdominal aortic aneurysm repair, which has not yet started recruitment. This trial lists death due to arterial disease and reintervention rates as primary outcomes.

AUTHORS' CONCLUSIONS

Because of a lack of data, we are uncertain whether any systemic or topical treatments used to reduce bleeding due to major vascular surgery have an effect on: all-cause mortality up to 30 days; risk of requiring a repeat procedure or operation due to bleeding; number of red cells transfused per participant up to 30 days or the number of participants requiring an allogeneic blood transfusion up to 30 days. There may be no effect of tranexamic acid on the risk of thromboembolic events up to 30 days, this is important as there has been concern that this risk may be increased. Trials with sample size targets of thousands of participants and clinically relevant outcomes are needed, and we look forward to seeing the results of the ongoing trials in the future.

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.

Factors associated with red blood cell transfusion among hospitalized patients: a cross-sectional study

Red blood cell (RBC) transfusion is a clinically effective therapy in anemia, for example in patients with malignancies (Shander et al., 2020), bleeding (Odutayo et al., 2017), and preoperative anemia (Padmanabhan et al., 2019). The past few decades have witnessed a shortage of blood for transfusion due to limited health insurance coverage for blood use and the rapid expansion of hospitals (Chen et al., 2011; Shi et al., 2014). Blood donation levels may easily be affected by general changes in the environment, policy, major events such as disasters, and public sentiment (Hu et al., 2019). Meanwhile, the transfusion of allogeneic RBC is a double-edged sword, increasing the possibility of infectious and immunological complications, and also leading to higher morbidity and mortality after transfusion (Frank et al., 2012). Considering that the continual shortfall has been increasingly prominent, identifying the factors associated with RBC transfusion could help blood transfusion departments to improve their supply of blood products as well as their inventory management (O'Donnell et al., 2018).

Machine learning-based prediction of transfusion

Background

The ability to predict transfusions arising during hospital admission might enable economized blood supply management and might furthermore increase patient safety by ensuring a sufficient stock of red blood cells (RBCs) for a specific patient. We therefore investigated the precision of four different machine learning–based prediction algorithms to predict transfusion, massive transfusion, and the number of transfusions in patients admitted to a hospital.

Study Design and Methods

This was a retrospective, observational study in three adult tertiary care hospitals in Western Australia between January 2008 and June 2017. Primary outcome measures for the classification tasks were the area under the curve for the receiver operating characteristics curve, the F₁ score, and the average precision of the four machine learning algorithms used: neural networks (NNs), logistic regression (LR), random forests (RFs), and gradient boosting (GB) trees.

Results

Using our four predictive models, transfusion of at least 1 unit of RBCs could be predicted rather accurately (sensitivity for NN, LR, RF, and GB: 0.898, 0.894, 0.584, and 0.872, respectively; specificity: 0.958, 0.966, 0.964, 0.965). Using the four methods for prediction of massive transfusion was less successful (sensitivity for NN, LR, RF, and GB: 0.780, 0.721, 0.002, and 0.797, respectively; specificity: 0.994, 0.995, 0.993, 0.995). As a consequence, prediction of the total number of packed RBCs transfused was also rather inaccurate.

Conclusion

This study demonstrates that the necessity for intrahospital transfusion can be forecasted reliably, however the amount of RBC units transfused during a hospital stay is more difficult to predict.